Computer Vision Based Identification and Mosaic of Gramineous Grass Seeds Xin Pan Xuanhe Zhao Weihong Yan Jiangping Liu Xiaoling Luo Tana Wuyun

Computer Vision based Identification
and Mosaic of Gramineous Grass

Seeds Xin Pan Xuanhe Zhao Weihong
Yan Jiangping Liu Xiaoling Luo Tana
Wuyun
Visit to download the full and correct content document:
https://ebookmeta.com/product/computer-vision-based-identification-and-mosaic-of-gr
amineous-grass-seeds-xin-pan-xuanhe-zhao-weihong-yan-jiangping-liu-xiaoling-luo-t
ana-wuyun/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Computer Vision based Identification and Mosaic of

Gramineous Grass Seeds Xin Pan
https://ebookmeta.com/product/computer-vision-based-
identification-and-mosaic-of-gramineous-grass-seeds-xin-pan/
Autonomous Underwater Vehicles Jing Yan Xian Yang

Haiyan Zhao Xiaoyuan Luo Xinping Guan
https://ebookmeta.com/product/autonomous-underwater-vehicles-
jing-yan-xian-yang-haiyan-zhao-xiaoyuan-luo-xinping-guan/
Structural Reliability Approaches from Perspectives of

Statistical Moments 1st Edition Yan-Gang Zhao
https://ebookmeta.com/product/structural-reliability-approaches-
from-perspectives-of-statistical-moments-1st-edition-yan-gang-
zhao/
Multimedia Sensor Networks 1st Edition Huadong Ma Liang

Liu Hong Luo
https://ebookmeta.com/product/multimedia-sensor-networks-1st-
edition-huadong-ma-liang-liu-hong-luo/
Anglo Chinese Encounters Before the Opium War 1st
Edition Xin Liu
https://ebookmeta.com/product/anglo-chinese-encounters-before-
the-opium-war-1st-edition-xin-liu/
Advanced Methods and Deep Learning in Computer Vision

(Computer Vision and Pattern Recognition) 1st Edition
E. R. Davies
https://ebookmeta.com/product/advanced-methods-and-deep-learning-
in-computer-vision-computer-vision-and-pattern-recognition-1st-
edition-e-r-davies/
Computer Vision for Microscopy Image Analysis Computer

Vision and Pattern Recognition 1st Edition Mei Chen
Ph.D (Editor)
https://ebookmeta.com/product/computer-vision-for-microscopy-
image-analysis-computer-vision-and-pattern-recognition-1st-
edition-mei-chen-ph-d-editor/
Computer Vision: Object Detection In Adversarial Vision

1st Edition Bhowmik
https://ebookmeta.com/product/computer-vision-object-detection-
in-adversarial-vision-1st-edition-bhowmik/
Design and Applications of Emerging Computer Systems

1st Edition Weiqiang Liu
https://ebookmeta.com/product/design-and-applications-of-
emerging-computer-systems-1st-edition-weiqiang-liu/
Xin Pan, Xuanhe Zhao, Weihong Yan, Jiangping Liu, Xiaoling Luo and
Tana Wuyun
Computer Vision based Identification

and Mosaic of Gramineous Grass Seeds
1st ed. 2021
Xin Pan
College of Computer and Information Engineering, Inner Mongolia
Agricultural University, Inner Mongolia, China
Xuanhe Zhao
Inner Mongolia Agricultural University, Inner Mongolia, China
Weihong Yan
Grassland Research Institute, Chinese Academy of Agricultural Sciences,
Inner Mongolia, China
Jiangping Liu
College of Computer and Information Engineering, Inner Mongolia
Xiaoling Luo
Inner Mongolia Agricultural University, Inner Mongolia, China
Tana Wuyun
College of Mechanical and Electrical Engineering, Inner Mongolia
ISBN 978-981-16-3500-7 e-ISBN 978-981-16-3501-4

https://doi.org/10.1007/978-981-16-3501-4
Jointly co-published with China Agricultural Science and Technology

Press, Beijing, China
The print edition is not for sale in China (Mainland). Customers from
China (Mainland) please order the print book from: China Agricultural
Science and Technology Press
Jointly co-published with China Agricultural Science and Technology

Press, Beijing, China. The print edition is not for sale in China
(Mainland). Customers from China (Mainland) please order the print
book from: China Agricultural Science and Technology Press
© China Agricultural Science and Technology Press 2021
This work is subject to copyright. All rights are solely and exclusively
licensed by the Publisher, whether the whole or part of the material is
concerned, specifically the rights of reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other
physical way, and transmission or information storage and retrieval,
electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks,

service marks, etc. in this publication does not imply, even in the
absence of a specific statement, that such names are exempt from the
relevant protective laws and regulations and therefore free for general
use.
The publisher, the authors and the editors are safe to assume that the
advice and information in this book are believed to be true and accurate
at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, expressed or implied, with respect to the
material contained herein or for any errors or omissions that may have
been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer

Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04
Gateway East, Singapore 189721, Singapore
Preface
Grassland is the refreshable natural resource for human beings, and the
dynamic monitor and digitalized government of grassland are vital for
the sustainable strategy. At present, the research on the grassland
digitalization is rare. Hence, it is of great importance in theory and
reality to realize grass recognition by computer vision, which will make
contribution to improve the accuracy of grass auto recognition, and
data acquisition.
Computer vision is a multi-disciplinary technology that takes
advantage of computer and image acquisition system to imitate the
human vision in converting image into digital signal. Different from the
remote sensing technology-based digital system of the grassland,
computer vision emphasizes on digital images of grassland and forage
captured by digital cameras. Hence, the book focused on intrinsic
feature extraction, so as to realize auto recognition of forage, and
microscope images mosaic. The computer vision-based digital platform
of grassland and forage was developed by integrating the images
mosaic and some related function modules.
The book is an initial effort in the research of grassland digitization
based on computer vision, especially focuses on the identification and
mosaic of gramineous grass seeds. It is a general research report of our
research team in Inner Mongolia Agricultural University and Grassland
Research Institute of Chinese Academy of Agricultural Sciences, which
aggregates our collective intelligence and efforts in the research. The
book was written by Pan Xin, Gao Xiaojing, Liu Jiangping, Wuyuntana,
Ma Yubao, and Yan Weihong. Among them, Yan Weihong and Ma Yubao
are from Grassland Research Institute of Chinese Academy of
Agricultural Sciences, and the other team members are from Inner
Mongolia Agricultural University. The final compilation and editing of
the whole book was completed by Pan Xin.
This work is dedicated to Researcher Zhu Xu and Dr. Feng Hao for
their great contributions. We are also inspired by their deep love of the
grassland, which has become the eternal part of their lives. We owe
gratitudes to Researcher Liu Guixiang, the Doctoral supervisor in
Grassland Research Institute of Chinese Academy of Agricultural
Sciences. Thanks to the help and advices from Professor Ma Shuoshi,
Professor Xue Heru, Professor Ding Xuehua, Professor Han Guodong,
Professor Wei Zhijun, Professor Liu Zhongling, Professor Luo Xiaoling,
and Associate Professor Jiang Xinhua. Millions of the researchers and
herders work and live in the magnificent grassland, protecting our
beautiful home. We are very proud to make all our efforts to be one of
them and does something minor to their great career. In addition, we
are also grateful at the contributions of all graduates, Chen Tong, Ning
Lina, Zhao Xuanhe, Duan Junjie, Zhang Jianhua, Zhai Lin, Han Fan, Zhao
Dandan, Cen Yao, Xiang Yang, Wu Lin, Peng Jingjing, Wang Xinyu, Qiao
Xue, Guo Yubo, and Dai Xuemin.
This work is supported by the National Natural Science Foundation
of China under Grant No. 61562067 and No. 61962048, Central Public-
Interest Scientific Institution Basal Research Fund (No
1610332020020) the Postdoctoral Science Foundation of China No.
20100480370 and No. 201104179, the Regional Natural Science
Foundation of Inner Mongolia under Grant No. 2012MS919 and No.
2012MS915, the Basic Scientific Research Foundation for Central Public
Research Institutes No. 1610332015007, and the Foundations of Inner
Mongolia Agricultural University NDPYTD 210-9, BJ09-43, and
JCYJ201201.
The research on the grassland digitization based on computer
vision is still at the very beginning stage, where there is a long way to
explore in future. More researchers are required to get involved in the
work. Limited by the professional knowledge in multi-discipline and
scientific vision scope, mistakes are inevitable in the book, and we are
grateful to the comments and corrections from the readers.
Xin Pan
Inner Mongolia, China
Contents
1 Introduction
1.1 Digitization Grassland
1.2 Computer Vision Based Research Status of Grassland
Digitization
1.2.1 Grassland Digitization Based on Remote Sensing
1.2.2 Grassland Digitization Based on Common Digital
Images
1.2.3 Researches on the Related Fields
1.3 Research Content
References
2 Forage Identification and Experimental Materials
2.1 Definition of Forage Automated Identification
2.2 Automatic Identification of Herbage Based on Seed Images
2.3 Experimental Materials
2.3.1 Image Acquisition
2.3.2 Image Preprocessing
2.4 Seed Image DatabaseI
References
3 Identification of Gramineous Grass Seeds Using Gabor and
Locality Preserving Projections
3.1 Feature Extraction
3.1.1 Gabor Filters
3.1.2 Locality Preserving Projections
3.2 Feature Matching
3.3 Results and Discussion
3.3.1 Experiments on Seed Image DatabaseI
3.3.2 Experiments on Seed Image DatabaseII
3.4 Summary
References
4 Identification of Gramineous Grass Seeds Using Difference of
Local Fractal Dimensions
4.1 Fractal and Fractal Dimension
4.2 Difference of Local Fractal Dimension
4.3 Results and Discussions
4.3.1 Local Division Strategy
4.3.2 Comparison of Different Seed Identification
Algorithms
4.4 Summary
References
5 Identification of Gramineous Grass Seeds Using Local Similarity
Pattern and Linear Discriminant Analysis
5.1 Identification Based on LSP and LDA
5.2 Experimental Results and Discussions
5.2.1 Image Database and Preprocessing
5.2.2 Identification Experiments
5.3 Summary
References
6 Identification of Gramineous Grass Seeds Using Local Similarity
Pattern and Gray Level Co-occurrence Matrix
6.1 Fusion of LSP and GLCM in Feature Extraction
6.1.1 Gray Level Co-occurrence Matrix
6.1.2 Fusion of LSP and GLCM for feature extraction
References
7 Microscopic Image Mosaic of Gramineous Grass Seeds
7.1 Image Mosaic
7.2 Materials and Methods
7.2.1 Experimental Materials
7.2.2 Mosaic Method
7.4 Summary
References
8 Digital Information Platform of Grassland and Forage Based on
Computer Vision
8.1 Digital Information Platform of Grassland and Forage
8.1.1 Main Development Tools and Models for Platform
8.1.2 Main Function Modules of the Platform
8.2 Key Technological Problems and Solutions of the Platform
8.2.1 Hybrid Programming of Java and Matlab
8.2.2 Design and Implementation of Module Interface
8.2.3 Design and Implementation of GUI for Image Mosaic
8.3 Implementation of the Platform
8.4 Summary
References
Appendix:Gramineous Grass
X. Pan et al., Computer Vision based Identification and Mosaic of Gramineous Grass
Seeds
https://doi.org/10.1007/978-981-16-3501-4_1
1. Introduction
Xin Pan1 , Xuanhe Zhao2 , Weihong Yan3 , Jiangping Liu4 ,
Xiaoling Luo5 and Tana Wuyun6
(1) College of Computer and Information Engineering, Inner Mongolia
(2) Inner Mongolia Agricultural University, Inner Mongolia, China
(3) Grassland Research Institute, Chinese Academy of Agricultural
Sciences, Inner Mongolia, China
(6) College of Mechanical and Electrical Engineering, Inner Mongolia
Xin Pan (Corresponding author)

Email: 1565083121@qq.com
Xuanhe Zhao
Weihong Yan
Email: yanweihong7037@163.com
Jiangping Liu
Email: liujiangping@imau.edu.cn
Xiaoling Luo
Email: luoxl@imau.edu.cn
Tana Wuyun
Email: wytn@imau.edu.cn
1.1 Digitization Grassland

Grassland is the refreshable natural resource for human beings and
provides the basic materials and life guarantee for the rational and
sustainable use of human beings. The unique geographical unit and
ecological environment have shaped the diversity of rich species and
become one of the most important living environments for human
beings. However, the degradation situation of grassland in China is very
astonishing because of many factors, such as long-term overgrazing,
extensive management, climate variation, etc.. At present, the grassland
degradation has caused grass yield decline, dust storms, soil erosion
and other ecological environmental problems (Tang et al. 2009). The
state has attaches great importance to the maintenance of grassland
ecosystem, but the mere traditional measurements for grassland
monitoring cannot effectively solve ecological environmental problems
such as grassland desertification (Wu 2003). Therefore, the dynamic
monitoring and digitalization of grassland industry are the basic ways
to rationally protect and utilize grassland resources and ecological
environment, and further to realize the national sustainable
development strategy.
Digital prataculture is a process of realizing digital and visual
expression, design, management and control of the relevant objects and
the whole process involved in grass industry by using digital
technology. It is an important branch of digital agriculture. Within the
basic framework of standards and specifications of digital agriculture, it
is necessary to construct a digital modern management technology and
the theoretical framework, including technologies of computer,
network communication, space information, automation and basic
subjects of pratacultural science, geography, ecology, to detect, manage
and control the elements of grassland ecosystem (environmental
factors, biological factors and economic factors, etc.) and vital process.
The digital prataculture can macroscopically monitor and evaluate the
ecology and economy of regional grass industry, and provide decision
support for the sustainable development of grass industry. At a micro
level, digital management and control of grassland production process
can be carried out by the expert system and knowledge engineering
construction, so as to optimize production input, yield and benefit to
the maximum extent (Tang et al. 2009). The digitization of grassland is
the realized form of digital prataculture. It is the process to store,
transmit process and represent the related grassland information in
digitized form, and establish the appropriate digital system according
to the data models to monitor and manage grassland. Grassland
digitization is an emerging research direction to apply information
technology into grassland ecosystem management, which has potential
values in both theoretical research and practical application.
At present, researches on grassland digitization mainly focus on
three aspects: simulation technology of grassland ecosystem process,
dynamic monitoring of grassland ecosystem and expert system and
decision support of prataculture management (Tang et al. 2009).
Images are important digital resources of grassland, which can be
generally divided into two categories according to the acquisition
means. One is the macroscopic hyperspectral images and remote
sensing images obtained by satellites and spectrometers. The other is
commonly visible images obtained by ordinary digital cameras.
Remote sensing images can be obtained by using sensors carried on
satellite to detect electromagnetic wave radiation information. Remote
sensing information is widely used in the study of grassland dynamic
monitoring of on account of the merits
in its cyclical, macro and systematic aspects, which can be used to
monitor the land surface from a macro perspective (Xiao and Zhang
2008) in a medium and long distance. Most of the investigations are
based on remote sensing images at a large scale, such as detecting
grassland variations (Lausch et al. 2013; Li et al. 2014), and biomass
estimation (Potter 2014). However, the overall application of remote
sensing cannot meet the practical requirements, such as the
synchronous, quasi-synchronous data acquisition, analysis in real-time
monitoring and data processing capacity, and automatic recognition of
remote sensing images, etc. (Zhao 2003). In addition, the remote
sensing images still have limitations in the following three aspects.
First, the acquisition cost. Owing to the images are mainly collected by
satellite from a long distance, the image number, popularity can not be
compared with ordinary digital images directly collected by digital
cameras by the researchers. The interpretability of satellite based
remote sensing is still a challenge for common investigators. Secondly,
the large scale. Generally, remote sensing images fail to capture
grassland information from a microscopic perspective, such as
individual plant growth status of herbage, composition of herbage
species in community landscape, etc.. A comprehensive grassland
digital system requires both remote sensing images obtained by
satellite and oridianry images by digital cameras. Finally, redundant
information. Remote sensing image has a large amount of data and a
large amount of redundant information, restricted in data storage,
processing, transmission, display and other links. Therefore, the
digitization of grassland merely dependent on remote sensing
technology is incomplete and one-sidedness.
Digital camera can capture the herbage from grassland from a
micro-scale. As the basis and important part of grassland resources,
forage plays an important role in providing grassland animal
husbandry, environmental protection and soil conservation (Liu et al.
2009). With the popularity of digital cameras, numerous grass images
can be easily captured and transferred to computers. Hence, herbage
images are as well crucial components of grassland image database and
digital grassland. Some of the images can be used for further
identification of grass and forage with the help of computer vision.
Computer vision technology is an interdisciplinary technology that
utilizes computer and image acquisition equipment to simulate human
vision acquisition images and convert them to digital signals, and
realizes image transmission, processing and understanding and other
visual information processing through computer. Image recognition can
be realized for automatic identification and classification by the stages
of preprocessing feature extraction and feature matching, which form a
whole process of pattern recognition. These images are mainly forage
plant and community landscape images. Computer vision technologies
such as digital image processing and pattern recognition can be used to
accurately judge forage species and community component analysis
from micro scale, calculate corresponding grassland biological
information, and integrate a complete grassland digital system with
macro-scale remote sensing images.
Different from the remote sensing technology of geographic
information system, computer vision technology mainly studies the
digital image acquired by digital camera, which can obtain the required
information from a micro scale. This kind of images can be directly
collected by digital camera, and then transmitted to the computer for
storage and processing, which reduces the error and improves the
efficiency of data collection and analysis in grassland field. With the
rapid development of image technology, especially the widespread use
of digital cameras, researchers have acquired a large number of
grassland herbage pictures, which provides a research foundation for
the comprehensive digitalization of grassland herbage.
Forage identification is not a simple task due to great complexity
among different varieties, which belongs to plant taxonomy, a highly
laborious task, consisting of scientific classification of our planet’s flora
(Yanikoglu et al. 2014).
Therefore, this book focuses on the study of grassland herbage
images to extract the characteristics of herbage image with the help of
computer vision technology, so as to realize the functions of automatic
classification and recognition of grassland forage, and microscopic
images mosaic. On the above basis, a digital platform based on
computer vision technology was developed to provide new ideas for
automatic data acquisition of grassland digitization.
1.2 Computer Vision Based Research Status of

Grassland Digitization
In general, there are relatively few studies on digital grassland and
forage using image recognition technology, and the methods are
relatively simple. Therefore, in this chapter, the main related researches
are categorized into 3 types. The first is about remote sensing images—
the focus of the studies at present, the second is about the common
grassland images captured by CCD digital camera, and the third is about
the application in precise agriculture using computer vision.
1.2.1 Grassland Digitization Based on Remote

Sensing
Most researchers focus on the processing and understanding of remote
sensing images, taking advantages of remote sensing images and
geographic information technology to investigate and dynamic monitor
grassland resources and environment, such as estimation of grassland
biomass, early warning of grassland diseases and pests, and monitoring
of grassland degradation. (Wu 2003; Zhang 2007; Qian and Ma 2009).
Among them, the Grassland Research Institute of Chinese Academy of
Agricultural Sciences started the work earlier, accumulated a lot of
research experience. During the 8th and 9th “Five-Year Plan” period,
they have undertaken some important research projects, such as
National Key Projects, Ministry of Agriculture and National Natural
Foundation Project, such as “Research on the Application of Remote
Sensing in the Resources Survey of Inner Mongolia Grassland”, “A
comprehensive investigation and research of Inner Mongolia grassland
pasture on protection forest based on remote sensing”, and “the
research on dynamic monitoring technology of breeding balance
between forage and livestock in grassland northern China”, etc.. During
the 10th “Five-Year Plan” period, the first information professional
network system of Chinese grassland resources based on 3S technology,
www.grassland.net.cn, has been established in China, with which the
queries of 10 ecological information items in forage natural distribution
areas. The above works make a frame for digitized grassland.
1.2.2 Grassland Digitization Based on Common

Digital Images
However, few researches on grassland digitization based on ordinary
visible images, mainly focus on image-based grassland vegetation
coverage. For example, Chi et al. (2007) measured the coverage of
grassland vegetation by taking close-up images of grassland vegetation
with a camera, and measured the coverage of grassland vegetation
through 4 steps: data acquisition, image processing, number of pixels
extraction and coverage calculation. Compared with traditional
coverage measurement methods, it has the advantages of objectivity,
consistency and high efficiency and accuracy.In the studies of the
impact of grazing on grassland plant communities (Zhao et al. 2010),
Cannon PC1057 digital camera was used to capture the images of
sample plots under different grazing intensity. The specific method was
to divide each area of 3 m × 100 m into 100 consecutive 1.5 m × 2 m,
and take 100 consecutive vertical projection photos of each rectangle
from a height of 2.5 m. After cutting the photos along the edge of the
quadrat with Photoshop, the images of the sample belt were accurately
splicing together. Then, the geographic information system software
ArcGIS9 was used to make layers of the patches and plant communities
in the photos and finally the patch area, plant composition, biomass,
density and other indicators under different strengths of the
corresponding quadrat can be measured. All above studies show that
the grassland images captured by ordinary digital cameras can be used
to analyze grassland data, which is conducive to reducing the cost of
manual data collection and improving the data accuracy.
To date, only a few researches have demonstrated the feasibility of
forage identification. For example, Wang et al. (2010) classified
leguminous forage based on shape features of leaf images. The global
shape features include the axis ratio, rectangularity, and 7 invariant
moments of leaf images. The roughness of the leaf edges are extracted
as the local features. Both global and local features are used as the input
to probabilistic neural network (PNN) and back propagation network
(BP) for classification, yielding correct classification rates of 85% and
82.4%, respectively, in a database of 560 training samples and 1,400
testing samples comprising 14 species.
1.2.3 Researches on the Related Fields

Despite rare research conducted in grass identification, most relevant
research on classification of weeds and cultivated crops in precision
agriculture provide us some valuable experiences for forage
classification. Precision agriculture has been going on for at least
20 years, aiming at the improvement of crop productivity. In
comparison, computer vision technology based on visible images has
been widely studied and applied in the field of precision agriculture,
such as weed identification, pest control, crop growth control, quality
monitoring of agricultural products, vegetation coverage monitoring,
etc. (Burks et al. 2000; Lv et al. 2001; Ma et al. 2009; Liu 2010), which
has the advantages of low cost, high efficiency and high precision.
Natural weed images are used for weed detection and classification
to discriminate crops from the weeds (Ishak et al. 2009; Evert et al.
2009) for a reduction in herbicide use and environmental pollution
(Burgos-Artizzu et al. 2010). Gerhards and Christensen (2003)
proposed a system of site-specific weed control in sugarbeet, winter
wheat and winter barley. Bi-spectral images were taken and converted
into binary ones to get the contour of the weeds. Color combined with
morphological filtering and line fitting were used to segment row crop
plants from weeds, yielding highest and lowest classification rates of
92% and 68% over 12 images, respectively (Onyango and Marchant
2003). Meyer and Neto (2008) improved the vegetation index using
Excess Green minus Excess Red, instead of commonly used Excess
Green, resulting in a higher precision in normalized difference indices
(NDI). Burgos-Artizzu et al. (2010) processed images captured in the
fields for the estimation of percentage of weed, crop and soil. Tellaeche
et al. (2011) developed an automatic computer vision system for the
identification of avena sterilis (a special weed seed growing in cereal
crops). Wang et al. (2013) proposed an adaptive thresholding algorithm
using OTSU and CANNY operators to segment single leaves from leaf
images captured from an online system. Yanikoglu et al. (2014)
proposed an automatic plant identification system using a rich variety
of shape, texture and color features, yielding 61 and 81% accuracies in
classifying the 126 different plant species in the top-1 and top-5
choices.
Liu (2010) developed a prototype of rice leaf disease diagnosis
system based on image pattern recognition. The recognition rate of 4
kinds of rice leaf diseases and pests can reach more than 83%. Han et
al. (2011) put forward pests classification method based on
compression perception, extracted by using pest image training area/
standard product /perimeter and 10 morphological parameters, such
as red, green, blue, bright, 4 color characteristics of the training sample
space of 12 classes of pests and 110 common pests classification. In
four different experimental conditions, the correct classification rate of
92.94%, 98.29%, 78.87% and 61.59%, respectively.
Yu (2007) extracted the features of the regional gray level from both
sides of cucumber leaf main veins. They established the regression
model, from the point of experiment to determine a cucumber point
water shortage area.
Li (2009) used machine vision technology to quantitatively evaluate
the appearance quality of tea image extraction, and proposed for the
first time the classification method of tea based on different particle
distribution states. The images based on granular stacking can reflect
the geometric features and macroscopic texture features of tea leaves,
and improve the efficiency of tea leaf differentiation. 18 shape features
and 15 texture features were selected in the study, and the recognition
accuracy was 93.8% higher.
Zhang et al. (2009) developed a portable vegetation coverage
photography instrument with digital camera as the main collection
equipment based on the actual field conditions, and developed the
corresponding automatic calculation software to interpret and estimate
vegetation coverage.
Owing to seeds are more stable and unlikely to be easily affected by
their surroundings, Granitto et al. (2002) assessed the discriminating
power of 57 weed seed species. The 12 feature vectors are composed of
six morphological, four color and two textural seed characteristics.
Bayes Classifier was testified surprisingly good for the classification
performance. In their later research (Granitto et al. 2005), a large scale
seed base containing 10,310 images of 236 different weed species were
used to testify the effectiveness of the algorithm, the recognition rates
still reached 99.3% and 98.2% when color images and black and white
seed images are testified, respectively.
Shi and Ji (2009a, b) proposed a seed identification system for
leguminous weeds. In their work, a seed database of 5,181 microscope
images comprising 808 species was constructed. The feature vector
composed of 16 components includes the shape geometry and the
inner structures of the seeds and umbilicus. The concrete features
include the major axis, surface area, perimeter, position and angle
between the center points, and 7 Hu invariant moments. The BP
Network and SVM classifier are used for classification, yielding a
recognition accuracy of 89.29%.
Similarly, identification of nine Iranian wheat seed varieties were
conducted on bulk sample images by Pourreza et al. (2012) by
extracting 131 textural features from gray level, GLCM (gray level co-
occurrence matrix), GLRM (gray level run-length matrix),LBP(local
binary patterns), LSP (local similarity patterns) and LSN (local
similarity numbers) matrix. LDA (linear discriminant analysis)
classifier was employed for classification using top selected features,
yielding an average classification accuracy of 98.15% with top 50
selected features. Hong et al. (2015) compared some simple feature
extraction techniques and obtained an average accuracy of 90.45%
using Random Forest method for the identification of six different rice
seed varieties in Vietnam.
It can be seen from the above comparison that visible images based
on digital cameras already have a certain foundation in precision
agriculture. Nevertheless, in grassland digitization, the overall research
lags behind. Most of the micro-scale data collection and measurement
of grassland mainly rely on manual completion, with a low automation
and subjective observation results. It is urgent to apply image
processing and pattern recognition technology to the research of grass
industry, especially the research of grass digitization, so as to effectively
use information resources and improve the level and efficiency of grass
digitization management.
The grassland digitization research based on computer vision can
draw on some successful experiences of precision agriculture
application, such as the classification of herbage and weed
identification, which are very similar in species and traits. The above
works mainly focused on the leaves of the weeds and plants. Although
forage identification is similar to weed identification in some ways, they
differ from each other at least in two aspects, research purpose and
identification methods. The former aims at the determination of
grassland species for research purpose and further environment
protection, while the latter is to remove the weed for less pesticide
usage and more grain productivity. So the grass identification mostly
focuses on the closely related species of the same families which are
similar in shape and appearance, providing valuable information the
grassland researchers and farmers. However, weed identification in
precision agriculture does not necessarily concentrate on closely
related species in the same family. They can be a very large varieties of
species widely distributed in any fields where they affect the grain and
crop growth whose visual features differ more obviously than forage
identification. Due to the different research purposes, weed
identification is mainly used to distinguish between weeds and crop,
differences between different weeds, weeds and crops are relatively
obvious; the grass identification mainly takes the dominant species on
the grassland as the research object, to distinguish the different forage
species. Especially the different species in the same family are highly
identical, which is sometimes difficult for the experts to distinguish.
Therefore, this book mainly uses computer vision technology to
extract information of grassland forage image, so as to obtain relevant
expert knowledge information of the forage species, and finally
establish digital platform of grassland forage based on image
recognition technology to improve the level of grassland digital
management.
1.3 Research Content

Computer vision based grassland digitization is only in its early stage,
and a lot of work needs to be done in theory and practice. Therefore,
the research mainly focuses on the visible images obtained by digital
cameras, and focuses on the forage identification based on seed images,
image Mosaic and the construction of digital platform for gramineous
grass in grassland. Among them, the realization of forage automatic
classification is the basic element of forage digitization research. Image
Mosaic technique is required in practical cases because the limitations
by the visual range of camera lens or microscope, multiple images are
fused to obtain the overall visual understanding for further studies on
forage traits. Finally, on the basis of previous studies, the relevant
research results of grassland forage have been integrated into a
platform for grassland scientific research workers and herdsmen,
named Digital Information Platform of Grassland and Forage. Image
mosaic can be realized on the platform, which is the highlight of the
platform.
This paper is divided into 8 chapters, and the specific content is
arranged as follows.
Chapter 1 is introduction. In this chapter, we introduce the
background and significance of grassland digitization research based
on computer vision technology. Then, the research status and
applications of image recognition technology in digital grass industry
and precision agriculture are emphatically compared and elaborated, to
clarify the important role of image recognition technology in grass
industry science. Finally, the main architecture of the book content is
listed.
Chapter 2 introduces the concept of forage identification in our
research and the image database for experiments.
Chapter 3 presents the identification method of gramineous grass
seeds using Gabor wavelets and Local Preserving Projections for
intrinsic manifold features. Chapter 4 presents the identification
method of gramineous grass seeds using the deviations of local fractal
dimensions based on the self-similarity of the seeds. Chapter 5 presents
the identification method of gramineous grass seeds using local
similarity pattern (LSP) and linear discriminant analysis (LDA). Chapter
6 presents the identification method of gramineous grass seeds using
local similarity pattern (LSP) and GLCM, which is an improved method
of Chapter 5.
Chapter 7 presents the mosaic method of microscopic image of
gramineous seeds. The whole procedure of image mosaic
implementation, including image acquisition, preprocessing, feature
extraction and feature fusion are introduced respectively.
Chapter 8 introduces the realization process of digital platform
based on computer vision. The main contents include the overall
framework of the platform, the key technical problems and solutions of
the platform, and the implementation effect of the platform.
References
Burks, T. F., S. A. Shearer, R. S. Gates. 2000. Classification of weed species using color
texture features and discriminant analysis[J].Transaction of the ASAE, 43(2): 441–
448.
Burgos-Artizzu, X. P., A. Ribeiro, A. Tellaeche, G. Pajares, and C. Fernández-Quintanilla.

2010. Analysis of natural images processing for the extraction of agricultural
elements[J]. Image Vision Computer. 28: 138–149.
Chi, H. K., G. S. Guang, Z. Z. Xu, and W. P. Yuan. 2007. Measuring coverage of grassland
vegetation using remote sensing over short distances[J]. Acta Prataculture Sinica.
16(2): 105–110. (in Chinese).
Gerhards, R., and S. Christensen. 2003. Real-time weed detection, decision making
and patch spraying in maize, sugarbeet, winter wheat and winter barley[J]. Weed Res.
43, 385–392.
Granitto, P. M., H. D. Navone, P. F. Verdes, and H. A. Ceccatto. 2002. Weed seeds

identification by machine vision[J]. Computers and Electronics in Agriculture. 33,
91–103.
Granitto, P. M., P. F. Verdes, and H. A. Ceccatto. 2005. Large-scale investigation of weed

seed identification by machine vision[J]. Computers and Electronics in Agriculture.
47, 15–24.
Han A. T, X. H. Guo, Z Liao, Z. Q. Chen, and J. Q. Han. 2011. Classification of agricultural

pests based on compressed sensing theory[J], Transactions of the Chinese Society of
Agricultural Engineering. 27(6): 203–207. (in Chinese).
Hong, P., T. Hai, L. Lan, V. Hoang, V. Hai, and T. Nguyen. 2015. Comparative study on
vision based rice seed varieties identification[C], 7th International Conference on
Knowledge and Systems Engineering. 377–382.
Ishak, A. J., A. Hussain, and M. Mustafa. 2009, Weed image classification using Gabor
wavelet and gradient field distribution[J]. Computers and Electronics in Agriculture,
66, 53–61.
Lausch, A., M. Pause, I. Merbach, S. Zacharias, D. Doktor, and M. Volk, R. Seppelt. 2013.
A new multiscale approach for monitoring vegetation using remote sensing-based
indicators in laboratory, field, and landscape[J]. Environmental Monitoring and
Assessment, 185(2): 1215–1235.
Li, X. L. 2009. Reasearch on nondestructive determination of tea quality based on

machine vision and spectroscopy techniques[D]. Hangzhou: Zhejiang Uninversity. (in
Chinese).
Li, Q., D. W. Zhou, Y. H. Jin, M. L. Wang, Y. T. Song, and G. D. Li. 2014. Effects of fencing
on vegetation and soil restoration in a degraded alkaline grassland in northeast
China[J]. J. Arid Land, 6(4):478-487.
Liu, L. B. 2010.Research of diagnostic technologies for rice leaf diseases based on

image[D].Beijing: Chinese Academy of Agricultural Sciences.(in Chinese).
Liu L. X., G. M. Li, Y. X. Lu, D. L. Wang, 2009. Current status and prospect in the
biotechnology of forage breeding in China[J]. Acta Agrectir Sinica, (3): 389-398. (in
Chinese).
Lv C. H., X. G. Chen, W. F. Wu, H. X. Zhao. 2001. The technique of detecting and
identifying field plants by machine vision[J]. Natural Science Journal of Jilin
University of Technology, 31( 3): 90–94. (in Chinese).
Ma Y. P., Y. L. Bai, X. Z. Gao, Y. Lu , W. Lei, and J. Du. 2009. Application and expectation
of computer vision technology in agriculture production[J]. Chinese Journal of
Agricultural Resources and Regional Planning. 30(4): 21–27. (in Chinese).
Meyer, G., and J. Neto. 2008. Verification of color vegetation indices for automated
crop imaging applications[J]. Comput. Electron. Agr. 63, 282–293.
Onyango, C., and J. Marchant. 2003. Segmentation of row crop plants from weeds
using colour and morphology[J]. Comput. Electron. Agr. 39, 141–155.
Potter, C. 2014. Monitoring the production of Central California coastal rangelands

using satellite remote sensing[J] . Journal of Coastal Conservation, 18(3): 213–220.
Pourreza, A., H. Pourreza, M. Abbaspour-Fard, and H. Sadrnia. 2012. Identification of

nine Iranian wheat seed varieties by textual analysis with image processing[J].
Comput. Electron. Agr. 83, 102–108.
Qian J. B., and M. G. Ma. 2009. A review of poisonous weeds detection using remote
sensing technology[J]. Remote Sensing Technology and Application. 24(5): 685–690.
(in Chinese).
Shi C. J., and G. R. Ji. 2009a. Study of recognition method of leguminous weed seeds
image[J]. International Workshop on Intelligent Systems and Applications, Seoul,
Korea, 1–4.
Shi, C., and G. Ji. 2009b. Study of recognition method of leguminous weed seeds
image[C]. in: Proceedings of International Workshop on Intelligent Systems and
Applications, 1–4. (in Chinese).
Tang H. J., X. P. Xin, G. X. Yang, B. H. Zhang, X. Wang, H. B. Zhang, and Y. C. Yan. 2009.
Advance and prospects in theories and techniques of modern digital grassland[J].
Chinese Journal of Grassland. 31(7): 1–8. (in Chinese).
Tellaeche, A., G. Pajares, X. P. Burgos-Artizzu, and A. Ribeiro. 2011. A computer vision

approach for weed identification through Support Vector Machines[J]. Applied Soft
Computing. 11, 908–915.
Van Evert, F K., G. Polder, GWAM. Van Der, C. Kempenaar, and L. Lotz. 2009. Real-time
vision-based detection of Rumex obtusifolius in grassland[J]. European Weed
Research Society Weed Research, 49,164–174.
Wang, J. X., Q. Feng, Y. T. Wang, and X. Q. Shao. 2010. Study on classification for
leguminous forage based on image recognition technology[J]. Acta Agrestia Sinica.
18, 37–41. (in Chinese).
Wang, J., J. He, 451 Y. Han, C. Ouyang , and D. Li. 2013. An adaptive Thresholding
algorithm of field leaf image[J]. Comput. Electron. Agr. 96:23–39.
Wu, X. T. 2003. Study on methods and application of grassland desertification

monitoring by remote sensing [D]. Beijing: Chinese Academy of Agricultural
Sciences. (in Chinese).
Xiao L. X., and Z. X. Zhang. 2008. Processes on the boundary definition of agro-pastor
al zone in China[J]. Progress in Geography. 27(2): 104–111. (in Chinese).
Yanikoglu, B. A., E. Aptoula, and C. Tirkaz. 2014. Automatic plant identification from
photographs[J]. Machine Vision and Applications. 25:1369-1383.
Yu C. L. 2007. Lossless detecting research of Water quantity in the leaves based on

image processing technology [D].JiLin: Ji Lin University. (in Chinese).
Zhang H. B. 2007. Study on variation of grassland vegetation based on multi-sensor

remote sensing data [D]. Beijing: Chinese Academy of Agricultural Sciences. (in
Chinese).
Zhang W. B., B. J. Lu, and S. Shi. 2009, Determination of vegetation coverage by

photograph and automatic calculation[J]. Bulletin of Soil and Water Conservation,
29(4): 39–42. (in Chinese)
Zhao, Y. S. 2003. Principles and methods of remote sensing applicaions and analysis
[M]. Beijing: Science Publication House. (in Chinese).
Zhao D. L., Z. L. Liu, G. X. Yang, X. P. Xin, C. Z. Liang, W. Wang, L. X. Wang, and W. P. Yuan.
2010. Grazing impact on distribution pattern of the plant communities and
populations in Stipa krylovii steppe [J]. Acat Prataculturae Sinica, 19(3): 6–13. (in
Chinese).
Seeds
https://doi.org/10.1007/978-981-16-3501-4_2
2. Forage Identification and

Experimental Materials

Xuanhe Zhao
Weihong Yan
Jiangping Liu
Xiaoling Luo
Tana Wuyun
2.1 Definition of Forage Automated Identification

The classical plant classification is mainly based on the external
morphology of plants, taking advantages of simple observation tools,
the experts analyze and compare plants indoors or in the field for their
similarity and variability, so as to distinguish and determine the species
(Cui 2010). Forage is the most important element in grassland in
providing food for cattle and keeping balance of our ecosystem. To
protect the grassland resources more effectively, we must know the
grass species and their attributes as the first step. Owing to wide
varieties between species and similarities of some closely related
species, it is hard to identify the forage directly, even for the experts.
The current situation hinders further development and practical
application for the forage and grassland. To solve the above issues, it is
necessary to develop automatic forage identification, so as to further
improve the grassland management. Therefore, machine vision is an
essential approach to enhance forage images and extract their
distinguishing features, thus the accuracy of forage identification can be
guaranteed with less dependence on experts.
Computer vision is a newly developed multi-discipline technology
to simulate human vision with computer and sensors, with which the
images can be acquired and converted to digital signals, and further
visual information processing including image transmission, processing
and understanding can be realized. In recent years, human biometrics
based on the intrinsic information has attracted wide attention as an
important ways to solve the public security issues. Many physiological
and behavior characteristics of the human body, such as human face,
fingerprint, iris, palm print and etc., can satisfy the requirements such
as universality, uniqueness, permanence, collectability, acceptability,
safety (Zhang 2000; Zhang et al. 2004). Therefore, face recognition,
fingerprint recognition and other biometric technologies and products
have become the research focus of the industry due to their accurate,
rapid and stable identification results (Ho and Chellappa 2013).
Analogy to biometric features used for personal identification, such
as face (Wang et al. 2014), fingerprints, palm-prints (Cui 2014) or gait,
forage classification can be categorized as a problem of pattern
recognition. In fact, each forage species is individually independent and
physiologically distinguishable from each other based on the seeds,
leaves, and flowers or even the entire plant. As shown in Fig. 2.1,
differences in shape, texture and color of forage can be observed in
seeds, plants, flowers and other forage organ images. These features
can be used as the input of computer vision to achieve automatic
classification and recognition. First forage images are acquired by
ordinary digital cameras, including seeds, leaves, flowers, organs, and
then the image preprocessing, then feature extraction and feature
matching process achieve the goal of forage classification by using
computer. In the process of research, the image forage database of the
living stages in the whole life cycle, from seed to seed can be
established. Scanner or digital camera can be used to collect local
pictures of seeds, stems, leaves, flowers and other different growth
stages to establish the grassland herbage database. By using computer
vision technology, the contour, texture, color and other inherent
features of the forage image can be extracted to quickly determine the
species.
Fig. 2.1 Images of forage seeds, plant and flowers

According to the function division of a pattern recognition system,
the automatic classification and recognition of forage can be realized by
four functional modules, including data acquisition and preprocessing,
feature extraction, pattern matching and decision (Fig. 2.2) (Duda et al.
2001; Bian and Zhang 2000).
Fig. 2.2 The function modules of forage auto classification system
Data acquisition refers to the use of optical sensor equipment such

as digital cameras or video cameras to capture images. The acquired
images are disturbed by the external environment, such as uneven
illuminations, motion distortion and deformation, so preprocessing the
acquired images is necessary in the following. Preprocessing usually
consists of two steps, one is to remove noise from the original images
with filters and smooth, so as to remove the influence of noise on the
images. Then the image is segmented to obtain ROI (Region of Interest)
for further processing.
Features are the information that describes the difference between
a certain original data and other original data. It is required to satisfy
the conditions of the minimum within class distance and the maximum
between classes at the same time as possible. Therefore, feature
extraction is the most critical part in forage recognition. It determines
the recognition performance of the system and becomes the key
research issue in the field of pattern recognition. The feature matching
module is to match the features to be recognized with the templates in
the template database and calculate the similarity (matching distance)
between them to determine the recognition result. It is the core
problem of pattern recognition and machine learning. The decision
module makes a decision on the matching result to determine the
category of the tested sample.
In conclusion, forage automatic recognition can be realized by
computer vision. Images can be acquired directly by digital camera,
which reduces both the difficulties and cost in sample collection and
saving. The recognition results with improved accuracy will no longer
rely seriously on the judgment of experts, the subjectivity of the
judgment results can be solved. In short, forage automatic classification
and identification has the characteristics of low cost, high stability and
strong anti-noise ability.
2.2 Automatic Identification of Herbage Based on

Seed Images
Forage seeds are the material basis for improving degraded grassland,
planting artificial grassland and improving grassland animal husbandry
production. They are as well as the basic materials for ecological
engineering construction, soil erosion and soil conservation
engineering construction and urban green space construction in arid
and semi-arid areas. Due to the diversity of leaves and flowers, seed
traits are relatively stable and less influenced by the environment. Seed
morphology is one of the most stable characteristics of plants.
Moreover, each forage seed is distinguished from other seeds due to its
unique morphological structure (Han and Mao 2011), which is an
important basis for automatic classification and identification of
herbage seeds based on seeds, authenticity identification of herbage
seeds, classification and inspection of herbage.
The external characteristics of forage seeds include shape, size,
color, seed coat surface characteristics and their appendages, as well as
the position, shape, size, concavity and color of the umbilicus.
Traditional classification of forage seeds is mainly based on the external
characteristics of seeds with observation, measurement and
comparative analysis of seeds carried out indoors by means of naked
eye, magnifying glass, anatomic mirror or measuring scale to study
their similarity and anisotropy, so as to determine the species of forage
seeds (Han and Mao 2011). Our research aims at the identification of
individual grass using common CCD images, which provides some
valuable information at a relatively small scale as compared to remote
sensing images. The research can be categorized to automatic plant
identification from photographs, with potential uses in two aspects: one
is to identify forage by the images of seeds or other organs for
researchers and farmers to protect the forage seeds resources, the
other is to assess grass seed to distinguish false seeds and inferior
seeds sneaked into seed market from the high quality seeds. The
automatic identification system is hopeful to solve these problems.
There are few reports on the studies forage seeds identification.
Similar studies include studies on identification of weed seeds and crop
seeds, which can provide references for our research. As introduced in
Sect. 1.3, the works of Granitto et al. (2002, 2005), Shi and Ji (2009)
concentrated on the outer appearance and texture description from the
gray level image, such as the color, geometrical features, and etc.
Pourreza et al. (2012) chose 131 textural features as the new features
for identificaiton, however, the multiple features selected manually for
optimum combination are insufficient to represent the seed images of
different species. In recent years, some new methods have been applied
in seed identification. Hong et al. (2015) used Random Forest for
feasible identification results. Cai et al. (2010) used compressed
sensing theory to identify 87 species of weed seeds, with the highest
recognition rate up to 90.80%.
It can be concluded from the above works that the study on the
automatic identification of forage seeds is necessary to be carried out.
The shape and texture of seeds are important distinguishing features
for seed identification. In addition, the images are presented as pixel
intensity matrix of gray scales in spatial domain, whereas the individual
pixel based data generally do not provide connection structure and
correlation information explicitly. Therefore, more robust global and
interior texture features should be explored to improve the
classification performance.
Gramineous grass seeds are the essential material basis for artificial
grassland construction and grassland improvement. Due to their high
yield, good quality, strong palatability and wide adaptability, grass
seeds play an important role in China's grassland animal husbandry.
Considering the importance of gramineous grass in grassland forage
and urban landscaping, we focus on the identification of the
gramineous grass seeds. Moreover, despite the small size of grass seeds
whose length is about 1 cm, clear images can be acquired by ordinary
digital CCD camera, which reduces the collection cost within a feasible
collection conditions. Through more subtle observations, it is found the
higher similarity between the gramineous seed, especially for the
different species in the family, such as Roegneria tunczaninovii
(Drob.)Nevski var. macrathera Ohwi, and (b) Roegneria varia Keng
shown in Fig. 2.3. Therefore, we emphasized on the textural features in
our research, the methods includes Gabor filters, Local Preserving
Projections (LPP), fractal dimension, Local Similarity Pattern (LSP),
GLCM, and so on. The corresponding contents were involved in Chaps.
3–6. In the remaining of this chapter, the acquisition method and
preprocessing of seed images are introduced respectively.
Fig. 2.3 Gramineous seeds a Roegneria tunczaninovii(Drob.)Nevski var. macrathera

Ohwi, and b Roegneria varia Keng
2.3 Experimental Materials

The seeds, 14 species belong to 5 genera (Elymus, Agropyron, Bromus,
Roegneria and Stipa), collected from grassland, which are listed in
Table 2.1 by name and origin, were provided by the Grassland Research
Institute of the Chinese Academy of Agricultural Sciences.
Table 2.1 The Image database of gramineous grass seeds
No Scientific name Origin

1 Elymus nutans Griseb Ameng, Inner Mongolia,
China
2 Elymus cylindricus (Frmch) Hinda Humeng, Inner Mongolia,
China
3 Elymus dahuricus Tunoz Ximeng, Inner Mongolia,
China
No Scientific name Origin
4 Agropyron cristatum var. pectiniforme (Roem.et Youyu, Shanxi, China
Schult) H Yang
5 Agropyron cristatum (Linn.) Gaertn Ximeng, Inner Mongolia,
China
6 Agropyron mongolicum Keng Ximeng, Inner Mongolia,
China
7 Agropyron desertorum (Fisch.) Schult Ximeng, Inner Mongolia,
China
8 Bromus inermis Leyss Ximeng, Inner Mongolia,
China
9 Elymas sibiricus Linn Ximeng, Inner Mongolia,
China
10 Roegneria tunczaninovii (Drob.) Nevski var. Yili, Xinjiang, China
macrathera Ohwi
11 Roegneria varia Keng Milin, Xizang, China
12 Roegneria kamoji Ohwi WuTai Moutain, Shanxi,
China
13 Stipa grandis Yanchi, Ningxia, China
14 Leymus chinensis (Trii) Tzvel Changping, Beijing, China
In this section, we will introduce the construction procedure of
database for experiments from three aspects, including image
acquisition, image preprocessing, and image database.
2.3.1 Image Acquisition

The gramineous grass seed images were captured by a commercial CCD
camera (DSLR-A350, Sony) under natural daylight illumination inside a
room (with non-controlled illumination). The seeds were put on the
rough surface of a black mouse mat for less diffuse reflectance, with
camera taken photos about 20 cm above. Seeds are generally composed
of seed coat, embryo, endosperm and other parts. Gramineous grass
seeds are usually caryopsis, in which the embryo is located on the back
side of the base of caryopsis (opposite to the side of lemma) with a
round or oval concave (Han and Mao 2011). A single kernel gramineous
seed has two different sides, the ventral side and the back side. The
seed lemma (back side) has relatively uniform and single textures (as
shown in Fig. 2.4b) with fewer features. The frontier ventral images
(Fig. 2.4a) were used to construct the experimental database because of
the multiple inner levels as compared to the back images. Auto flash
was used when the light was not strong enough from 5 p.m. to 6 p.m. in
winter. The focus lens was less than 100 mm. The camera was set to an
automatic focus and exposure. The other options were set
automatically under the standard program of the camera. No more
restrictions in the process of seed image acquisition provide a relatively
free style for the users. Figure 2.4 illustrates an original seed image
with a pixel resolution of 4592 × 3056.
Fig. 2.4 Original seed image (Elymus nutans Griseh). a Ventral side, b back side
2.3.2 Image Preprocessing

As can be seen from Fig. 2.4, sparkles and redundant background are
present in the image. The sparkles are the stains scattered in the mouse
mat which causes noise for further processing. Therefore, the images
are preprocessed to eliminate the noise and segment the Region of
Interest (ROI).
Preprocessing comprises three main steps. Initially, considering the
gramineous seeds are mostly medium brownish without unobvious
color difference, the original RGB color images can be converted into
gray images for efficient storage and computation. The images were
then converted into its binary image with a proper threshold of gray
levels (Fig. 2.5).
Fig. 2.5 Binary image
Subsequently, an open mathematical morphology operation is

applied to the binary image to eliminate the bright white sparkles of the
background. Suppose A and B represent the image matrix and
structuring element matrix, respectively. and denote the
erosion and dilation of image A by structuring element B. Then the
opening of image A by structuring B can be defined as follows
(Gonzalez and Woods 2008)
(2.1)
It can be seen that the opening operation comprises two steps,
erosion and dilation. Image A is first eroded by element B to eliminate
microscopic spots, and then recovered by dilation. Figure 2.6 shows the
result of open operation for a gramineous seed. Lastly, the seed
borderline is acquired from the binary image. Rotate the image to keep
the principal axis be horizontal, and determine the minimum enclosing
rectangle of the seed, thus the ROI is cropped from the original image,
as shown in Fig. 2.7.
Fig. 2.6 open operation of the seed
Fig. 2.7 ROI after segmentation
2.4 Seed Image DatabaseI

We constructed Seed Image DatabaseI of gramineous grass seed image
database using image acquisition techniques described in Sect. 2.2. In
view of difficulties in field seed acquisition of the gramineous grass, a
database contained 700 seed images comprising 14 species from 5
genera (Elymus, Agropyron, Bromus, Roegneria and Stipa) was
constructed.
In the image database, each species was represented by 50 images
captured from 10 seeds, 5 images per seed were captured with
variations of focuses, orientations and positions. For practical purpose,
the classification was confined to the species, rather than each seed, i.e.,
there were 14 classes in the database, and each class harbored 50
image samples. The original image resolution was 4592 × 3056 pixels of
3.30 MB. The images were stored in a jpeg compressed format. The
main bodies of the seeds were cropped without a fixed size due to the
seed diversity and the use of the auto focus during image acquisition.
Figures 2.8, 2.9 and 2.10 give some examples of the original seed
images and their corresponding ROIs in the image database. The ROI
resolution and size are listed in parentheses after the figure name.
Figure 2.11 shows all the classes in the seed image database of
Gramineous Grass.
Fig. 2.8 Agropyron cristatum var.pectiniforme(Roem.et Schult) H Yang, a original

image, b extracted ROI(1205 × 170, 30 KB)
Fig. 2.9 Agropyron cristatum (Linn.) Gaertn, a original image, b extracted ROI (1002
× 167, 21.2 KB)
Fig. 2.10 Elymas sibiricus Linn, a original image, b extracted ROI (1970 × 289,
66.4 KB)
Fig. 2.11 Seed image database of Gramineous Grass (from left to right, from top to
bottom. Row 1: Elymus nutans Griseb, Elymus cylindricus (Frmch) Hinda; Row 2:
Elymus dahuricus Tunoz, Agropyron cristatum var.pectiniforme (Roem.et Schult) H
Yang; Row 3: Agropyron cristatum (Linn.) Gaertn, Agropyron mongolicum Keng; Row
4: Agropyron desertorum (Fisch.) Schult., Bromus inermis Leyss.; Row 5: Elymas
sibiricus Linn, Roegneria tunczaninovii (Drob.) Nevski var.macrathera Ohwi; Row 6:
Roegneria varia Keng; Roegneria kamoji Ohwi; Row 7: Stipa grandis, Leymus
chinensis (Trii) Tzvel.)
References
Bian, Z. Q., and X. G. Zhang. 2000. Pattern recognition (2nd Version)[M]. Beijing:
Tsinghua University Press. (in Chinese).
Cai, C., M. Zhang, and J. P. Zhu. 2010. Weed seeds classification based on compressive
sensing theory[J], Science China: Inforamtion Science, 40(Supplement): 160-172. (in
Chinese).
Cui, D. F. 2010. Plant taxonomy(3rd Version)[M]. Beijing: Chinese Agriculture Press.
(in Chinese).
Cui, J. 2014. 2D and 3D palmprint fusion and recognition using PCA plus TPTSR
method[J]. Neural Comput& Applic, 24(3-4):497-502.
Duda, R. O., P. E. Hart, and D. G. Stork. 2001. Pattern Classification(2nd Edition)[M].

U.S: John Wiley & Sons.
Gonzalez, R. C., and R. E. Woods. 2008. Digital Image Processing[M]. Prentice Hall.
Han, J. G., and P. S. Mao. 2011. Forage seed science (2nd Version)[M]. Beijing: China
Agricultural University Press. (in Chinese).
Ho, H. T., and R. Chellappa. 2013. Pose-Invariant face recognition using markov
random fields. IEEE Transactions on Image Processing, 22(4):1573–1584.
Hong, P., T. Hai, L. Lan, V. Hoang, V. Hai, and T. Nguyen. 2015. Comparative study on
vision based rice seed varieties identification[C], 7th International Conference on
Knowledge and Systems Engineering. 377–382.
Granitto, P. M. , Navone, H. D. , Verdes, P. F. , & Ceccatto, H. A. . 2002. Weed seeds

identification by machine vision[J], Computers and Electronics in Agriculture, 33:91-
103.
Granitto, P. M., Verdes, P. F., Ceccatto, H. A., 2005. Large-scale investigationof weed
seed identification by machine vision. Computers and Electronics in Agriculture.
47,15-24.
Pourreza, A., H. Pourreza, M. Abbaspour-Fard, and H. Sadrnia. 2012. Identification of

nine Iranian wheat seed varieties by textual analysis with image processing[J].
Comput. Electron. Agr. 83:102-108.
Shi, C., and G. Ji. 2009. Study of recognition method of leguminous weed seeds
image[C]. in: Proceedings of International Workshop on Intelligent Systems and
Applications, 1–4.
Wang, N., Q. Li, A. El-Latif, J. Peng, and X. Niu. 2014. An enhanced thermal face
recognition method based on multiscale complex fusion of Gabor coefficients[J].
Multimed. Tools Appl.72:2339-2358.
Zhang, D. Automated Biometrics-Technologies and Systems[M]. Kluwer Academic

Publishers. 2000.
Zhang, D. Palmprint authentication[M]. Kluwer Academic publishers, 2004.
Seeds
https://doi.org/10.1007/978-981-16-3501-4_3
3. Identification of Gramineous Grass

Seeds Using Gabor and Locality
Preserving Projections

Xuanhe Zhao
Weihong Yan
Jiangping Liu
Xiaoling Luo
Tana Wuyun
Therefore, considering the importance of gramineous grass in

grassland forage and urban landscaping, we propose a classification
system of gramineous grass based on seed images using Gabor filter
and locality preserving projections (Pan et al. 2016). The classification
system comprises 4 modules: image acquisition, image preprocessing,
feature extraction and feature matching (Fig. 3.1). At first, the seed
images are captured by a common digital camera with a macro lens of
100 mm under daylight illumination. Then morphological operations
are used for noise elimination and the Region of interest (ROI)
segmentation. In the third module, Gabor wavelets and LPP are applied
for robust feature extraction. The former can provide robust features
against varying image brightness and contrast (Zhang et al. 2003), and
the latter can preserve the image manifold structure (He et al. 2005) for
efficient dimensionality reduction. The integration yields robust
features for seed image presentation. Finally, the nearest neighbor (NN)
classifier with Euclidean distance, and linear discriminant analysis
(LDA) classifier are used for classification. The experimental results
demonstrate the effectiveness of the proposed algorithm. The novelty
lies in applying Gabor filters and LPP to extract the textural manifold
features of the seed images, not limited to the appearance and
geometric features.
Fig. 3.1 Block diagram of the identification system for gramineous grass seeds
Therefore, in the proposed system in this study, Gabor filters and

locality preserving projections (LPP) are attempted to extract textural
manifold features of gramineous seeds for forage identification.
In this chapter, we introduce a classification algorithm of
gramineous grass based on seed images using Gabor filter and locality
preserving projections (LPP). It is a feature extraction module in the
recognition system, executed after image acquisition and preprocessing
that were introduced in Sect. 3.2 (Fig. 3.1). The former Gabor wavelets
can provide robust features against varying image brightness and
contrast (Zhang et al. 2003), and the latter LPP can preserve the image
manifold structure (He et al. 2005) for efficient dimensionality
reduction. The integration yields robust features for seed image
presentation. Finally, the nearest neighbor (NN) classifier with
Euclidean distance, and linear discriminant analysis (LDA) classifier are
used for classification. The experimental results demonstrate the
effectiveness of the proposed algorithm. The novelty lies in applying
Gabor filters and LPP to extract the textural manifold features of the
seed images, not limited to the appearance and geometric features.
In the remainder of this chapter, we will predominantly focuses on
the feature extraction with Gabor and LPP, and feature matching with
NN and LDA in Sect. 3.1. The experimental results and discussions are
listed in Sect. 3.2. Section 3.3 highlights the conclusions.
3.1 Feature Extraction

In order to extract the robust features of the gramineous seeds images
effectively, Gabor filters and LPP are integrated for manifold texture
extraction against variations, especially illumination variations of the
seed images. The application in face recognition shows the robust
characteristics in texture analysis (Hu 2012).
3.1.1 Gabor Filters

The imaging condition has a significant impact on the classification
efficiency (Pourreza et al. 2012). In consideration of the sophisticated
surroundings of image acquisition procedure of the proposed system,
such as non-controlled illumination under daylight, auto flash used for
light compensation, automatic focus, and handheld camera without
mounting facilities, Gabor filters were used to alleviate the effects
caused by imaging condition before further processing. Numerous
studies have demonstrated that Gabor wavelets can provide robust
features against varying image brightness and contrast in addition to
accurate time frequency location. The successful application of Gabor
filters in texture analysis involves a large quantity of areas, such as iris
recognition, face recognition, etc. The utility of Gabor filters in the
separation of weed and crops from natural images proves the good
performance on weed texture (Ishak et al. 2009).
2D Gabor filters exhibits the following general form (Daugman
1980):
(3.1)
where , u is the frequency of the sinusoidal wave, controls
the orientation of the function, and is the standard deviation of the

Gaussian envelope. For an input seed image I(x, y), the convolution of
Gabor filter and the seed image yields Gabor filtered images .
The transformed results include real and imaginary parts, which can be
used to calculate the amplitude. Let OR and OI represent the real and
imaginary parts, respectively, the magnitude and phase can be
computed by the following equation:
(3.2)
The imaginary part and the amplitude of the Gabor features are
displayed in Fig. 3.2a, b, respectively. For N training images, all of the
Gabor-filtered vectors form the Gabor feature space GS = {O1, O2,…, ON}.
Fig. 3.2 Gabor filtered image, a imaginary part, b amplitude
3.1.2 Locality Preserving Projections

Locality preserving projections, an appearance-based linear subspace
method, preserve the intrinsic geometry of the data and local manifold
structure modeled by a nearest neighbor graph of the images (He et al.
2005; Cai et al. 2006). In the proposed system, LPP is applied onto the
Gabor feature space GS of the seed images. The integration can provide
robust features for varying brightness and image contrast while
preserving the manifold structure of the images for efficient
dimensionality reduction. The following specific steps are listed:
(1) Principal component analysis (PCA) projection. The Gabor feature
space is first projected onto the PCA subspace, and N principle
components are preserved because the number of training
samples is relatively small compared to the dimension of training
vectors.
(2) Construction of the nearest neighbor graph and the similarity

matrix S. When each node corresponds to a Gabor filtered vector,
the nearest neighbor graph of N nodes can be constructed. For a
node i, it can be connected with node j if they are close, where
their distance is among the k nearest one of node i and all other
nodes in the graph. The similarity matrix S can be defined as an N
× N matrix,
(3.3)
where , if node j is among k nearest neighbors of i,

otherwise , i, j = 1, 2,…, n.
(3) Eigenmap: Computation of the optimum LPP transform matrix by

solving the following generalized eigenvalues problem:
Another random document with
no related content on Scribd:
TABLE SHOWING THE NUMBER OF ALIENS ADMITTED TO AND DEPARTED FROM THE
UNITED STATES, AND THE NET GAIN OR LOSS IN POPULATION RESULTING THEREFROM
BY MONTHS, FROM 1907 TO 1910
Month 1907 1908
Admitted Departed Gain (+) or Loss Admitted Departed Gain (+) or
(−) Loss (-)
January 54,417 33,058 60,233 − 27,175
February 65,541 30,266 50,688 − 20,422
March 139,118 43,537 43,506 + 31
April 145,256 55,220 65,721 − 10,501
May 184,886 48,245 61,251 − 13,006
June 154,734 41,094 60,482 − 19,388
July 107,535 46,198 + 61,337 37,133 51,508 − 14,375
August 111,135 44,317 + 66,818 39,606 47,569 − 7,963
September 115,287 43,734 + 71,553 56,635 43,884 + 12,751
October 129,564 55,826 + 73,738 60,715 41,916 + 18,799
November 132,647 94,440 + 38,207 50,965 38,609 + 12,356
December 77,107 88,432 − 11,325 61,111 33,416 + 27,695
1909 1910
January 54,975 18,061 + 36,914 57,472 20,256 + 37,216
February 81,992 15,100 + 66,892 66,072 17,672 + 48,400
March 135,040 22,550 + 112,490 152,020 30,894 + 121,126
April 138,382 24,315 + 114,067 153,915 40,886 + 113,029
May 127,139 31,190 + 95,949 148,822 38,740 + 110,082
June 100,542 32,274 + 68,268 115,793 36,119 + 79,674
July 77,944 27,940 + 50,004 82,191 39,056 + 43,135
August 71,992 28,450 + 43,542 91,460 37,206 + 54,254
September 85,088 29,950 + 55,138 100,456 43,023 + 57,433
October 92,372 30,838 + 61,534 100,334 39,189 + 61,145
November 98,020 39,134 + 58,886 86,144 54,700 + 31,444
December 78,527 39,539 + 38,988 68,794 61,814 + 6,980
It is not necessary to take account of these discriminations for the
purposes of the present study.
Turning then to the table, we observe that the monthly average of
arrivals during the first six months of 1907 was a high one. Following
a large immigration during the last six months of the preceding year,
this made the fiscal year ending June 30, 1907, the record year for
immigration in the history of the country. For the next four months
the stream of immigration continued high, considering the season,
and the number of departures was moderate. Early in October,
however, there were signs of disturbance in the New York Stock
Exchange. On the sixteenth there was a crash in the market, and
within a week the panic had become general. It reached its height on
October 24, and continued for many weeks after.[318] The response of
the alien population to this disturbance was almost immediate, and
manifested itself first in the emigration movement. In November the
number of departures almost doubled. But the immigrants who were
on the way could not be stopped, and in spite of the large exodus,
there was a net gain of 38,207 during the month. The next month,
December, however, saw a marked decrease in the stream of arrivals,
which, accompanied by a departure of aliens almost as great as in
November, resulted in a net decrease in population of 11,325 for the
month. During the first six months of 1908 the number of arrivals
was small, and the departures numerous, so that, with the exception
of March, each month shows a net loss in population. During July
the number of departures began to approach the normal (compare
the months in 1908 with 1907 and 1910), but the arrivals were so few
that there was still a decrease for the months of July and August. In
September, 1908, the balance swung the other way, and from that
time to the present every month with the exception of December,
1911, has shown a substantial increase in population through the
movement of aliens.
Thus we see that the period during which the number of alien
laborers in the United States was decreasing was confined to the
months December, 1907, to August, 1908, inclusive.[319] By the end of
July, 1908, the effects of the crisis were practically over as far as
departures are concerned. It is evident, then, that the effects of the
crisis on emigration were immediate, but not of very long duration.
During the months of November and December, 1907, when the
distress was the keenest, there were still large numbers of aliens
arriving. But when the stream of immigration was once checked, it
remained low for some time, and it was not until about January,
1909, that it returned to what may be considered a normal figure.
The reasons for this are obvious. The stream of immigration is a long
one, and its sources are remote. It takes a long time for retarding
influences in America to be thoroughly felt on the other side. The
principal agency in checking immigration at its source is the
returning immigrant himself, who brings personal information of the
unfortunate conditions in the United States. This takes some time.
But when the potential immigrants are once discouraged as to the
outlook across the ocean, they require some positive assurance of
better times before they will start out again.
Now what catches the public eye in such an epoch as this is the
large number of departures. We are accustomed to immense
numbers of arrivals and we think little about that side of it. But heavy
emigration is a phenomenon, and accordingly we hear much about
how acceptably our alien population serves to accommodate the
supply of labor to the demand. But if we stop to add up the monthly
figures, we find that for the entire period after the crisis of 1907,
when emigration exceeded immigration, the total decrease in alien
population was only 124,124—scarcely equal to the immigration of a
single month during a fairly busy season. This figure is almost
infinitesimal compared to the total mass of the American working
people, or to the amount of unemployment at a normal time, to say
nothing of a crisis.[320] It is thus evident that the importance of our
alien population as an alleviating force at the time of a crisis has been
vastly exaggerated. The most that can be said for it is that it has a
very trifling palliative effect.
The really important relation between immigration and crises is
much less conspicuous but much more far-reaching. It rests upon the
nature and underlying causes of crises in this country. These are
fairly well understood at the present time. A typical crisis may be
said to be caused by speculative overproduction, or overspeculative
production. Some prefer to call the trouble underconsumption,
which is much the same thing looked at from another point of view.
Professor Irving Fisher has furnished a convenient and logical
outline of the ordinary course of affairs.[321] In a normal business
period some slight disturbance, such as an increase in the quantity of
gold, causes prices to rise. A rise in prices is accompanied by
increased profits for business men, because the rate of interest on the
borrowed capital which they use in their business fails to increase at
a corresponding ratio. If prices are rising at the rate of two per cent
annually, a nominal rate of interest of six per cent is equivalent to an
actual rate of only about four per cent. Hence, doing business on
borrowed capital becomes very profitable, and there is an increased
demand for loans.
This results in an increase of the deposit currency, which is
accompanied by a further rise in prices. The nominal rate of interest
rises somewhat, but not sufficiently, and prices tend to outstrip it
still further. Thus the process is repeated, until the large profits of
business lead to a disproportionate production of goods for
anticipated future demand, and a vast overextension of credit. But
this cycle cannot repeat itself indefinitely. Though the rate of interest
rises tardily, it rises progressively, and eventually catches up with the
rise in prices, owing to the necessity which banks feel of maintaining
a reasonable ratio between loans and reserves. Other causes operate
with this to produce the same result. The consequence is that
business men find themselves unable to renew their loans at the old
rate, and hence some of them are unable to meet their obligations,
and fail. The failure of a few firms dispels the atmosphere of public
confidence which is essential to extended credit. Creditors begin to
demand cash payment for their loans; there is a growing demand for
currency; the rate of interest soars; and the old familiar symptoms of
a panic appear. In this entire process the blame falls, according to
Professor Fisher, primarily upon the failure of the rate of interest to
rise promptly in proportion to the rise in prices. If the forces which
give inertia to the rate of interest were removed, so that the rate of
interest would fluctuate readily with prices, the great temptation to
expand business unduly during a period of rising prices would be
removed. It may well be conceived that there are other factors,
besides the discrepancy between the nominal and real rates of
interest, that give to business a temporary or specious profitableness,
and tend to encourage speculative overproduction. But the influence
of the rate of interest resembles so closely that resulting from
immigration, that Professor Fisher’s explanation is of especial service
in the present discussion.
The rate of interest represents the payment which the
entrepreneur makes for one of the great factors of production—
capital. The failure of this remuneration to keep pace with the price
of commodities in general leads to excessive profits and
overproduction. The payment which the entrepreneur makes for one
of the other factors of production—labor—is represented by wages. If
wages fail to rise along with prices, the effect on business, while not
strictly analogous, is very similar to that produced by the slowly
rising rate of interest. The entrepreneur is relieved of the necessity of
sharing any of his excessive profits with labor, just as in the other
case he is relieved from sharing them with capital. It would probably
be hard to prove that the increased demand for labor results in
further raising prices in general, as an increased demand for capital
results in raising prices by increasing the deposit currency. But if the
demand for labor results in increasing the number of laborers in the
country, thereby increasing the demand for commodities, it may very
well result in raising the prices of commodities as distinguished from
labor, which is just as satisfactory to the entrepreneur. This is exactly
what is accomplished when unlimited immigration is allowed. As
soon as the conditions of business produce an increased demand for
labor, this demand is met by an increased number of laborers,
produced by immigration.
In the preceding paragraph it has been assumed that wages do not
rise with prices. The great question is, is this true? This is a question
very difficult of answer. There is a very general impression that
during the last few years prices have seriously outstripped wages.
Thus Professor Ely says, “Wages do not usually rise as rapidly as
prices in periods of business expansion.”[322] R. B. Brinsmade stated
in a discussion at a recent meeting of the American Economic
Association that “our recent great rise of prices is acknowledged to be
equivalent to a marked reduction in general wages.”[323] Whether this
idea is correct, and if correct, whether this effect had transpired in
the years immediately previous to 1907, cannot be definitely stated.
The index numbers of wages and prices given in the Statistical
Abstract of the United States, for 1909 (p. 249), seem to show that
during the years 1895 to 1907 money wages increased about pari
passu with the retail prices of food, so that the purchasing power of
the full-time weekly earnings remained nearly constant.
But whether or not money wages rose as fast as prices in the years
from 1900 to 1907, one thing is certain, they did not rise any faster.
That is to say, if real wages did not actually fall, they assuredly did
not rise. But the welfare of the country requires that, in the years
when business is moving toward a crisis, wages should rise; not only
money wages, but real wages. What is needed is some check on the
unwarranted activity of the entrepreneurs, which will make them
stop and consider whether the apparently bright business outlook
rests on sound and permanent conditions, or is illusory and
transient. If their large profits are legitimate and enduring, they
should be forced to share a part of them with the laborer. If not, the
fact should be impressed upon them. We have seen that the rate of
interest fails to act as an efficient check. Then the rate of wages
should do it. And if the entrepreneurs were compelled to rely on the
existing labor supply in their own country, the rate of wages would
do it. Business expands by increasing the amount of labor utilized, as
well as the amount of capital. If the increased labor supply could be
secured only from the people already resident in the country, the
increased demand would have to express itself in an increased wage,
and the entrepreneur would be forced to pause and reflect. But in the
United States we have adopted the opposite policy. In the vast
peasant population of Europe there is an inexhaustible reservoir of
labor, only waiting a signal from this side to enter the labor market—
to enter it, not with a demand for the high wage that the business
situation justifies, but ready to take any wage that will be offered, just
so it is a little higher than the pittance to which they are accustomed
at home. And we allow them to come, without any restrictions
whatever as to numbers. Thus wages are kept from rising, and
immigration becomes a powerful factor, tending to intensify and
augment the unhealthy, oscillatory character of our industrial life. It
was not by mere chance that the panic year of 1907 was the record
year in immigration.
Against this point of view it may be argued that the legitimate
expansion of business in this country requires the presence of the
immigrant. But if business expansion is legitimate and permanent,
resting on lasting favorable conditions, it will express itself in a high
wage scale, persisting over a long period of time. And the demand so
expressed will be met by an increase of native offspring, whose
parents are reaping the benefit of the high standard of living. A
permanent shortage of the labor supply is as abhorrent to nature as a
vacuum. Expansion of any other kind than this ought to be
hampered, not gratified.
There is one other way in which immigration, as it exists at
present, influences crises. In considering this, it will be well to regard
the crisis from the other point of view—as a phenomenon of
underconsumption. Practically all production at the present day is to
supply an anticipated future demand. There can be no
overproduction unless the actual demand fails to equal that
anticipated. This is underconsumption. Now the great mass of
consumers in the United States is composed of wage earners. Their
consuming power depends upon their wages. In so far as
immigration lowers wages in the United States, or prevents them
from rising, it reduces consuming power, and hence is favorable to
the recurrence of periods of underconsumption. It is not probable, to
be sure, that a high wage scale in itself could prevent crises, as the
entrepreneurs would base their calculations on the corresponding
consuming power, just as they do at present. But a high wage scale
carries with it the possibility of saving, and an increase of
accumulations among the common people. It is estimated at the
present time that half of the industrial people of the United States
are unable to save anything.[324] This increase in saving would almost
inevitably have some effect upon the results of crises, though it must
be confessed that it is very difficult to predict just what this effect
would be. One result that might naturally be expected to follow
would be that the laboring classes would take the opportunity of the
period of low prices immediately following the crisis to invest some
of their savings in luxuries which hitherto they had not felt able to
afford. This would increase the demand for the goods which
manufacturers are eager to dispose of at almost any price, and would
thereby mitigate the evils of the depressed market. It is probably true
that the immigrant, under the same conditions, will save more out of
a given wage than the native, so that it might seem that an alien
laboring body would have more surplus available for use at the time
of a crisis than a native class. But the immigrant sends a very large
proportion of his savings to friends and relatives in the old country,
or deposits it in foreign institutions, so that it is not available at such
a time. Moreover, our laboring class is not as yet wholly foreign, and
the native has to share approximately the same wage as the alien.
Without the immense body of alien labor, we should have a class of
native workers with a considerably higher wage scale, and a large
amount of savings accumulated in this country, and available when
needed.
On the other hand, it may be argued that if the desire to purchase
goods in a depressed market should lead to a large withdrawal of
cash from savings banks and similar institutions, it might tend to
augment rather than alleviate the evils of a money stringency. There
seems to be much force to this argument. Yet Mr. Streightoff tells us
that in a period of hard times the tendency is for the poorer classes to
increase their deposits, rather than diminish them.[325] On the whole,
it seems probable that a large amount of accumulated savings in the
hands of the poorer classes would tend to have a steadying influence
on conditions at the time of a crisis, and that by preventing this, as
well as in other ways, immigration tends to increase the evils of
crises.
In closing this discussion, it may be interesting to note what are
the elements in our alien population which respond most readily to
economic influences in this country, and hence are mainly
accountable for the influences we have been considering. As stated
above, the annual reports of the Commissioner General of
Immigration give very complete data as to the make-up of the
incoming and outgoing streams by years. Thus in the fiscal year 1908
there were 782,870 immigrant aliens and 141,825 nonimmigrant
aliens admitted. Of the nonimmigrant aliens, 86,570 were
individuals whose country of last permanent residence and of
intended future residence were both the United States; that is, they
were alien residents of this country who had been abroad for a brief
visit. These are the birds of passage in the strictest sense, in which we
shall use the term hereafter. In the same year there was a total
exodus of 714,828 aliens, of whom 395,073 were emigrants and
319,755 nonemigrants. The former class includes those who have
made their fortune in this country and are going home to spend it,
and those who have failed, and are going home broken and
discouraged—a very large number in this panic year. The latter class
includes aliens who have had a permanent residence in the United
States, but who are going abroad to wait till the storm blows over,
with the expectation of returning again—true birds of passage
outward bound. There were 133,251 of these. The balance were aliens
in transit, and aliens who had been in this country on a visit, or only
for a short time. In 1909 there were 751,786 immigrant aliens and
192,449 nonimmigrant aliens. Of the nonimmigrants 138,680 were
true birds of passage according to the above distinction—a large
number and almost exactly equal to the number of departing birds of
passage in the previous year. The storm is over, and they have come
back. The departures in that year numbered 225,802 emigrant and
174,590 nonemigrant aliens. These numbers are considerably
smaller than in the previous year, but are still large, showing that the
effects of the crisis were still felt in the early part of this fiscal year.
The number of birds of passage among the nonemigrant aliens,
80,151, is much smaller than in the previous year. In 1910 there were
1,041,570 immigrant aliens and 156,467 nonimmigrant aliens. In the
latter class, the number of birds of passage, 94,075, again
approximated the corresponding class among the departures of the
previous year. The departures in 1910 were 202,436 emigrant aliens
and 177,982 nonemigrant aliens, of whom 89,754 were birds of
passage. This probably comes near to representing the normal
number of this class. A careful study of these figures confirms the
conclusion reached above. While a crisis in this country does
undoubtedly increase the number of departing aliens, both emigrant
and nonemigrant, and eventually cuts down the number of arrivals,
the total effect is much smaller than is usually supposed, and taken
in connection with the fact that the stream of arrivals is never wholly
checked, the influence of emigration in easing the labor market is
absolutely trifling.
Comparing the different races in regard to their readiness to
respond to changes in economic conditions, it appears that the
Italians stand easily at the head, and the Slavs come second. In 1908,
in the traffic between the United States and Italy, there was a net loss
in the population of this country of 79,966; in 1909 a net gain of
94,806. In the traffic between this country and Austria-Hungary
there was a loss in 1908 of 5463; in 1909 a gain of 48,763. In the
traffic with the Russian Empire and Finland there was a gain of
104,641 in 1908 and a gain of 94,806 in 1909. This shows how
unique are the motives and conditions which control the emigration
from the two latter countries. The emigrants from there, particularly
the Jews, come to this country to escape intolerable conditions on
the other side, not merely for the sake of economic betterment. They
prefer to endure anything in this country, rather than to return to
their old home, even if they could.
Hand in hand with the economic disparity caused by immigration
has come an increasing social stratification.[326] This is based partly
on wealth, partly on race. Already certain occupations are regarded
as the special province of certain nationalities, and native parents
recoil from the prospect of having their children enter them to work
side by side with the aliens. Only the beginnings of these changes are
as yet manifest, and no one can foretell what the outcome will be. But
even the beginnings must give us pause. There can be no more
pernicious social classification in a nation than one based on race.
Distinctions resting on wealth, religion, or education can be
overcome, potentially at least. Distinctions of birth affect only a small
proportion of a society, and exist only in nations long habituated to
them. But distinctions of race affect the entire population are
fundamental, and can never be obliterated except as assimilation is
so perfect that race is forgotten. No effort of the individual can blot
out his racial identification. The most familiar example yet developed
in the United States is that of the Hebrews. However sincerely we
may admire their fine racial traits, however closely we may associate
with individuals of the race, we cannot deny that they constitute a
separate body in our population in many respects.[327] Summer hotels
are closed to them, or else other people avoid those resorts.
Americans move out of the sections of cities where they are moving
in. Select clubs are closed to them. It is an indictment against the
American people that these things are so. We, who pose as the
friends of all races, however downtrodden and despised, should be
ready to take them into equality with us when they seek refuge on
our shores. Both Hebrews and Americans may resent the bald
statement of such facts. Can we deny their truth?
Nor is it only in high society, nor only among Americans, that this
friction is felt. In the slums of our cities bitter feeling exists between
the Italians and the Jews.[328] Nor is racial antagonism confined to
any two or three races.[329] Employers of labor find it wholly
expedient to arrange their workers in groups of the same nationality.
[330]
Austria-Hungary is an example of the conditions that may result
when too many jarring nationalities are included within a national
territory. But the racial groups in Austria-Hungary do not compare
in diversity with those which are gradually forming in the United
States.
In the political aspects of the immigration situation there has been
a peculiar reversal of public opinion in the last three quarters of a
century. In the days of the Native Americans and the Know nothings,
the uneasiness was mainly due to the fear that too many aliens would
acquire the rights of citizenship. Then it was the naturalized
foreigner who was the undesirable. Nowadays, the fear is that the
foreigners will ignore the privileges of citizenship, and a high
percentage of naturalization is a test of desirability in any foreign
group. This change may be attributed to a change in the situation of
the United States, and to a difference in the character and causes of
immigration. During the first half of the nineteenth century the
United States was essentially a new country. Political questions were
predominant, and the memory of the men who fell in the fight for
freedom was still fresh in the minds of their sons. The immigrants of
the period, on the other hand, were actuated to a large extent by the
desire for political freedom, and were keen to secure all the power
possible in this country. At the present time, the predominating
interests are wholly economic, and even the political questions of the
day have an economic flavor. At the same time, the motives of the
immigrants are almost wholly economic. So the jealousy between
native and foreigner now concerns itself mainly with the industrial
relations, and anything which indicates an inclination on the part of
the alien to ally himself permanently with the interests of the country
is welcomed. The temporary immigrant was an almost unknown
quantity in the old days.
The naturalization laws of the United States have undergone only
slight modifications in the past hundred years.[331] The main
provisions of the present laws are as follows: In order to become a
citizen of the United States an alien must follow out the following
method of procedure: At least two years before he is admitted he
must file a preliminary declaration of intention. To do this he must
be at least eighteen years old. This declaration shall state that it is his
bona-fide intention to become a citizen of the United States, and to
renounce all other allegiance to a foreign power, and shall set forth
his name, age, occupation, personal description, place of birth, last
foreign residence and allegiance, date of arrival in the United States,
name of the vessel, if any, by which he came, and present place of
residence in the United States. Not less than two years nor more than
seven years after he has made application, he shall present a petition
in writing, signed in his own handwriting, stating the essential facts
about himself, including his declaration of allegiance to the United
States, and disclaiming belief in anarchy, or belief in or practice of
polygamy.
This petition shall be verified by at least two credible witnesses,
who are citizens of the United States, who shall state that they have
known the applicant to be a resident of the United States for a period
of at least five years continuously, and of the state or territory at least
one year immediately preceding, and that they have personal
knowledge of his good moral character and general fitness to become
a citizen of the United States.
With this petition is filed a certificate from the Department of
Commerce and Labor, stating the date, place, and manner of his
arrival, and also his declaration of intention. He shall swear in open
court his allegiance to the United States and renounce all other
allegiance.
In accordance with a recent law, no alien can now be naturalized
without an ability to speak the English language, unless he has made
entry upon the public lands of the United States. No person may be
naturalized within thirty days preceding the holding of a general
election in the territorial jurisdiction of the court. Chinese are not
admissible to citizenship.
A woman who is married to a citizen of the United States is herself
a citizen, provided she herself might be legally naturalized. This
provision has been the subject of considerable attention lately on
account of the practice of women engaged in the white slave traffic
marrying a citizen in order to avoid deportation. The Commissioner
General in his report for 1910 recommended that a more definite
statement be made of this clause, admitting of no doubt as to its
interpretation.
Children of naturalized citizens who were under the age of twenty-
one at the time of the naturalization of their parents, if dwelling in
the United States, are considered citizens, as are children of citizens,
born outside of the United States.
If any alien who has received a certificate of citizenship shall,
within five years thereafter, go to the land of his nativity or to any
other foreign country, and take up permanent residence therein, it
shall be deemed evidence of his lack of intention to become a
permanent citizen of the United States at the time of filing his
application, and warrants the canceling of his certificate.
According to the regulations of September 15, 1910, clerks of
courts are instructed not to receive declarations of intention or file
petitions for naturalization from other aliens than white persons, and
persons of African nativity or of African descent.
Jurisdiction to naturalize aliens is conferred on the following
courts: United States circuit and district courts in any state, United
States district courts for the territories, the supreme court of the
District of Columbia, and the United States courts for the Indian
territory; also all courts of record in any state or territory, having a
seal, a clerk, and jurisdiction in actions at law or equity, or law and
equity, in which the amount in controversy is unlimited.
Since the establishment of the division of naturalization by the act
of June 29, 1906, the business of naturalization has been in the
hands of the Bureau of Immigration and Naturalization.
The statistics of naturalization for the five years 1908–1912 are as
follows:
Year Declarations filed Petitions filed Certificates granted
1908 137,229 44,029 25,963
1909 145,794 43,161 38,372
1910 167,226 55,038 39,206[332]
1911 186,157 73,644 55,329
1912 169,142 95,627 69,965
332. Repts. Comm. Gen. of Imm.
In addition to the certificates granted there were, in 1912, 9635
certificates denied. These were for a variety of causes, the most
important of which was failure of the petitioners to prosecute them,
so that they were stricken from the docket. Of those which were
actually refused the largest single cause was incompetent witnesses.
There has been a large amount of fraud and trickery in connection
with naturalization, and presumably it has not wholly ceased. This
has been due partly to a lax attitude on the part of some of the court
officials, and partly to the physical impossibility of giving proper
attention to the number of candidates who apply, with the existing
machinery. There is a story of one judge in New York City who issued
nearly seven thousand papers in October, 1891, at the rate of two a
minute.[333] Many states have been very lax in their requirements for
voting. In some states aliens have been allowed to vote in both state
and federal elections, sometimes after a residence of only six months.
[334]
Even where naturalization is desired by recent immigrants, it is

not always for the most commendable reasons. Sometimes the
motive is the desire for a better chance of securing employment,[335]
sometimes the facilitating of entrance into the United States after a
trip abroad. Natives of some foreign countries, particularly Turkey,
have come to the United States with the express intention of securing
citizenship, in order to return to their native land, and carry on
business under the protection of the American flag, which carries
with it greater guarantees than their own. A special law, passed to
put a stop to such practices as these, provides that when a
naturalized alien has resided two years in the foreign state from
which he came, or five years in any other foreign state, he forfeits his
citizenship.[336]
Of all foreign races, the Irish have taken by far the largest place in
politics in this country. According to Professor Commons, the “ward
boss” is the logical product of the mixture of nationalities in the
various divisions of a city, and the Irishman is the logical man for the
work.[337] “The Irishman has above all races the mixture of ingenuity,
firmness, human sympathy, comradeship, and daring that makes
him the amalgamator of races.”[338] Possibly a sense of humor ought
to be added to these qualifications. In the eyes of Professor
Commons, such a system makes it the merest chance if the best man
is elected, and subverts our whole system of representative
government.[339] It seems beyond question that the existence of
separate racial groups in a community, each with its own prejudices
and group loyalty, must have a very disturbing influence on the
course of elections. Measures become of much less import than men
in the minds of the voters, and in the choice of men race rather than
fitness is often the determining element.
CHAPTER XVII
THE NEW PROBLEM OF IMMIGRATION
It was stated on an earlier page that the immigration situation, in

most of its important characteristics, presents an entirely new aspect
to the men of this generation, and that these changes might be
looked for under six general heads, as follows: race, volume,
distribution, economic condition of the United States, native birth
rate and quality of the immigrants. We are now prepared to consider
the truth of this assertion.
In regard to race, nothing further need be said. Sufficient facts are
already before the reader to establish the fact that the racial aspect of
the situation has undergone a sweeping and significant change in the
last thirty years. The change in volume has naturally been one of
degree, not of kind. But the change in degree has been a profound
one—more so than is often admitted. It has been pointed out
occasionally, as a sedative to the fears aroused by the immense
immigration of the twentieth century, that while the positive
immigration has increased tremendously, it has not increased at so
great a rate as the population of the country. The ratio between
immigration and total population was higher in the early fifties and
early eighties than at any subsequent period. The assumption is that
if we could successfully assimilate the immigrants of the earlier
period, we certainly ought to be able to take care of those of to-day.
But the question of assimilation depends not only upon the ratio of
immigrants to total population, but upon the proportion of foreign-
born population already in the country. In this connection the
following figures are significant. The number of foreign-born to
100,000, native-born in the population of the country at the time of
the last seven censuses was as follows:
1850 10,715
1860 15,157
1870 16,875
1880 15,365
1890 17,314
1900 15,886
1910 17,227
It thus appears that the proportion of foreign-born, even at the time

of the census of 1900, after a decade of very slight immigration, was
much higher than at the time of the beginning of large immigration,
while the last census, after the enormous immigration of the past ten
years, shows a proportion of foreign-born higher than at any
previous census, except that of 1890. Now it is the proportion of
foreign-born to native-born which determines the assimilating power
of the nation, so that without this correction the comparison between
immigration and total population is inadequate and misleading. It is
as if a fireman whose steam boiler lacked a safety valve was warned
that his gauge was going up more and more rapidly all the time, and
he replied, “Never mind, the pressure is not coming in so fast,
compared to what I already have, as it was awhile ago.”
Another circumstance which affects the ability of the country to
assimilate immigrants, and in which there has been a marked change
during the history of immigration, is the ratio of men to land, upon
which much emphasis has already been laid. As the amount of
unappropriated and unsettled land diminishes in any country, the
need of new settlers also diminishes, while the difficulty of
assimilation and the possible evils resulting from foreign population
proportionally increase. In the case of the United States the first and
simplest comparison to make is that between immigration and the
total territory of the nation. In this, as in the subsequent
comparisons, it will be desirable to leave Alaska out of consideration.
The enormous extent of that inhospitable region, to which practically
none of our immigrants ever find their way, if included in the
reckoning, would simply confuse the issue. The gross area of the
United States, exclusive of Alaska and Hawaii, at the time of the
different censuses, has been as follows: 1790 and 1800, 827,844
square miles; 1810, 1,999,775 square miles; 1820, 2,059,043 square
miles; 1830 and 1840, the same; 1850, 2,980,959 square miles; 1860
down to the present, 3,025,600 square miles.[340]
Estimating the immigration before 1820 at 10,000 per year, and
using the official figures after that date, we find that the immigration
by decades from 1791 to 1910 was as follows:
1791–1800 100,000
1801–1810 100,000
1811–1820 98,385
1821–1830 143,439
1831–1840 599,125
1841–1850 1,713,251
1851–1860 2,511,060
1861–1870 2,377,279
1871–1880 2,812,191
1881–1890 5,246,613
1891–1900 3,687,564
1900–1910 8,795,386
Combining these two sets of figures, it appears that for each

immigrant coming to this country during the decades specified, there
was at the close of the decade the following number of square miles
of territory in the United States:
1800 8.278
1810 19.998
1820 20.927
1830 14.355
1840 3.437
1850 1.739
1860 1.205
1870 1.273
1880 1.076
1890 .570
1900 .824
1910 .347
This table illustrates forcibly the fact that from the point of view of
the need of new settlers immigration at the present time is a vastly
different matter from what it has ever been before in the history of
our country. This impression is strengthened if we make another
comparison, which is even more significant for our purposes, viz. the
relation of immigration to the public domain, that is, to the land
which still remains unclaimed and open to settlement. If there were
still large tracts of good land lying unutilized, and available for
settlement, as there have been in other periods of our history, we
could take comfort in the thought that as soon as the incoming aliens
caused too great a congestion in any region, the surplus inhabitants
would overflow, by a natural process, into the less thickly settled
districts. Let us consider what the facts have to show in this respect.
In 1860 there were, as nearly as can be estimated, 939,173,057
acres of land lying unappropriated and unreserved in the public
domain. In 1906 there were 424,202,732 acres of such land,
representing the leavings, after all the best land had been chosen. In
other words, for each immigrant entering the country during the
decade ending 1860 there were 374 acres in the public domain, at
least half of it extremely valuable farm land. In 1906, for each
immigrant entering during the previous ten years, there were 68.9
acres, almost wholly arid and worthless.
The fact that the immigrants in this country do not, to any great
extent, take up this unclaimed public land does not destroy the
significance of this comparison. As long as there was a strong
movement of the native population westward, it was not so much a
matter of concern, if large numbers of foreigners were entering the
Atlantic seaboard. And this was exactly the case during the middle of
the nineteenth century. This was the period of the great internal
migration to the new lands of the Middle West. In point of fact also,
at this time, many of these pioneers were actually immigrants. It is
scarcely necessary to say that nothing comparable to this is going on
at the present time. The frontier, which has had such a determining
influence on our national life, is a thing of the past. Of the
424,202,732 acres remaining in the public domain in 1906, only a
very small part consisted of valuable farm lands, such as existed in
great abundance when the Homestead Act was passed in 1862.
Evidence of this fact is furnished by the act recently passed allowing
homesteads of 640 acres to be taken up in certain sections of
Nebraska, where it is impossible for a man to make a living from less.
Not only are the incoming hordes of aliens not now counterbalanced
by an important internal migration, but there is an actual movement,
of noteworthy dimensions, of ambitious young farmers from the
United States to the new and cheaper wheat lands in Canada.
This set of conditions may be stated in another way by saying that
the United States has changed from an agricultural to a
manufacturing and commercial nation.[341] In the early nineteenth
century the rural family was the typical one, to-day it is the urban
family. Then the simplicity and independence of the farm gave
character to the national life; to-day it is the complexity and
artificiality of the city which govern. The nineteenth century was a
period of expansion. Particularly in the earlier part of it was the
subduing of new land the fundamental consideration of national
development. This was the period of internal improvements, the
building of roads and canals, and later of railroads. It was the
adolescence of the American people. At such a period the great
demand is for accessions of population, and it is no wonder that
many of the writers of that day were frank in their demands for the
encouragement of immigration. And even in the thirties and forties,
though the miserable shipping conditions and the large number of
incoming paupers aroused a countercurrent of opinion, still the
immigrants found a logical place on the great construction works of
the period, as well as on the vacant arable lands.
This period is past. The labors of the typical alien are not now
expended on the railroad, the canal, or the farm, but in the mines
and foundries, the sweatshops and factories. The immigrants of to-
day are meeting an economic demand radically different from that of
a century or half a century, yes, we may say a quarter of a century
ago.[342]
This change is further exemplified by the increased concentration
of population in cities which the United States has witnessed in the
past century. In 1790 there were only 6 cities in the United States
with over 8000 population each, containing 3.4 per cent of the total
population. In 1840 the percentage of population in cities of this size
was still only 8.4. But in 1900 there were 545 cities of over 8000,
counting among their inhabitants 33.1 per cent of the total
population. In other words, the ratio between city and country
dwellers (taking the city of 8000 as the dividing line) has changed
from one to twenty-eight in 1790 to one to two in 1900. At the same
time the average density of population of the country as a whole has
increased from 3.7 per square mile in 1810 to 10.8 in 1860, 17.3 in
1880, and 25.6 in 1900.
Hand in hand with these changes has come a sweeping change in
the scale of production, which must have an important bearing on
the immigration situation. The early immigrants, to a very large
extent, came into more or less close personal relations with their
employers, often working side by side with them on the farm or in
the shop. Now foreigners are hired by the thousands by employers
whom they perhaps never see, certainly never have any dealings
with, the arrangements being made through some underling, very
likely a foreigner himself. Working all day side by side with others of
their own race, or of other races equally foreign, and going home at
night to crowded dwellings, inhabited by aliens, and with a European
atmosphere, the modern immigrants have but slight commerce with
anything that is calculated to inculcate American ideas or contribute
any real Americanizing influence.
Mention of the declining native birth rate in the United States had
already been made (Chapter XI), with some consideration of the

Computer Vision Based Identification and Mosaic of Gramineous Grass Seeds Xin Pan Xuanhe Zhao Weihong Yan Jiangping Liu Xiaoling Luo Tana Wuyun

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Computer Vision Based Identification and Mosaic of Gramineous Grass Seeds Xin Pan Xuanhe Zhao Weihong Yan Jiangping Liu Xiaoling Luo Tana Wuyun

Uploaded by

Copyright:

Available Formats

Computer Vision based Identification

and Mosaic of Gramineous Grass

Computer Vision based Identification and Mosaic of

Autonomous Underwater Vehicles Jing Yan Xian Yang

Structural Reliability Approaches from Perspectives of

Multimedia Sensor Networks 1st Edition Huadong Ma Liang

Advanced Methods and Deep Learning in Computer Vision

Computer Vision for Microscopy Image Analysis Computer

Computer Vision: Object Detection In Adversarial Vision

Design and Applications of Emerging Computer Systems

Computer Vision based Identification

ISBN 978-981-16-3500-7 e-ISBN 978-981-16-3501-4

Jointly co-published with China Agricultural Science and Technology

Jointly co-published with China Agricultural Science and Technology

© China Agricultural Science and Technology Press 2021

The use of general descriptive names, registered names, trademarks,

This Springer imprint is published by the registered company Springer

Xin Pan (Corresponding author)

1.1 Digitization Grassland

1.2 Computer Vision Based Research Status of

1.2.1 Grassland Digitization Based on Remote

1.2.2 Grassland Digitization Based on Common

1.2.3 Researches on the Related Fields

1.3 Research Content

Burgos-Artizzu, X. P., A. Ribeiro, A. Tellaeche, G. Pajares, and C. Fernández-Quintanilla.

Granitto, P. M., H. D. Navone, P. F. Verdes, and H. A. Ceccatto. 2002. Weed seeds

Granitto, P. M., P. F. Verdes, and H. A. Ceccatto. 2005. Large-scale investigation of weed

Han A. T, X. H. Guo, Z Liao, Z. Q. Chen, and J. Q. Han. 2011. Classification of agricultural

Li, X. L. 2009. Reasearch on nondestructive determination of tea quality based on

Liu, L. B. 2010.Research of diagnostic technologies for rice leaf diseases based on

Potter, C. 2014. Monitoring the production of Central California coastal rangelands

Pourreza, A., H. Pourreza, M. Abbaspour-Fard, and H. Sadrnia. 2012. Identification of

Tellaeche, A., G. Pajares, X. P. Burgos-Artizzu, and A. Ribeiro. 2011. A computer vision

Wu, X. T. 2003. Study on methods and application of grassland desertification

Yu C. L. 2007. Lossless detecting research of Water quantity in the leaves based on

Zhang H. B. 2007. Study on variation of grassland vegetation based on multi-sensor

Zhang W. B., B. J. Lu, and S. Shi. 2009, Determination of vegetation coverage by

2. Forage Identification and

Xin Pan (Corresponding author)

2.1 Definition of Forage Automated Identification

Fig. 2.1 Images of forage seeds, plant and flowers

Fig. 2.2 The function modules of forage auto classification system

Data acquisition refers to the use of optical sensor equipment such

2.2 Automatic Identification of Herbage Based on

Fig. 2.3 Gramineous seeds a Roegneria tunczaninovii(Drob.)Nevski var. macrathera

2.3 Experimental Materials

Table 2.1 The Image database of gramineous grass seeds

No Scientific name Origin

2.3.1 Image Acquisition

2.3.2 Image Preprocessing

Fig. 2.5 Binary image

Subsequently, an open mathematical morphology operation is

Fig. 2.6 open operation of the seed

Fig. 2.7 ROI after segmentation

2.4 Seed Image DatabaseI

Fig. 2.8 Agropyron cristatum var.pectiniforme(Roem.et Schult) H Yang, a original

Duda, R. O., P. E. Hart, and D. G. Stork. 2001. Pattern Classification(2nd Edition)[M].

Granitto, P. M. , Navone, H. D. , Verdes, P. F. , & Ceccatto, H. A. . 2002. Weed seeds

Pourreza, A., H. Pourreza, M. Abbaspour-Fard, and H. Sadrnia. 2012. Identification of

Zhang, D. Automated Biometrics-Technologies and Systems[M]. Kluwer Academic

3. Identification of Gramineous Grass

Xin Pan (Corresponding author)

Therefore, considering the importance of gramineous grass in