Image Retrieval

e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science

( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:05/Issue:04/April-2023 Impact Factor- 7.868 www.irjmets.com
SATELLITE IMAGE RETRIEVAL AND DETECTION SYSTEM USING

MACHINE LEARNING ALGORITHEM
Divyansh Sinha*1, Prof. (Dr) Samta Gajbhiye*2
*1M.Tech Research Scholar In Computer Science Engineering, Shri Shankarcharya
Technical Campus Bhilai India.
*2Professor, Computer Science Engineering, Shri Shankarcharya Technical
Campus Bhilai India.
ABSTRACT
This study demonstrates a method for extracting any information from a satellite picture, including colours,
forms, textures, and other attributes. Using Texture filters and SVM to realize it (Support Vector Machine).
Important urban elements, including buildings and gardens, as well as rural characteristics, like native plants,
water features, and fields, may be identified using this image processing approach. Based on how many textures
are found in the image, Texel is used to represent the textures, which are then grouped into a variety of sets. In
order to retrieve the right picture using SVM, we need satellite images.
Keywords: Retrieval, Satellite Images, Colour And Texture, SVM (Support Vector Machine).
I. INTRODUCTION
Digital images are widely employed in a variety of fields, including business, law enforcement, fingerprint
recognition, healthcare, surveillance, engineering, fashion, architecture, and graphic design, as well as in
education, government, and historical research, among others. This would necessitate a rise in retrieval
precision and a fall in retrieval time. The prior approaches were solely based on Searching using only text, not a
visual feature. Several occasions, a single term linked to numerous photographs also results in incorrect
outcomes. The limitation of text-based retrieval is thus overcome by Content Based Image Retrieval (CBIR)
[DevyaniSoni, 2015]. In the early 1990s, content-based image retrieval first appeared. The primary goal of a
content-based image retrieval system is to search for and locate specific images within a huge database using
their visual contents, such as colour, form, and texture, among other things. According to Mujtaba Amin Dar
(2017), the two fundamental ideas used by Content Based Image Retrieval systems for picture retrieval are as
follows:
• Feature extraction
• Matching
Support Vector Machine (SVM): A SUPPORT Vector Machine (SVM) is a discriminative classifier that is officially
described by a separating hyperplane.
In other words, the algorithm generates an ideal hyperplane that classifies fresh samples given labeled training
data. This hyperplane, which divides a plane into two portions in two-dimensional space, has one class on
either side. The principle of structural risk minimization (SRM) is roughly applied in this situation. It produces a
classifier whose Vapnik-Chervonenkis (VC) dimension is minimised. An upper constraint on the generalisation
error rate is minimised via SVM. Error Rate is constrained by total training. Take the issue of splitting a set of
training vectors into two classes, for instance.image retrieval issue, where +1 indicates a positive example and -
1 a negative example. Support vector machines (SVMs) are supervised learning models and learning algorithms
that examine data and identify patterns for classification and regression analysis. A non-probabilistic binary
linear classifier is created by an SVM training method using a set of training examples that have been identified
as belonging to one of two categories. An SVM model is a representation of the instances, which are points in
space mapped in such a way that the examples of the various categories are separated by as obvious a gap as is
feasible. where the new exam subjects are then projected into that same space and classified according toWhat
side of the gap they are on is important to note [Jagbir Singh, 2016; TattaSugamya, 2016].
www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[5815]
e-ISSN: 2582-5208
II. LITERATURE REVIEW
Known a squery by image content and content based visual information retrieval is the system in which
retrieval is based on the content and associated information of the image. The XYZ and HSV algorithm on the
measurement of Euclidean distance for Content Based Image retrieval. The results evaluate that the proposed
algorithmis quiet better as compare to XYZ for set of dataset of trademark images on the basis of color, text and
logo.
Figure 1. H1 does not separate the classes. H2 does, but only with a small margin. H3 separates
them with the maximum margin
Katta Sugamya et al. A new two-step strategy in which first step is feature extraction using low level features
(color, shape and texture) while SVM classifier is used in the second step to handle the noisy positive examples.
Thus, an efficient image retrieval algorithm based on color-correlogram for color feature extraction, wavelet
transformation for extracting shape features and Gabor wavelet for texture feature extraction is proposed.
Nikita Upadhyaya et al. The focus is mostly on extracting features from the queried image and from the images
stored in the database for finding the similarity between these features to retrieve images which are similar
visually. CBIR becomes tougher once focus goes to reducing the semantic gap or the linguistics gap between low
level features and high level semantics.
A.Komali et al. CSIR can be developed by using K-Means algorithm for getting retrieval results of similar image
efficiently. By using K-Means algorithm, more number of iterations occurred. In order to reduce the number of
iterations we use codebook algorithm. This CSIR can be used in several applications such as photo sharing sites,
forensic lab, etc. CLARANS is the normal method which is used to reduce the bugs occurred in the existing
algorithms. Sanjiv K. Bhatia etal. The task may be simplified in part by correlating geographical coordinates to
observation but that may lead to omission of similar condition sin different regions. Work on an image search
engine that can eﬃciently extract matching image segments from a database of satellite images. This engine is
based on an adaptation of RISE (Robust Image Search Engine) that has been used successfully in querying large
databases of images.
III. METHODOLOGY
Designing an effective CBIR for satellite pictures using SVM is the proposed task. The precision, recall, and
accuracy of the results are assessed, and they are found to have improved as discussed in the result section.
CBIR: The Content Based Image Retrieval (CBIR) technique searches and retrieves digital images based on the
content of the images. Systems for retrieving images using content rather than language have been developed
in response to issues with text-based image retrieval. A group of methods known as content-based image
retrieval use autonomously generated picture attributes to find semantically relevant photos in an image
database. CBIR's primary objective is efficiency in image indexing and retrieval, which lowers the demand for
human intervention in the indexing process.
Creation of the Database: Some of the sample images are shown in figure 2
[5816]
e-ISSN: 2582-5208
Fig.2. The above figure shows the satellite images from the dataset
Feature Extraction: The proposed CBIR system for satellite images will be based on the extraction of colour
texture and shape features.
 Color-The colour feature is one of the most important and often used in CBIR. The colour feature is an
element of the image that cannot be changed in terms of orientation.
and size of the object, making it simple to analyse and extract. Color descriptor extracts the proportion of colour
within the query image and is used for similarity activities with database photos that have a large or less similar
percentage of colour to generate output pictures that are similar.
 Texture-Texture is defined as the appearance of a visual pattern in an image with the features of uniformity
that do not emerge from the presence of a single colour or intensity in an image. Textural qualities are assigned
to a specific section of an image or subimage
 Shape-Shape is one of the fundamental basic visual elements utilised to convey information about image
content. While analysing shapes within an image, a representation of an object or form is created by combining
shape border and interior content. Shape descriptors must efficiently locate similar shapes from a pool of
photos in order to retrieve images accurately.
SVM: The Support Vector Machines (SVM) algorithm is an approximation of the structural risk minimization
(SRM) principle. It generates a classifier with the smallest possible Vapnik-Chervonenkis (VC) dimension. SVM
reduces the upper constraint on the generalisation error rate. The total of training limits the error rate.
(A) Performance Measures
 PRECISION: It is the proportion of relevant records found to the total number of irrelevant and relevant

