IJCRT22A6204

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

www.ijcrt.

org © 2022 IJCRT | Volume 10, Issue 6 June 2022 | ISSN: 2320-2882

DEEP LEARNING BASED PLANT DISEASE


PREDICTION USING CONVOLUTIONAL
NEURAL NETWORK
Satyalakshmi B1*, Ananthi J2, Dr. S. Uma2
1
PG Student, 2Assistant Professor, 3Professor,
Department of Computer Science and Engineering, Hindusthan College of Engineering and Technology, Coimbatore, INDIA-
641032.

Abstract: Plant disease diagnosis is a huge issue in increasing the production of yield in the agriculture sector. Identifying plant
diseases is crucial to prevent loss of yield and quantity of agricultural products. The recent advances in computer vision helps the
researchers in using the Artificial intelligence technology for finding and detecting the type of plant disease of the crop. In this
current research, a deep learning technology is used for detecting and to classify the type of disease through image processing
technique. Convolutional Neural Network (CNN) is utilized for image processing for detection of diseases, including image
acquisition, image preprocessing, image segmentation, and feature extraction and classification. This describes a method for
detecting the disease of plants by using the image of their leaves. It also describes the algorithm for extracting some segmentation
and functionality used in the detection of plant diseases.

Index Terms - Artificial Intelligence, Convolutional Neural Network, Image Processing.


1. INTRODUCTION
Plant pathology is the scientific study of diseases in plants caused by pathogens and environmental conditions. Organisms that
cause infectious disease include fungi, oomycetes, bacteria, viruses, viroids, virus-like organisms, phytoplasmas, protozoa, nematodes
and parasitic plants. The focus of this research is to develop a technology in the field of agriculture that is based on engineering
techniques. Crops are now subjected to a wide range of characteristics and diseases. One of the key traits/diseases is the damage
produced by the bug. Because pesticides can be hazardous to birds, they are not always shown to be effective. It also has an adverse
effect on natural animal food systems. Plant scientists frequently assess the damage to a plant (e.g., leaf, stem) based on the proportion
of the diseased region seen by naked eye on a large scale. This research presents an advanced strategy for employing image processing
to analyse plant diseases and characteristics. The strategies investigated are for boosting throughput and lowering inaccuracy in plant
disease detection caused by human specialists. Engineering technology is used to detect leaf illness, and mathematical theory is used
to interpret the results.
2. LITERATURE SURVEY
A smart agricultural system with the necessary infrastructure is a new technology that may increase the quality and quantity of
agricultural produce in the country, such as tomatoes. Since tomato plants are cultivated from various variables such as environment,
soil, and amount of solar radiation, the presence of the disease cannot be avoided. Deep learning has enabled improvements in the
field of computer vision in recent years, opening the door for camera-assisted illness detection as mentioned. This study produced a
novel technique for effective disease detection in tomatoes. To detect and identify leaf diseases, a motor-controlled image capture
box is utilised to record images of all four sides of each tomato plant. A specific tomato variety, Diamantemax, was used for testing.
The system is designed to identify diseases that are spot rot and target spotlights. Using controlled conditions on the affected and
healthy tomato plant leaf collection image dataset. We trained a deep convolutional neural network to recognize three diseases. In
farming, plant diseases are the first criminals. Another tomato disease can be encountered by agronomists and farmers. It can be
detected or the leaves can be located on the roots of stem fruits and plants. Common symptoms of plant damage are mycoses, bacteria,
viruses, and spots where nematodes are the source of leaves, brown or black lesions, last death of lower leaves, and lower leaves and
dark spots. It has a yellowing stem. This is caused by changes in the environment such as humidity every day in the world. Each
disease has another designated control or prevention to avoid the disease. Common methods used are cultural practices, the use of
disease resistant varieties and chemicals [1]. Real-time image processing is related to the typical frame rate of all processed frames
required after the image is captured. This paper proposes a real-time edge detection technique for identifying images and their
hardware trefoil disease (rubber leaves). The leaves of the three large rubber trees are the spots of Corynespora leaves, the spots of
the bird's eye and the disease of Collectotrichum leaf disease in the image comparison used in this study. Diseases on the leaves can

