Identification and Classification of Disease For Tomato Plant Using IP and CNN

Volume 4, Issue 5, May – 2018 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Identification and Classification of Disease for

Tomato Plant using IP and CNN
Vrushabh Shende Dr. Mahendra Pund
Computer Science and Engineering Computer Science and Engineering
PRMIT&R, Badnera PRMIT&R, Badnera
Amravati, India Amravati, India
Abstract:- Tomato Leaf disease Identification is basically II. RELATED LITERATURE

a system application which uses IP (Image Processing)
and CNN (Convolution Neural Network) techniques. This section of Literature Survey eventually reveals
Image processing is very important part of the system some facts of Tomato Plant Disease Identification System
which includes image acquisition and image conversion. based on the analysis of many authors work as follows:
The convolution neural network is use for classification.
To train the model for the CNN architecture, data is Debashis Ghosh [1], aimed to develop a mobile vision
collected from the authenticated online source. 4000 based plant leaf recognition system which monitor crop
samples leaf images of four classes, including the healthy diseases having different patterns. It classifies its suitable
leaf images are used to train the model. At the end the class which can be used to help the botanical students in
result shows some tomato plant diseases such as Late their research work. The feature factor used in the system
Blight, Yellow Leaf Curl and Bacterial Spot are detected. planned by the author in this work has enormous
computation cost. The quality of captured image is affected
Keywords:- Image Processing, Convolution Neural by the shadow and season.
Network, Late Blight, Yellow Leaf Curl, Bacterial Spot.
S. Bani-Ahmad [2], proposed an improved solution for
I. INTRODUCTION classification of leaf diseases by using algorithms like k-
means and neural networks. Otsu’s method are used in
Agriculture has much more importance than simply a segmentation to mark green pixels and then the boundary
means to nourish ever growing populations. However, plant pixels are removed. In proposed work clustering and
diseases are threatening the livelihood of this important classification of diseases is done.
source. Due to the plant diseases there are major economic
losses in agriculture and forestry. Therefore, early detection S. Arivazagan [3], proposes to develop a system that
and identification of plant diseases plays the utmost automatically finding the solutions for plant leaf disease by
important role to take timely Measures. There are several analyzing the texture of leaf. At the initial stage diseases are
ways to detect plant pathologies. Some diseases are not analyzed. The identification rate is short which needs to be
identify at the early stage. They need some time to identify optimized to avoid the misclassification of the variable
the disease. In these cases, it is necessary to perform symptoms of the plant disease.
sophisticated analysis, usually by means of powerful
microscopes. In some cases, the signs can only be detected III. PROPOSED ARCHITECTURE
in parts of the electromagnetic spectrum that are not visible
to humans. We develop a system which helps the farmers to
identify the tomato plant diseases. We are using Image
The aim of this research is to develop a user friendly Processing for extracting the features and Convolution
system for the farmers that will help farmers in finding out Neural Network for classification of the tomato plant
diseases of tomato plant without bringing an expert to the disease. For that purpose we are using the dataset of total
field. We are providing remedies of tomato plant diseases to 4000 images of tomato plant leaf. These 4000 images are
the user. For this purpose we first get the image from user classified in four categories as follows: 1.Late Blight
then perform feature extraction on that image. For extracting Disease 2.Yellow Leaf Curl Disease 3.Bacterial Spot
the features we use conversion of image and resize the Disease 4.Healthy. Each category of disease contain 1000
image. The convolution neural network is use for classifying images.
the diseases. The system can uses one image databases
which contain 4000 images, 70% of the data is use for For classification of images we are using convolution
training the model and 30% of data is use for testing the neural network. For this first we have to train the model then
model. test the accuracy of that model by testing the model. So 70%
of data is use for training the model and 30% of data is use
for testing the model. The main objective of this system is to
identification of tomato plant disease and provide remedies
for that disease.
IJISRT19MY246 www.ijisrt.com 72
ISSN No:-2456-2165
Fig 1:- Image Dataset
 Methodologies: database, higher the size of database higher the accuracy of

the model. There are total seven layers present in this
A. Image Processing for Feature Extraction: convolution neural network. First layer is input layer and
First process after getting the image from camera or the last layer is output layer. There are total five hidden
any other source is to process the image for feature layers present in this convolution neural network. The
extraction. We get the image of different resolution so first output layer has four neuron present in it for classification
we need to resize the image of 50 by 50 pixel. Then of the disease. Each neuron represent the unique disease of
conversion from RGB to HIS model and flood fill tomato plant. Such as one neuron for Late Blight Disease
algorithm is use for feature extraction. and others neurons for different diseases. In this model we
are using the learning rate of 1e-3 that is 1*(10 to the power
 Conversion from RGB to HIS model: -3) for training the model. After finding the tomato plant
Basically the input image is in format of RGB model. diseases, system provides the remedies of this disease.
The RGB model, which encodes by their intensity of color
of Red, Green, and Blue. Each color red, green and blue
defines their intensity of color by an integer value between
0 to 255. The HSI model, which encodes colors according
to their Hue, Saturation, and Intensity. The HIS model is
used by some graphics software and image monitoring
system alongside the RGB Model.
 Flood Fill Algorithm:

For segmentation purposed flood fill algorithm is very
important. Flood fill algorithm is also called region
growing algorithm. At a first view, it looks very simple
algorithm to implement but it has very interesting
implementation for segmenting the image. Flood fill
algorithm finds the same color area from starting point to
last point with same color. Flood fill algorithm includes 4
or 8 direction of neighbors that check for condition to same
area color and perform the operation recursively for
neighbors in case they passed that condition.
B. CNN for Disease Classification:

The classification of disease is depend on the Fig 2:- System Architectural Diagram
model that we are created by training the 70% of the data.
The accuracy of the model is depend on the size of
ISSN No:-2456-2165
IV. EXPERIMENTATION AND RESULTS V. CONCLUSION
The dataset is divided 70% for the training and 30% The system is capable of detecting the disease at the
for testing. Different models with different architectures earlier stage as soon as it occurs on the leaf. Hence saving
and learning rate are tested. ReLu activation function is the loss and reducing the dependency on the expert to a
used since researches have shown that ReLU result in faster certain extent is possible. It can provide the help for a
training. The result obtained is shown below. person having less knowledge about the disease. The
system periodically monitors the cultivated field. Tomato
plant diseases are detected in early stage by using Image
Processing and CNN techniques. Convolution Neural
Network techniques are used to train the model which helps
to identify the diseases. The remedies are provided to the
farmer to control the disease.
VI. FUTURE SCOPE
This research has been considered only for three

diseases and further it can be extended for various diseases.
In future it can be extended to find the percentage of the
infection in accordance with the area affected. In future the
system may be implemented by adding extra services like
nearby government stores, price list for the pesticides,
nearby open market and many more.
Fig 3:- System GUI for Identification of Disease REFERENCES
Fig 3 shows the ‘Status’ of the leaf as HEALTHY or [1.] Shitala Prasad • Sateesh K. Peddoju • Debashis Ghosh
UNHEALTHY. If the plant is UNHEALTHY, then the “Multi-resolution mobile vision system for plant leaf
name of disease is specified. Knowing the disease is not disease Diagnosis”, Springer-Verlag London 2015
enough, so remedies for the disease is provided. To get the [2.] H. Al-Hiary, Z, M. Braik and S. Bani-Ahmad,” Fast
remedies the button as ‘Remedies’ is shown in the above and Accurate Detection and classification of Plant
figure. By pressing this button a process of searching Diseases” International Journal of Computer
appropriate decease remedies carried out simply by Applications (0975 – 8887) Volume 17– No.1, March
switching to block of matched index value out from 2011.
analysis algorithm. [3.] S. Arivazagan, R. Newlin Shebiah*, “Detection of
unhealthy region of plant leaves and classification of
Thus after detecting the disease, remedies for the plant leaf diseases using texture features”, CIGR
disease is displayed on the resulting screen provided in the journal March-2013.
following Fig 4. [4.] Jyotismita Chaki,, Ranjan Parekh, “Plant Leaf
Recognition using Shape based Features and Neural
Network classifiers.”
[5.] Mr. Sachin B. Jagtap1, Mr. S.M. Hambarde2
“Agricultural Plant Leaf Disease Detection and
Diagnosis Using ImageProcessing Based on
Morphological Feature Extraction” university Of
Pune , Jspm’s Jscoe, Pune, India.
[6.] Sindhuja Sankarana, Ashish Mishraa, Reza Ehsania,∗,
Cristina DavisbA review of advanced techniques for
detecting plant diseases .
[7.] 1Suneeta Budihal, 2Sandhya R., 3Soumya D
Hajawagol, 4Soumya R Navi B.V.B.C.E.T., Hubli,
India,” Detection of Disease in Tomato Leaf”.
[8.] C. Stroutopolos, I. Andreadis, Multithresholding of
color and gray-level images through a neural network.
[9.] Karimi, Kaivan, and Gary Atkinson. "What the IoT
Fig 4:- System GUI for Remedies of Disease wants to become a reality." White Paper, FreeScale
and ARM (2013).
The remedies are useful to avoid the disease, so the [10.] Gubbi, Jayavardhana, et al. "IoT: Architectural
plant will be Healthy. elements ,a vision and future directions." Future
Generation Computer Systems 29.7 (2013): 1645-
1660.
ISSN No:-2456-2165
[11.] "Understanding the Internet of Things (IoT) “, July
2014.
[12.] P. Y. Simard, D. Steinkrans and J .C. Platt, “Best
Practice for Convolutional Neural Network Applied
to Visual Document Analysis”, The 7th ICDAR,
(2003).
[13.] B. Kowlek, “Detection Face Using CNN and Gabor
Filters, Artificial Neural Networks”, Biological
Inspirations, vol.3696, (2005)
[14.] Doug Tidwell, James Snell, Pavel Kulchenko
“Programming Web Services with SOAP”,First
edition, December 2001.
[15.] Vrushabh Shende, Mahendra Pund “IoT Based
Preventive Crop Disease Model Using IP and
CNN”,Volume 6, Issue III, International Research
Journal of Engineering and Technology (IRJET) Page
No: 1270-1272
[16.] M. A. Pund, A. P. Jane “Recognition of Similar
Shaped Handwritten Marathi Characters Using
Artificial Neural Network”, Volume 3, Issue II,
International Journal of Engineering Research and
Applications (IJERA) Page No.63-67
[17.] M. A. Pund, M. S. Makeshwar, N. K. Rajgure
“Object Identification using Neural Network”
International Journal of Computer Science and
Application.

Identification and Classification of Disease For Tomato Plant Using IP and CNN

Uploaded by

Copyright:

Available Formats

Identification and Classification of Disease For Tomato Plant Using IP and CNN

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

Identification and Classification of Disease For Tomato Plant Using IP and CNN

Uploaded by

Copyright:

Available Formats

Volume 4, Issue 5, May – 2018 International Journal of Innovative Science and Research Technology

Identification and Classification of Disease for

Abstract:- Tomato Leaf disease Identification is basically II. RELATED LITERATURE

Fig 1:- Image Dataset

 Methodologies: database, higher the size of database higher the accuracy of

 Flood Fill Algorithm:

B. CNN for Disease Classification:

VI. FUTURE SCOPE

This research has been considered only for three

Fig 3:- System GUI for Identification of Disease REFERENCES

You might also like