[5817]
e-ISSN: 2582-5208
records found. It is most commonly stated as a percentage. PRECISION = Number of Relevant Images Retrieved
/ Total Images Retrieved
 RECALL: It is the proportion of relevant records retrieved to total relevant records in the database. It is most
commonly stated as a percentage. RECALL=Number of Retrievable Images/Number of Retrievable Images in
the Database
 ACCURACY: ACCURACY:final_acc = 100*sum (diag (cmat))./sum(cmat(:));fprintf('SVM(1-against-
1):\naccuracy =%.2f%%\n',final_acc);
Through this code we get the values of the accuracy from the retrieval image
IV. RESULTS AND DISCUSSION
The image dataset description may be seen in Fig. 3, which is taken from Google Image Search. The image is
384*256, a jpg file with a size of 12.9KB.
Working Methodology: We will cover the project's working steps in this section
Performance Evaluation: In this section, we calculate the recall, precision, and accuracy of the retrieval system
in order to assess its performance. Precision assesses the system's ability to recover only the relevant models,
whereas recall measures the system's capacity to retrieve all relevant models. Accuracy indicates the worth of
an accurate image retrieval from the search.
Fig.3. Image in the data set

STEP 1: Initialization Phase
STEP 2: The data sets are loaded in this section

[5818]
e-ISSN: 2582-5208
STEP3: In this section we take an image from data set s and get the number of query images.
STEP4: After getting the query images we proceed It to get SVM of it.