IJCRT22A6204 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org b581


www.ijcrt.org © 2022 IJCRT | Volume 10, Issue 6 June 2022 | ISSN: 2320-2882
be detected by edge detection using the Sobel edge detection algorithm. This is to compare the real-time edge detection results that
occur through the FPGA Cyclone IV E through a monitor that is compared to the Sobel Edge Detection Algorithm generated by
MATLAB. The algorithm is implemented in FPGA, and the result of the image on the VGA monitor can be processed in the
embedded system of many applications such as MATLAB simulation of object detection, X-ray image intensifier, character
recognition, etc. as it is now [3]. A diagnostic system for plant leaf diseases presented from color images using unsupervised neural
networks. The image is processed using the color and texture features. Disease Feature Extraction and Disease Classification: The
system consists of two main steps. The disease feature extraction process analyzes the appearance of features using a statistically
based gray-level simultaneous occurrence matrix and texture feature equations. The disease classification process deploys an
unsupervised simplified fuzzy ARTMAP neural network for disease classification types. It is used to test the classification
performance of four systems of vine leaf disease. Rust prevention is scab, downy mildew and disease-free. The desired result was
achieved with an accuracy of 90% or higher. The proposed system can be completely applied to diagnose other types of plant diseases.
Vegetables and fruits are one of the most important agricultural products in many countries. This is clear because the world always
needs more food. Product quality control is basically required to produce higher value-added products. One of the most important
effects of such A quality is plant disease. Therefore, minimized plant diseases provide a significant improvement in product quality.
It requires the ability to detect diseases and diagnose plant diseases in a timely, accurate and responsive manner to minimize damage.
Today, precision agriculture and smart farming technologies are widespread and are being developed in two aspects: product growth
and quality control of the most agricultural products. Image processing and computer vision are the most effective technologies for
different types of applications. This work uses color images and disease feature analysis to propose a plant leaf disease diagnostic
system. Simultaneous treatment of the I and H components allows the appearance of features to be effectively significant. The gray
level simultaneous occurrence matrix and texture feature equations are used as statistical data for fuzzy simplified ARTMAP to
completely classify image types of vine leaf disease. Results are desired and can be appropriately applied to the detection and
classification of leaf diseases in real-world plants [4]. Rice is the most significant crop in Asian nations. Most people eat rice since it
is considered the staple diet of Asian countries. Many illnesses impair rice yields, causing farmers to lose money. This approach
suggests a technique for detecting blast furnace and brown spot infections. To categorise data, the global threshold approach and the
k-nearest neighbour classification were utilised. Consequently, the recommended strategy was successful. Brown spot disease is
discovered utilising image processing and pattern recognition algorithms in this method and rice blast. Plant diseases may
substantially reduce output and even kill crops. Direct and indirect losses are both types of losses. Reduced plant stands,
accommodation, spotted kernels, fewer per plant, and a general decline in smaller grain and plant output are examples of direct losses.
The expense of fungicides needed to control the illness is one example of an indirect loss. This use has various limitations, including
the high cost and low return associated with cultural practices related with disease reduction. In terms of greatest output, this is
ineffective. Image processing has been used to develop methods for identifying leaf diseases in agricultural plants. The author begins
by photographing the diseased leaf with a digital camera and then converting the RGB picture to an HSV colour model. As there is
no lack of infected regions, the segmentation of the infected area is further separated into several significant information patches
chosen by the author, taking into consideration the HSV colour space and the partial sound components of the segmentation. GLCM
was used to extract the statistical function. Based on typical morphological changes, identify, and diagnose leaf diseases in agricultural
plants. Images were preprocessed, and image contrast analysis histograms were created by performing threshold modifications. The
traditional fuzzy C-means (FCM) method can be used for segmentation. Color, shape, and texture aspects of images are employed
for segmentation. Artificial neural networks (ANN) should be classified [5]. Image capture, picture preprocessing, image
segmentation, feature extraction, and classification are all phases in detecting and identifying plant leaf diseases. This method can be
used in the description and leaf disease classification image segmentation algorithms to preprocess the images used in the study of
automatic identification and leaf disease classification algorithms for various plants. Therefore, fast, low cost and accurate methods
have practical significance for large farms to consider automatic identification and identification of diseases from plant leaves. This
decision support system (DSS) is needed to establish a call center, and farmers need to give detailed information about the foliage of
the plant verbally. DSS based on image processing helps improve agricultural productivity. This study proposes a system focused on
the detection and identification of diseases that are useful for decision making. The proposed system includes four main stages:
pretreatment, segmentation, feature extraction and classification. In this method, we focus on image classification techniques that are
different from image splitting [6].
3. RESEARCH METHODLOGY
3.1 Existing System
It is planted disease identification is an important issue in the existing system. Generalized problem identified plant leaf disease
makes your computer more intelligent and reducing the workload of humanity. Something in common with these two works have
a clear-cut picture of the lesion. The problem can be resolved with the diversity of data limitations. The difference is that the leaves
can be converted into a healthy image to various types of diseases. This is the image of the conversion method to the image from
the image, but the existing one is converting the image from the noise to generate a marked image conversion. The same type of
image. There is a problem with the difficult process of classification or regression of noise on the existing system.

