Volume 7, Issue 9, September – 2022 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Plants Diseases Detection:

(A Brief Review)
1st Siddhant Dev Mannu 2nd Dr. Sourabh Jain
Department of Information TechnologyIndian Institute of Department of Information TechnologyIndian Institute of
Information Technology, Sonepat Information Technology, Sonepat
Haryana, India Haryana, India

Abstract:- Diseases in plant/crops cause major losses in exact. In plants, a few general diseases are seen like yellow
terms of economy and productions as well as reduction in spots and brown, early and late burn, and others are viral,
quantity and quality of agricultural goods/products. Plant bacterial and fungal diseases[7]. Image processing is used for
disease affect the growth of their species. About 58% of estimating impacted area of diseases and to decide the
Indian population, agricultural is the primary/elementary distinction in the color of the impactedregion[6].
source for their livings. Since, 70% of Indian Economy is
highly depends only on agricultural yields, so there is need II. BASIC STEPS FOR IMAGE PROCESSING
to control the losses done by the diseases. To preventing
the losses in quantity and improving the quality of A. Image Acquisition
agricultural goods, therefore early recognizance is very By and large, the task of reestablishing a image from the
important. Still traditional methods being followed by the beginning is defined to be an image acquisition. The basic
experts such as naked eye observation that is time- state that is required consistently for the grouping of work
consuming and expensive. In this paper, image processing process is Image acquisition because of the need of a image
techniques and some others technique and methods are in image processing. The image is regular which is acquired.
used for the detecting the diseases in plant are disscused. First step, is to capture the plant leaf by using digital camera
It’s includes several steps like in image processing are: or mobile phone with high resolution for better quality and
image acquisitions, image pre-processing, image better results[10]. The captured image is in RGB (Red,
segmentations, feature extractions and classifications of Green, Blue) formats.
plant diseases.
B. Image Pre-processing
Keywords:- Plants Diseases Detection, Image Processing, In second step, captured image is pre-processeds to
Classifications. enhance the quality of image or to remove the noise inimage
by removing undesirable distortions and enhances images
I. INTRODUCTION details which is necessary for further processing and analyze
task. Different techniques are used in image pre- processing.
India is a developing and cultivated country and its Image clipping is based on ROI (Region Of Interest)[10].
economy and populations highly depends on agriculture
products. Agricultural represents roughly 17%of total GDP, Image smoothing helps to improving the image quality
giving more than 60 % of the populace with anddone by using smoothing filter. Image enhancement is
employment[7],[12]. In field of agricultural, detection of helpsto increasing contrast of the images to get the better
diseases parts in plant plays an critical role. In the modern results. RGB image is converted into grey-image by using
times, farmers are as yet dealing with issues to finding the colour conversion equation are written below:
suitable pesticides and herbicides to plants[7]. The current F(x)=0.2989*R+0.5870*G +0.114 (1)
technique for plants diseases detection is only naked eye
observation in which experts can identify the disease present After that the histogram equalization is done which
on plants. What's more, it is extremely challenging to the assign the intensities of image is fuctional on the images to
large teams of experts to monitoring the diseases in plants improve the leaf disease image. The cumulative distribution
continuously. Mostly in backward area, Framer do not have functions that distribute the intensity values.
any ideas and facility regarding how to contact the experts.
In case farmer contact to expert it would be pricey and time- C. Image Segmentation
consuming. And it is much difficult to the farmer to overcome Image segmentation is that method in which an image is
with new plant diseases. Therefore looking for new method splits into various parts of same feature or having some
which would be fast and less expensive and must be accurate, likeness that is expressive and easy to analyse[6]. The
and which detect the diseases that appear on the leaf of plant segmentation should possible use, converting RGB to HSI
automatically[14]. model, Thresholding, K-mean clustering, Otsu's algorithm
and etc.
Plants diseases identifications by visual way is more
difficult task and simultaneously, less accurate and shouldbe
possible just in limited regions. If automatic detectionmethod
is used then it takes less time, less effort and become more