[5819]
e-ISSN: 2582-5208
Table 1: Shows the result of the different values
S.no. Queryimage Retrieved Precision Recall Accuracy(%)
1. 10 0.1 0.02 85.20%
2. 15 0.15 0.03 85.20%
3. 5 0.05 0.01 86.40%
4. 8 0.08 0.16 89.00%
5. 10 0.1 0.02 84.60%
6. 12 0.12 0.24 89.20%
7. 10 0.1 0.02 87.60%

[5820]
e-ISSN: 2582-5208
8. 15 0.10 0.01 89.20%
TOTAL 0.8 0.7 87.05%

Table 2. The comparative analysis
S.No Precision Recall Accuracy
Simar deep Kaurand
0.6 0.5 67.75%
Dr.VijayKumar
Bangaetal. 2013
Proposedwork 0.8 0.7 87.05%
Comparative analysis
The enhanced outcome and effective value of precision, recall, and accuracy are inferred from the preceding
table.The outcome derived from the earlier work is more than the work completed by us.
V. CONCLUSION
In this study, we modelled a CBIR system that searches for pertinent images using a query image. The findings
demonstrate that SVM is the best technique for determining colour space for colour feature extraction.
Compared to other methods, it produces good outcomes.
ACKNOWLEDGEMENTS
I am very thankful to my guide and friend who help in research work in this section.
VI. REFERENCES
[1] Darshana Mistry Computer Engineering, Gandhinagar Institute Of Technology Color and Texture based
Image Retrieval Using SVM for Relevance Feedback.
[2] DevyaniSoni, K. J. Mathai. Department of Computer Engineering and Applications National Institute of
Technical Teachers’ Training and Research, Bhopal. An Efficient Content Based Image Retrieval System
using Text, Color Space Approach and Color Correlogram.
[3] Jagbir Singh Gill Assistant Professor Department of CSE, Chandigarh Engineering College Landran,
Mohali. CBIRof Trademark Images in different color spaces using XYZ and HSI Volume 6, Issue 5, May
(2016).
[4] Katta Sugamya, 2016. Suresh Pabboju A Cbir Classification Using Support Vector Machines March 03-
05, R. L. Jalappa Institute of Technology, Doddaballapur, Bangalore, India.
[5] Komali, A., Veera Babu, R. 2015. An Efficient Content Based Image Retrieval System for Color and Shape
Using Optimized K-Means Algorithm VOL.15 No.4, April.
[6] Lakhdar LAIB1 and SamyAit-Aoudia National High Schoolof Computer Science ESI, Algiers, algeria
Efficient Approach for Content Based Image Retrieval using Multiple SVM in yacbir
[7] Mujtaba Amin Dar, Ishfaq Gull, 2017. Sahil Dalwal Content Based Image Retrieval with SURF, SVM and
BAYESIAN Vol. 5, Issue 2, February.
[8] Nikita Upadhyaya and Manish Dixit Department of CSE/IT Madhav Institute of Technology and Science
Relating Low Level Features to High Level Semantics in CBIR, Vol.9, No.3 (2016).
[9] Sanjiv K. 2007. Bhatia1, Ashok Samal RISE-SIMR: A Robust Image Search Engine for Satellite Image
Matching and Retrieval ISVC 2007, Part II, LNCS 4842, pp. 245–254.

[5821]

Image Retrieval

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

Image Retrieval

Uploaded by

Copyright:

Available Formats

e-ISSN: 2582-5208

International Research Journal of Modernization in Engineering Technology and Science

SATELLITE IMAGE RETRIEVAL AND DETECTION SYSTEM USING

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

Fig.3. Image in the data set

STEP 2: The data sets are loaded in this section

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

1. 10 0.1 0.02 85.20%

2. 15 0.15 0.03 85.20%

3. 5 0.05 0.01 86.40%

4. 8 0.08 0.16 89.00%

5. 10 0.1 0.02 84.60%

6. 12 0.12 0.24 89.20%

7. 10 0.1 0.02 87.60%

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

8. 15 0.10 0.01 89.20%

TOTAL 0.8 0.7 87.05%

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

You might also like