3.2 Proposed System


Deep learning is an artificial intelligence subfield. In recent years, the benefit of extraction of autonomous learning and function
has garnered a wide variety of academic and corporate interest. Image and video processing are commonly employed in speech
processing and natural language processing. Simultaneously, it is artificial deep learning of the application plant disease recognition
and pest range assessment of plant disease recognition that it is possible to avoid the disadvantages caused by selection, the hot spot
of research in the field of agriculture plant protection, such as and it has become a disease of the spot is equipped with a more
feature extraction of objective makeup plant diseases, and improve the exchange rate.of efficiency and technology of research. The
Proposed system is as shown in Figure.1.

IJCRT22A6204 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org b582


www.ijcrt.org © 2022 IJCRT | Volume 10, Issue 6 June 2022 | ISSN: 2320-2882

Dataset Pre-processing data Feature extraction

Feature selection

Predict the result Classification using


CNN

Figure 1: Proposed diagram


3.2.1 Use case diagram
A use case diagram is a graphical representation of the interactions between system parts as shown in Figure.2. The
intended behaviour and the means for achieving it will be specified in use cases. Once defined, use cases can have both
written and visual representations.

Input image data

Preprocessing

Plant data
set
Segmentation

Output
Feature extraction

Classification

Figure 2. Use case diagram


3.2.2 Sequence diagram
A sequence diagram as shown in Figure.3. depicts distinct processes or objects that exist concurrently as parallel vertical
lines (lifelines), and the messages passed between them as horizontal arrows in the order in which they occur.

3.2.3 Activity diagram


Another significant diagram in UML for describing dynamic characteristics of the system is the activity diagram. An
activity diagram is essentially a flowchart that depicts the flow from one activity to another. The action can be defined
as system functioning. The control flow is directed from one operation to the next. The above Figure.4 represents the
flow from one activity to another activity, the activity starts from input leaf image through digital camera, and then
input leaf is preprocessed and extract the features like color, shape, texture and so on. Now, the processed image is
classified as Normal or Abnormal, if Abnormal is found in the leaf, then remedies will be suggested.

IJCRT22A6204 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org b583


www.ijcrt.org © 2022 IJCRT | Volume 10, Issue 6 June 2022 | ISSN: 2320-2882

Input Preprocess Segmentation Extraction Classification

Upload image
Preprocess
image Leaf Disease
Analysis images prediction

Segmenting
images Disease or
healthy

Figure.3 Sequence diagram

Input leaf images Preprocessing images Feature extraction

Diseases

CNN Classification

Healthy

Figure.4. Activity diagram


3.2.4 Data flow diagram
A data flow diagram is the graphical illustration as depicted in Figure 4 of the "flow" of data through an information system
that models the process characteristics of the system. A DFD is frequently used as a preliminary stage to produce an overview
of the system that may subsequently be expanded.
3.3 Dataset for plant leaf
The data set collection is the plant leaf disease image data collected from the information on the performance. Then, data
cleaning and data reduction are done from the collected plant leaf dataset records. These data of the plant disease, plant leaf
information for the leaf size, leaf color, quality and then plant characters are collected from the dataset performance.

4. Results and Discussions

4.1 Process of data preprocessing using Gaussian filter


There is usually no complete and consistent data in the real world, which cannot be used for direct data mining or the effect is
not ideal. To improve the quality of data pre-processing, Gaussian filter is used data cleaning and data integration, data
conversions, and data compression are the many ways of preprocessing data.

4.2 Process of data cleaning


The plant disease using correction data clearing process means identify the record set, table, database, incorrect, data inaccurate
records, please replace, then replace, delete data. This is divided plant disease prediction analysis has been used in the process
for the information theory in the plant leaf for change color and sizes is divided into types are to predict the data and descriptive
data cleaning prediction deployed the data structure.