IJISRT22SEP051 www.ijisrt.com 98
Volume 7, Issue 9, September – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
a) RGB to HSI model: edge in a image is a critical neighborhood changes in the
The RGB images is converted into the HSI (Hue, images intensity. Morphological tasks apply an organizing
Saturation, and, Intensity) model for segmentation[8]. component to an input image, making a result image of a
Boundary and spot detection helps to track down the similar size and shapes[8].
contaminated part of the leaf. The thought of connectedness
of pixels is expected for distinguishing the boundary. After Interesting features of an image from where the required
that boundary and spot detection’s algorithm is done. information’s are extracted is called as feature extraction. The
component of the Region Of Interest (ROI) will be smaller
b) Thresholding; than the original images. GrayLevelCo-occurrenceMatrix
Image thresholding is an easy and effective method of (GLCM) is one of the best methods for texture analysis[10].
convey a image into a frontal area and foundation. This is a It uses 2nd order statistics methods for estimating the image
sort of image segmentation method that isolates object by properties. Resultant will be the amount of event of the pixel
changing it from grayscale images into twofold images. This with specific intensity in the spatial space. Size of the GLCM
is most effective in image with elevated degrees of variety. will be founded on the quantity of gray level.
Selection of threshold Values is the key attribute in the
processing[14]. Feature extractions is the process which is done after
segmentation. As Demonstrated by the segmented data and
c) K-means Clustering: predefined dataset a few features of the image can be
The Kmeans clustering is utilized for characterization remove. This extraction could be the any of structural,
of objects based on set of features into Knumber of classes. statistical or signal processing. Gray Level Co-occurrence
And its utilized to get the require feature from the leaves. Matrices (GLCM), Color co-occurrence, Spatial Gray-level,
Furthermore, the feature of the contaminated part of leaves SGDM technique, Gabor Filters, Wavelet Transform[14],
are improved by getting the contrast images of the leave and and many more analysis have a fewtechniques utilized for
from the neural network[12].The primary point of this feature extraction.
algorithm is to limit the amount of distances between the
objects and their relating clusters. a) Color Co-occurrence Method:
The color cooccurrence matrix for various spatial
d) Ostu’s thresholding: distances is characterized in light of the most extreme/least
Ostu thresholding pick out the threshold values to of color parts between the three parts (Red, Green, Blue) of a
minimizing the intra class difference of the threshold high pixel. The GLCM function distinguish the texture of a image
contrast pixels. Ostu algorithm automatically execute the by computing how frequently sets of pixel with explicit
cluster based image thresholding, or the reducing the gray- values and in a predefined spatial relationship happen in a
level image to a Binary Image by adjusting all pixels. image, making a GLCM, and afterward removing statistical
measurement from the matrix. The GLCM matrix is a
The working of Ostu algorithm are follows: methods to evalute the spacial relationship of a image pixel.
Step-1: According to the ostu threshold, distinct pixels into Co-occurrence matrix estimates the possibility of appearance
two different clusters. of sets of pixel values situated at distance in image. And
Step-2: Then find out mean of the each cluster. Step-3: Square this algorithm is said to be GLCM. The matrix characterizes
the variance between the means. the possibility of joining two pixels. In this GLCM method
Step-4: Multiply the quantity of each pixels in asingle cluster both texture and color are taken into record to get a unique
time the number in other. feature for that image. For that the RGB image is changed
over into the HSI model. For the texture calculation, the
The contaminated leaf shows the side effects of the SGDM matrix are produced and uses GLCM function so that
disease by changing the shade of the leaf. Thus, it would be feature is calculated.
possible to detect the contaminated part of leaf by using
greenness on leaves. Than the RGB, part are extract from E. Classifiers:
the image. The thresholds is determined using the Otsu Classifiers are utilized to recognize and classify the
methods. Then the green pixels is concealed and taken out if various diseases that arise on plant leaves based on acquire
the green pixel intensities are lesser than calculated threshold. features. There are the linear and nonlinear classifier that
have been used in past work to identify the plants diseases
D. Feature Extraction are Support Vector Machine (SVM), KNearest Neighbors
To identification of an objects, feature extraction have (KNN),Artificial Neural Network(ANN),Convolutional
an important role. There are many application of an image Neura l Network(CNN),Radialbasis functions, Probailistic
processing where the feature extraction are used. Texture, Neural Net work(PNN) and Back Propagation Network and
color, edges, morphology, and so on are the elements which so on[14].There are some platforms such as MATLAB that
canbe used to detect plant diseases and these elements are are used toprepare and test these classifiers.
considered in paper [8]. They have observed that
morphological outcome gives improved outcome than the
other elements. Texture means how the color is dispersed in
the image, the harshness, hardness of the image. It is also be
utilized for the recognition of contaminated plant regions. An

IJISRT22SEP051 www.ijisrt.com 99
Volume 7, Issue 9, September – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
III. RELATED WORK The two kinds of ANN are the feed forward ANNs in
which the way of behaving of any layer won't affect that
A. CNN equivalent layer and the feedback ANNs in which signals
Convolution Neural Networks is a network architecture spread in the two directions by requiring network loops.
for deep learning that have the capacity of handling complex
data[2]. The function of CNN is to simplify an images into a As displayed in figure 2, Artificial neuron involves a
simpler to-process form, without compromising the features few data sources that can take any values in the range of 0 and
that are necessary for getting a good prediction. There are 1, however just a single output[15]. For each inputs
different accessible architectures for CNN, for example, information the neuron has weight what's more, an overall
AlexNet[11], GoogLeNet[11], VGGNet, LeNet, ResNet, bias.
ZFNet and so on[13]. Its development has created a lots of
interest among researcher in different fields of software
engineering. In agriculture, these architectures has been used
for the arrangement of plants diseases.

The CNN model involves an input layer, convolution

layer, pooling layer, the fully connected layer and a
result/output layer. In input, an images are provided so that
easy to classify the disease in plants . The convolution layer
is utilized for extricating the feature from the input images.
The pooling layer figures the component values from the Fig 2: Classification using ANN classifier.
extracted component. Depending upon complexity of
images, the convolution, pooling layer can be additionally  Kumari et al. [3] in 2019 has proposed a automated
expanded to get more details. Fully connected layer utilizes framework/system for recognition of four classes of
the result of previous layers and changes them into a solitary diseases in cotton and tomato plants. The sample of 20
vector that can be utilized as a input for next layer. The result plants images are taken from Plant Village dataset which
layer at long last systematize the plant diseases. is used in the paper. Images are segmented by using
Kmeans clustering and identify features are extracted
through GLCM technique, that are given as input to the
ANN.

C. SVM
SupportVectorMachine is the famous Supervised
Learning algorithm, which is utilized for the Classifications
and Regressions problems. The main objectives of SVM
Fig 1:Plant disease classification through CNN classifier.
algorithms is to makes the best line/decision boundary that
can sort out n-dimensional area into the classes, so we can
 Kamal et al. [2] in 2019 has proposed two models
undoubtedly put the new data points in the right order later on
specifically Changed MobileNet and Reduced
and the best decision boundary is supposed to be
MobileNet by uses the depthwise distinct convolution
hyperplane[4].SVM pick out the extreme point/vector that
architecture and their results were contrasted with
help in making the hyperplane. These outrageous cases are
MobileNet, VGG and AlexNet. There are some
supposed to be support vectors and thus calculation is named
optimizers like Adam, SGD and Nadam are used in it.
as support vector machine. Consider the given chart, there are
Nadam performed better and with a quicker convergence
2 unique categories that are characterized using a hyperplane
rate than the other two optimizers. In this work 82,161
or decision boundary:
images having 55 well defined classes of unhealthy and
healthy plants were used from openly accessible
PlantVillage dataset for the training and testing of the
model.

B. ANN
ANN is a non linear statistical data handling model
inspired by how data is handled by a biological system for
example the brain[13]. It comprises of processing elements
(PEs) or artifical neuron that are connect with coefficients,
which form the neural structure. They assemble the
information by identifying data patterns and connections and
they learn through experience and not by programming.
Artificial Neural Networks can be utilized for pattern
extraction due to their capacity of getting significance from
complex information. Fig 3: Classification using SVM classifier.

IJISRT22SEP051 www.ijisrt.com 100

Volume 7, Issue 9, September – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
 Bhimte et al.[4] in 2018 introduced a model which IV. COMPARISON
identified Bacterial scourge and Magnesium Deficiency
in the cotton plants. The dataset comprised of 130 Table 1: Summary of work done using CNN, ANN, SVM
captured images, taken by the camera. Quality of an and KNN technique’s.
images are upgrad by using preprocessing techniques and
afterward k-means clustering technique is used for
segmentation. The improved images are then classifyby
using the SVM classifier after the features have been
removed by using GrayLevelCo-occurrence Matrix.

D. KNN
K-Nearest Neighbor is the most straightforwards
Machine Learning Algorithms. K-NN can be used as
Classification, Regression problems, and it's a nonparametric
structural. It is additionally called a lazy learner algorithm
since it doesn't learn from the training set promptly rather it
stores the data and at the time of characterization, it perform
an action on the data. KNN at the training stage simply stores
the data and when it collects the new data, then it specify that
data into a classes that is much related to new data. Here, the
classification is done based on the calculated Euclidean
distance metric.

V. CONCLUSIONS

Serious diseases in plants leads the annual losses of the

agricultural yields and highly affect the economy of any
country. Consequently, detecting the plant diseases at a
beginning phase is very important for the prevention of such
extreme losses later on. In this paper, Image processing steps
Fig 4: Classification using KNN classifier. and Classification techniques that are generally broadly used
for the identification and detection the diseases on plants have
For K-NN classification three fundamental viewpoints been reviewed. The latest work has been anatomized and it is
as per the following: illustrate in the above tables. It is concluded that among the
> Simple resultant output interpretation. various techniques that already used in the current work done,
> Short computational time. The deep learning concepts, CNN approach gained the
> High prediction rate. highest accuracy as compared to others techniques.

This technique generally used in regions like pattern REFERENCES

recognition, text mining, and in agriculture/farming for
characterizing different diseases in plants. [1]. S. D. Khirade and A. B. Patil, "Plant Disease Detection
Using Image Processing," in International Conference
 Hossain et al. [5] in 2019 have examined plant diseases for on Computing Communication Control and
example, anthracnose, alternaria alternata, bacterial blight, Automation, Pune, 2015, pp. 768-771.
a leaf spot and used K-NN to characterize them. The [2]. Kamal, K.C., Zhendong Yin, Mingyang Wu, and Zhilu
dataset involves 237 leaf images obtained from the Wu, (2019), "Depthwise Separable Convolution
Arkansas plant diseases data base. The feature of plants Architectures for Plant Disease Classification"
have been removed by using the GLCM method. To Computers and electronics in agriculture.
exclude overfitting, the five fold cross approval was [3]. C. U. Kumari, S. Jeevan Prasad and G. Mounika, "Leaf
applied on the trained dataset. Disease Detection: Feature Extraction with K-means
clustering and Classification with ANN," 2019 3rd
International Conference on Computing Methodologies
and Communication (ICCMC), Erode, India, 2019, pp.
1095-1098.

IJISRT22SEP051 www.ijisrt.com 101

Volume 7, Issue 9, September – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[4]. N. R. Bhimte and V. R. Thool, "Diseases Detection of
Cotton Leaf Spot Using Image Processing and SVM
Classifier," 2018 Second International Conference on
Intelligent Computing and Control Systems (ICICCS),
Madurai, India, pp. 340-344.
[5]. E. Hossain, M. F. Hossain and M. A. Rahaman, "A
Color and Texture Based Approach for the Detection
and Classification of Plant Leaf Disease Using KNN
Classifier," 2019 International Conference on
Electrical, Computer and Communication Engineering
(ECCE),
[6]. Cox'sBazar, Bangladesh, 2019, pp. 1-6
[7]. Vijai Singh, A.K. Misra, “Detection of Plant Leaf
Diseases using Image Segmentation and Soft
Computing Techniques”, Information Processing in
Agriculture 4 (2017), pp. 41–49
[8]. S, Santhosh Kumar; Raghavendra, B. K. (2019). [IEEE
2019 5th International Conference on Advanced
Computing & Communication Systems (ICACCS) -
Coimbatore, India (2019.3.15-2019.3.16)] 2019 5th
International Conference on Advanced Computing &
Communication Systems (ICACCS) - Diseases
Detection of Various Plant Leaf Using Image
Processing Techniques: A Review.
[9]. Monica Jhuria, Ashwani Kumar, and Rushikesh Borse,
“Image Processing For Smart Farming: Detection Of
Disease And Fruit Grading”, Proceedings of the 2013
IEEE Second International Conference on Image
Information Processing (ICIIP-2013
[10]. P. Revathi, M. Hemalatha, Classification of Cotton
Leaf Spot Diseases Using Image Processing Edge
Detection Techniques, IEEE International Conference
on Emerging Trends in Science, Engineering and
Technology, pp-169-173, Tiruchirappalli, Tamilnadu,
India, 2012.
[11]. Dr. Sridhathan C"Plant Infection Detection Using
Image Processing.“International Journal Of Modern
Engineering Research (IJMER), vol. 08, no. 07, 2018,
pp.13-16
[12]. Mohanty, S. P., Hughes, D. P., & Salathé, M. (2016).
Using deep learning for imagebased plant disease
detection. Frontiers in Plant Science, 7(September),
[1419]
[13]. “Indian agriculture economy.”. Available: http://
statistics times.com/economy/sectorwise-gdp-
Contribution-ofindia. Php
[14]. Saleem MH, Potgieter J, Arif KM. Plant Disease
Classification: A Comparative Evaluation of
Convolutional Neural Networks and Deep Learning
Optimizers. Plants. 2020; 9(10):1319.
[15]. Gavhale, Ms & Gawande, Ujwalla. (2014). An
Overview of the Research on Plant Leaves Disease
detection using Image Processing Techniques. IOSR
Journal of Computer Engineering.
[16]. Sladojevic, Srdjan & Arsenovic, Marko & Anderla,
Andras & Stefanović, Darko. (2016). Deep Neural
Networks Based Recognition of Plant Diseases by Leaf
Image Classification. Computational Intelligence and
Neuroscience.

IJISRT22SEP051 www.ijisrt.com 102