Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

Plant Leaf

COE 121 FINAL PROEJCT PRESENTATION

Identification
Introduction
SIGNIFICANCE OF PLANTS
Statement of the
Problem
In agriculture, edible leaves of vegetables are important sources of energy
of animals and humans alike. To correctly identify weeds from actual
leaves of plants is an integral part of a farmer’s job to ensure the plants
grow correctly and create quality produce. Furthermore, with many
varieties of leafy vegetables farmers have to deal with, it can get very
confusing sometimes. With that, in mind, with the ever advancing of
technology, there could become a tool which helps farmers in that aspect
of their tough and grueling jobs.
Objectives
Specific Objectives:
General Objectives:

1. Identify features of a given image of a leaf


1. Identify locally abundant plants from their leaves
using Gabor filters
using image recognition.
2. Compare features with the database created
from image of leaves with classes identified.
Significance of the
study
The project could aid farmers in quickly, efficiently and
consistently identify plant species in the presence of a
multitude of plant species.

The project would also encourage creation of a catalog or


database of images of different plant species
indigenous to our country.

Lastly, the project will contribute in categorizing local and


native species of plants which is a very large task because
of the vast biodiversity of the country.
Related Works

PLANT LEAF LEAF LEAF RECOGNITION


RECOGNITION IDENTIFICATION OF VEGETABLES
USING GABOR USING FEATURE USING MATLAB
FILTER EXTRACTION AND
NEURAL NETWORK Used image recognition
Comparative study of Gabor techniques to an extensive
filters with different Paired simple feature number of plant species
parameters and how it affects extranction with adaptive
feature analysis neural networks for Caldito, N. , Dagdagan E. et al.
Chaki, J., Parekh, R. (2012) Singh, S., Bhamrah, M. (2016)
(2015)
Methodology Image Collection

Database Creation

Image Testing
Image
Collection
9 classes of vegetable leaves were collected. Each class has
20 images to be collected, 10 for training and 10 for testing

- Ampalaya
- Basil
- Cilantro
- Gabi
- Kangkong
- Laurel
- Lemongrass
- Malunggay
- Pechay
Database
Creation
Using the train images and the code, a database is created by
generating a Gabor Filter and filtering the image to create the
vector of the featues generated form the image. This feature
vector is then added to the database. This is done to 10
images of 9 different species of plants stated earlier.
Image Testing
The 10 images for test is used together with the Image Test
code. The image undergoes the same process as Database
creation where the feature vector is created using Gabor
Filters. Furthermore, the feature vector is not added to the
database and instead undergo KNN algorithm. KNN algorithm
calculates the distance by subtracting the feature vector from
every item in the database then getting the sum to find the total
distance. The database entry with the minimum value/distance
is then identified to find the class. This class can be said to also
be the same class of the inputted test image.
Incorrect (Basil) 0%
10%

Results
AMPALAYA
90% - Correct
10% - Incorrect (Basil)
Correct
90%
0%

Results
BASIL
100% - Correct
Correct
100%
Incorrect (Lemongrass) 0%
10%

Results Incorrect (Basil)

CILANTRO
30%

Correct

60% - Correct
60%

30% - Incorrect (Basil)


10% - Incorrect (Lemongrass)
0%

Results Incorrect (Laurel)


40%

GABI Correct
60%

60% - Correct
40% - Incorrect (Laurel)
Incorrect (Gabi) 0%
10%

Results Incorrect (Ampalaya)


10%

KANGKONG Correct
50%

50% - Correct
30% - Incorrect (Ampalaya)
10% - Incorrect (Gabi)
Incorrect (Basil)
30%

10% - Incorrect (Basil)


Incorrect (Lemongrass) 0%
10%

Results Correct

LAUREL
50%

Incorrect (Basil)

50% - Correct
40%

40% - Incorrect (Basil)


10% - Incorrect (Lemongrass)
0%

Results
LEMONGRASS
100% - Correct
Correct
100%
Incorrect (Lemongrass) 0%
10%

Results
Incorrect (Pechay)
10%

MALUNGGAY
Incorrect (Basil)
10%

70% - Correct
10% - Incorrect (Basil)
10% - Incorrect (Pechay)
Correct
70%

10% - Incorrect (Lemongrass)


Incorrect (Laurel)
0%
9.1%

Results
Incorrect (Lemongrass)
9.1%

PECHAY Correct
54.5%

50% - Correct
30% - Incorrect (Basil)
Incorrect (Basil)
27.3%

10% - Incorrect (Lemongrass)


10% - Incorrect (Laurel)
Problems Encountered
IMAGE QUANTITY AND QUALITY FEATURE VECTOR SIZE
Most plant species used are local so the Different image size create feature vector of
number of available images are small and the different size,so the vector size need to be
quality is not the best. This was not fully resolved to avoid erasing the current database
resolved and a compromise in sample size was and creating a different one. Resolved by
done simply resizing the image to create uniform
sizes
Conclusion
The objective of the project is to be able to
identify plants from their leaves using
image recognition. Combining a gabor
feature filter analysis for feature
detection and kNN for identification
resulted in a 71.11% (64 out of 90)
accuracy. Further improvements can be
made and the plant database could be
expanded to accommodate the large
number of plant species native to the
country.

You might also like