4.3 Data Reduction process


The purpose is to reduce the data with a small compression volume data set, while maintaining its original integrity. This is an
effective method but gives similar result this is useful for data analysis and the data reduction can be used as a subset of the
evaluation function by the classifier CNN algorithms which are used for the prediction of result. It takes advantage of the
improved accuracy of selecting the functions related to the CNN algorithm performance.
4.4 Classification using Convolution neural Network
The classification is performed using the proposed method Convolution Neural Network (CNN) performance on the plant
disease prediction. Prediction of diseases includes image data, data preprocessing, data cleaning, data reduction and
classification. This describes a method for detecting the disease of plants by using the image of their levels. It also describes
IJCRT22A6204 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org b584
www.ijcrt.org © 2022 IJCRT | Volume 10, Issue 6 June 2022 | ISSN: 2320-2882
the algorithm and functionality used in the prediction of plant diseases. The plant leaf is to be classified for the prediction and
classification. The result is input image dataset for preprocessing and data clean and reduction for the data performance.

Figure 5 Python code for Implementing Package

Figure 6 Python code for Dataset implemented

Figure 7 Code for testing scanned images

IJCRT22A6204 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org b585


www.ijcrt.org © 2022 IJCRT | Volume 10, Issue 6 June 2022 | ISSN: 2320-2882

Figure .8 Result for Training images

Figure 9 CNN-soft-max activation function

Figure 10 Prediction of effected leaf images


5 References
[1] De Luna, R. G., Dadios, E. P., & Bandala, A. A. (2018)"Automated Image Capturing System for Deep Learning-based
Tomato Plant Leaf Disease Detection and Recognition", TENCON 2018 - 2018 IEEE Region 10 Conference.
[2] Yusoff, N. M., Abdul Halim, I. S., Abdullah, N. E., & Ab. Rahim, A. A. (2018)"Real-time Hevea Leaves Diseases
Identification using Sobel Edge Algorithm on FPGA A Preliminary Study", IEEE Control and System Graduate Research
Colloquium (ICSGRC).
[3] Khitthuk, C., Srikaew, A., Attakitmongcol, K., & Kumsawat, P. (2018)"Plant Leaf Disease Diagnosis from Color Imagery
Using Co-Occurrence Matrix and Artificial Intelligence System”, International Electrical Engineering Congress
(IEEECON).
[4] Goel, N., Jain, D., & Sinha, A. (2018)"Prediction Model for Automated Leaf Disease Detection & Analysis", IEEE
International Advance Computing Conference (IACC).
[5] Suresha, M., Shreekanth, K. N., & Thirumalesh, B. V. (2017)"Recognition of diseases in paddy leaves using knn
classifier”, International Conference for Convergence in Technology (I2CT).
[6] Dhaware, C. G., & Wanjale, K. H. (2017)"A modern approach for plant leaf disease classification which depends on leaf
image processing", International Conference on Computer Communication and Informatics (ICCCI).
[7] Francis, J., Anto Sahaya Dhas D, & Anoop B K. (2016)"Identification of leaf diseases in pepper plants using soft computing
techniques", Conference on Emerging Devices and Smart Systems (ICEDSS).
[8] Mattihalli, C., Gedefaye, E., Endalamaw, F., & Necho, A. (2018)"Real Time Automation of Agriculture Land, by
automatically Detecting Plant Leaf Diseases and Auto Medicine", International Conference on Advanced Information
Networking and Applications Workshops (WAINA).

IJCRT22A6204 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org b586


www.ijcrt.org © 2022 IJCRT | Volume 10, Issue 6 June 2022 | ISSN: 2320-2882
[9] Herdiyeni, Y., Bakhtiar, T. S., Dewanto, V., Tjahjono, B., Siregar, B. A., & Oliveira, L. S. S. (2017) "Automatic
identification of acacia leaf diseases in plantation forests using wavelet energy and Shannon entropy”, International
Conference on Information and Communication Technology Convergence (ICTC).
[10] Zhang, X., Qiao, Y., Meng, F., Fan, C., & Zhang, M. (2018)"Identification of Maize Leaf Diseases Using Improved Deep
Convolutional Neural Networks", IEEE Access,

IJCRT22A6204 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org b587

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy