See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/349172952

Plant Disease Detection Techniques: A Review

Research · April 2019

DOI: 10.13140/RG.2.2.10724.30081

CITATIONS READS

7 2,977

2 authors:

Ufaq Khan Ashish Oberoi

Mohamed bin Zayed University of Artificial Intelligence Regional Institute of Management and Technology
9 PUBLICATIONS 46 CITATIONS 37 PUBLICATIONS 70 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Ufaq Khan on 10 February 2021.

The user has requested enhancement of the downloaded file.

 Ufaq Khan et al, International Journal of Computer Science and Mobile Computing, Vol.8 Issue.4, April- 2019, pg. 59-68

Available Online at www.ijcsmc.com

International Journal of Computer Science and Mobile Computing

A Monthly Journal of Computer Science and Information Technology

ISSN 2320–088X
IMPACT FACTOR: 6.199

IJCSMC, Vol. 8, Issue. 4, April 2019, pg.59 – 68

Plant Disease Detection

Techniques: A Review
*Ufaq Khan; **Dr. Ashish Oberoi
* [email protected], ** [email protected]
M.Tech (Student), Department of Computer Science & Engineering, RIMT University
Professor, Department of Computer Science & Engineering, RIMT University

ABSTRACT: The Plant disease detection is the technique which is applied to detect diseases from the
infected plants. The plant disease detection techniques consist of two phases, in the first phase segmentation
of the input image done which detect diseased portion from the input image. The technique of features
extraction is applied which will extract the features of the image and also classify extracted features using
various classifiers. In this paper, various techniques of segmentation, feature extraction and classification is
reviewed and discussed in terms of various parameters.

INTRODUCTION

1.1 Image Processing

Image consists of finite number of elements each of which has a particular location and value. These elements
are called picture element and pixels. Digital image processing refers to processing of digital images by means
of digital computers. It is a process to convert an image into digital form and perform some operations to get an
enhanced image and extract useful information from it. There are different types of image processing fields like
nuclear medicine, astronomical observation, signature recognition, number plate detection, agriculture etc.
Digital image processing is widely used in agricultural field. It is used in, harvest control, fruit grading, seeding
and food picking, plant disease detection etc.

1.2 Plant Disease Detection

Plant diseases cause several devastating economical, social and ecological losses. It is necessary to detect plant
disease accurately on its minor stage to protect the quality and quantity of crops. There are several ways to
detect plant pathologies. Some diseases do not have any visible symptoms associated and farmers can‟t
recognize it easily. In those cases, normally some kind of sophisticated analysis, usually by means of powerful
microscopes is necessary. In other cases, the signs can only be detected by imaging machine which cover almost

© 2019, IJCSMC All Rights Reserved 59

 Ufaq Khan et al, International Journal of Computer Science and Mobile Computing, Vol.8 Issue.4, April- 2019, pg. 59-68

the entire electromagnetic spectrum ranging from gamma to radio waves that are not visible to humans. Farmers
use naked eye observation method to detect the diseases with their experience but it is very hard to detect
disease at very early stage by this method. To identify different plant disease an automated system has been
implemented using different classifiers such as back propagation, PCA and SVD techniques of neural network
[1].

Training Set

Input
Feature
certain set Database
Extraction
of images

Test Set

Input
Feature
Single Classification
Extraction
Image

Segmentation

Fig: 1: General Process of Plant Disease Detection in Image Processing

1.3 Segmentation Techniques

In Image segmentation, an image is divided into small regions so that useful information can be extracted from
the data. There are two ways by which image segmentation can be done. First method is based on discontinuities
e.g.-edge detection and Second method is based on similarities e.g. thresholding done using Otsu‟s method.

1.3.1 Region based segmentation

1. Region Growing
2. Region Splitting and Merging

1. Region Growing

It is procedure that groups pixels or sub regions into larger regions by examining the neighboring pixels. It is
repeated for each boundary pixels. The approach start with a “seed “point and then grow regions by appending
to each seed those neighboring pixels that have properties similar to the seed .

2. Region Splitting and Merging

In this technique, a region R is subdivided into uniform regions e.g.-smaller and smaller regions into quadrant
and sub-quadrants. In this procedure, initial assumption is made that the entire image is a single region, then
computes the homogeneity criterion to check some properties, If FALSE, then the region is split into the further

© 2019, IJCSMC All Rights Reserved 60

 Ufaq Khan et al, International Journal of Computer Science and Mobile Computing, Vol.8 Issue.4, April- 2019, pg. 59-68

smaller regions. This process continues until further splitting is necessary .These small square regions are
then merged if they are similar irregular regions.

1.3.2 Partition Clustering

The segmentation based technique is a partition clustering technique which is used to partition n number of
observations into k cluster each object belongs to particular one cluster. Here k is the number of clusters in the
segmented image. The main advantage of segmentation based K-means clustering technique is that it works on
both local and global information of image. This algorithm is easy to implement, fast, robust and flexible [2].

1.3.3 Edge Detection

Edge Detection is a the approach used more frequently for segmenting image based on sudden change in
intensity values and results in a line drawing of the image. The lines represent changes in values such as
intersections and cross sections of planes, textures, colors, as well as differences in shading The main purpose
of this technique is to identify areas of an image where a large change in intensity values occurs e.g. canny,
sobel, prewitt edge detectors most widely used edge detection, canny edge detector to detect wide range of
edges of image.

1.4 Artificial Neural Networks

Artificial Neural Networks are the models which are inspired by the biological neural network. Artificial neural
networks process the information like the human brain but they are slow in processing but they achieve
performance near to biological neurons So, ANN‟s pattern are widely used in the automatic identification and
classification of plant diseases. Classification processes the data in groups or classes. Classification is usually
called as supervised learning technique [3].

Fig. 2: Artificial Neural Network

Artificial neural networks consist of processing units which are typically organized in the form of layers. It has
input layer from which signal is propagated to output layer through some hidden layers by which weights can be
updated. The data flows from an input layer to the output layer in feed-forward network. Some of the classifiers
used are.

1.4.1 BPA (Back Propagation Analysis)

BPA is algorithm which is used to train artificial neural network. It is mathematical algorithm in which signal
from outer layer propagate to the input layer. It is used to train the hidden layers. In this error is calculated and
propagated back to hidden layer by which we can train our network. In this any number of hidden layers can be
used but less than or equal to input layer. It provides supervised learning to neural networks.

© 2019, IJCSMC All Rights Reserved 61

 Ufaq Khan et al, International Journal of Computer Science and Mobile Computing, Vol.8 Issue.4, April- 2019, pg. 59-68

1.4.2 PCA (Principal component analysis)

Principal component analysis is a dimensionality reduction technique in data analysis [4]. First important
property comes from scheme of compressing a set of high dimensional vectors into a set of low dimensional
vectors and then reconstructing. Second parameters can be directly calculated from data. Third, compression and
decompression can easily performed on model parameters as they require only matrix multiplications.PCA is
widely used method for plant disease detection. But it is less accurate than SVM.

1.4.3 Singular Value Decomposition (SVD)

It is factorization of real and complex matrix. It is used in many applications like signal processing.
Special feature is that it can be performed on any real matrix [5]. It is effective method to split the components
which are linearly independent and has its own energy contribution. Several SVD properties are very
advantageous for images such as maximum energy packing, solving of least mean squares problem, calculating
pseudo- inverse of a matrix and multivariate analysis. . Let‟s say, we have a matrix A with m rows and n
columns with rank r ≤ n < m .then the a can be factorized into three matrices.
A = USVT

LITERATURE REVIEW
This section describes works done by researchers in different domains such as fruit grading system, weed
detection, classification of plants etc.

G. Saradhambal, et.al proposed an approach to produce a system for automatic plant disease detection. Research
was carried out to predict the infected area of the leaves by applying k-means clustering algorithm and the
Otsu‟s classifier. Both the shape and texture features were extracted in the proposed work. The shape oriented
features that were extracted in this work included area, color axis length, eccentricity, solidity and perimeter,
whereas the texture oriented features were contrast, correlation, energy, homogeneity and mean [6]. And lastly,
classification in this research was done using a neural network based classifier.

K. R. Aravind, et.al studied maize crop diseases for automatizing the plant disease detection system. Features
known as Speeded Up Robust Features (SURF) were extracted from each image. The features were clustered
using k-means algorithm. Two methods were used for feature extraction namely, histogram and GLCM. These
two methods were used for studying various textural features [7]. For classification, multi-class SVM was
applied based on various kernel functions like linear, polynomial and Radial Basis Function, etc. A best average
accuracy of 83.7% was achieved using the SURF features in the proposed work.

Santanu Phadikar and Jaya Sil describes a Software system for disease detection based on the infected images of
various rice plants. Images of the infected rice plants using digital camera are captured and then processed using
techniques like image growing, image segmentation and zooming to detect infected parts of the plants. Then
infected part of the leaf has been used for the classification using neural network. The methods employed in this
system are both image processing and soft computing technique [8].

H. Al-Hiary, S. Bani-Ahmad et.al. Proposed method to Accurate Detection and Classification of Plant Diseases
In the first step it identify green Colours pixels. Then pixels are masked based on particular threshold values
that are obtained using Otsu's method, and then mostly green pixels are masked. The other additional step is that
the pixels with zeros red, green and blue values and the pixels of infected clusters from the boundaries were
completely removed. SGDM matrix generated for H and S, and then GLCM Function is called to calculate the
features. The experimental results demonstrate that this technique is a powerful technique for the detection of
plant leaves diseases [9].

© 2019, IJCSMC All Rights Reserved 62

 Ufaq Khan et al, International Journal of Computer Science and Mobile Computing, Vol.8 Issue.4, April- 2019, pg. 59-68

Sabah Bashir, Navdeep Sharma surveyed on Remote Area Plant Disease Detection Using Image Processing. In
this paper, a method is proposed for detection of disease in malus domestica using methods like k-mean
clustering, texture and color analysis. Algorithm used for texture segmentation is CCM method. As RGB images
of leaves are converted into HSI color space representation .Then this is used to generate co-occurrence matrix
(CCM). By comparing texture and color images plant diseases can be detected [10].

Dr. A. N. Cheeran, Piyush Chaudhary proposed, an algorithm which uses image processing for disease spot
segmentation in plant leaf .in the first phase automatic detection and classification of plant diseases is done.
Colour transform of RGB image is performed for better segmentation of disease spots. For image smoothing
median filter is used. Otsu method is used to calculate the threshold. An algorithm which is independent of
background noise, plant type and disease spot colour was developed and experiments were carried out on
different “Monocot” and “Dicot” family plant leaves with both, noise free (white) and noisy background. In this
paper a comparison of the effect of CIELAB, HSI and YCbCr colour space in the process of disease spot
detection is done [11].

Haiguang Wang, Guanlin Li, in this study, the image recognition of two kinds of grape diseases (grape downy
mildew and grape powdery mildew) and two kinds of wheat diseases (wheat stripe rust and wheat leaf rust) was
conducted by using image processing technologies and BP networks. Based on the data of the extracted colour
features, shape features and texture features from disease images and their combined features, BP networks
constructed by using different function combinations were used as the classifiers to identify grape diseases and
wheat diseases, feature combinations respectively. The image recognition of plant diseases using BP networks
was also conducted based on the dimension-reduced data that were obtained by using PCA to process the data
[12].

Sagar Patil, Anjali Chandavale in these paper different plant disease detection techniques is discussed. Many
segmentation techniques and classifiers are given. From the schemes discussed, it is concluded that K-means
clustering method for segmentation is widely used by most of the researchers. For classification and feature
extraction, GLCM along with SVM classifier were found to be better in performance in comparison to others
[13].

Prof. Sunil Deokule, Shital Banker used edge detection and colour matching histogram techniques to detect
diseases of plants. The system includes two phases, in first phase; all the disease and healthy leaves are given as
input to the system.
In the Second phase, testing samples are given as input to the system. In both the phases, first the RGB
components are converted into three layers red, green and blue; greyscale image. For edge detection, Canny
Edge Detector is used. Then, histogram is plotted for each component of healthy and disease leaf image and
stored in the systems. A comparison is performed with stored results and disease infected or not infected plants
are identified [14].

Savita N. Ghaiwat, Parul Arora study different classifiers. To predict the class of a test example k-nearest-
neighbour is simplest algorithm but one of the disadvantage of the k-NN method is the time complexity of
making predictions. Additionally, neural networks are tolerant to noisy inputs, but it‟s difficult to understand
structure of algorithm. SVM was found competitive with the best available machine learning algorithms in
classifying high-dimensional data sets. In SVM computational complexity is reduced to quadratic optimization
problem and it‟s easy to control complexity of decision rule and frequency of error [15].

Hiteshwari Sabrol,satish kumar The algorithm begins with digital image acquisition of infected and non-infected
plants; perform image pre-processing, differentiate disease infected region from a non-infected region using
colour space conversion ,segmentation, extract features from segmented images for recognition and
classification based on Feature Analysis, Neural Network, Support Vector Machine and Fuzzy and Rule-Based
Classification. This survey expected to useful for researchers from plant pathology and pattern recognition field
[16].

© 2019, IJCSMC All Rights Reserved 63

 Ufaq Khan et al, International Journal of Computer Science and Mobile Computing, Vol.8 Issue.4, April- 2019, pg. 59-68

Prajakta Mitkal, Priyanka Pawar, in this paper concept of disease detection of sugarcane leaf based on feature
extraction is proposed. After doing review on various techniques and algorithms conclusion given is that
compared to other algorithms, SVM algorithm gives the better result. This approach can also be developed using
normal techniques like JAVA, but using Matlab gives the efficient and effective result. This application is
helpful for farmer and laboratory where they can easily protect their crops from disease in early stages and can
increase in growth of production [17].

Author Year Description Technique Parameters Outcome

2018 An approach was Research was Area, color Improvement in

proposed to carried out to axis length, accuracy level for
G. Saradhambal, produce a system predict the eccentricity, detecting plant
Dhivya, R., for automatic plant infected area of solidity, diseases from
Latha, S., disease detection. the leaves by perimeter, images was
Rajesh, R. applying k-means contrast, achieved by
clustering correlation, applying this
algorithm and the energy, proposed technique.
Otsu‟s classifier. homogeneity
and mean.

2018 Maize crop Two methods Features The best average

diseases were were used for known as accuracy of 83.7%
K. R. Aravind, studied for feature extraction Speeded Up was achieved using
Raja, P., automating the namely, histogram Robust the SURF features
Mukesh, K.V., plant disease and GLCM. These Features in the proposed
Aniirudh, R., detection system. two methods were (SURF) were work.
Ashiwin, R. used for studying extracted from
various textural each image.
features.

2008 Work is done on Images of the Feature Successful

Diseased rice infected rice extraction classification is
Santanu images which are plants using achieved for
Phadikar and classified using digital camera are different cases.
Jaya Sil SOM Neural captured and then
network .Zooming processed using
algorithm is used techniques like
for classification of image growing,
test images. image
segmentation and
zooming to detect
infected parts of
the plants.

© 2019, IJCSMC All Rights Reserved 64

 Ufaq Khan et al, International Journal of Computer Science and Mobile Computing, Vol.8 Issue.4, April- 2019, pg. 59-68

H. Al-Hiary, S. 2011 Green pixels are Then pixels are SGDM matrix Algorithm was
Bani-Ahmad masked using masked based on generated for tested on five
Otsu‟s method‟s particular H and S, and diseases on the
threshold values then GLCM plants- Early scorch,
.K-mean clustering
that are obtained Function is Cottony mold,
and neural network using Otsu's called to ashen mold, late
are formulated for method, and then calculate the scorch, tiny
clustering and mostly green features. whiteness. Precision
classification. pixels are masked. between 83% and
94%,is achieved.
Sabah 2012 This paper presents Algorithm used As RGB The experimental
bashir,navdeep an effective method for texture images of results indicate that
sharma for detection of segmentation is leaves are proposed
diseases in Malus CCM method. converted into Approach
Domestica using HSI colour significantly
methods like K- space enhances accuracy
mean clustering, representation.
in automatic
CCM, Then this is
color and texture used to detection of normal
analysis. generate co- and affected
occurrence produce.
matrix (CCM).
Piyush 2012 In this paper a Process of disease Detection of In this method
Chaudhary, comparison spot detection is disease spots, different disease
Anand K. of the effect of done. Median and affects spots are detected
CIELAB, HSI and filter is used for caused on accurately and
Chaudhari, Dr.
YCbCr colour image smoothing. background, results are not
A. N. Cheeran et space in the type of leaf, affected by
al. type of disease background, type of
spot and leaf, type of disease
camera. spot and camera.
Haiguang Wang, 2012 Two grape disease Image processing Fitting Grape diseases was
Guanlin Li, (grape downy technologies and accuracy and obtained as the
Zhanhong Ma, mildew and grape BP networks were prediction fitting accuracy was
powdery mildew) used in this accuracy. 100% and the
Xiaolong Li
and two kinds of mechanism. PCA prediction accuracy
wheat diseases was used to was 97.14%, and
(wheat stripe rust feature extracted that for wheat
and wheat of diseased diseases was
leaf rust) was images. obtained as the
conducted. fitting accuracy and
the prediction
accuracy were both
100%.
2013 Many classification For classification
Sagar Patil1, In this paper segmentation and feature and feature
Anjali discussed existing techniques and extraction extraction, GLCM
Chandavale segmentation classifiers were along with SVM
method along with compared in terms classifier were
classifiers for of different found to be better in
detection of performance performance in
diseases in factors to comparison to
Monocot and Dicot determine their others
family plant. outputs.

© 2019, IJCSMC All Rights Reserved 65

 Ufaq Khan et al, International Journal of Computer Science and Mobile Computing, Vol.8 Issue.4, April- 2019, pg. 59-68

Shital Bankar, 2014 The colour features The system

Ajita extractions are includes two
Dube,Pranali applied on samples phases, in first Accuracy and Canny edge
Kadam, Prof. phase; all the efficiency detection and
that are healthy leaf
Sunil Deokule disease and
of plant and the histogram matching
healthy leaves are
diseased leaf of the given as input to gives accurate
plant. Plant the system. results.
diseased is detected In the Second
by using histogram phase, testing
matching. samples are given
as input to the
system.

2014 This paper present In SVM Tolerance to SVM was found

Savita N. survey on different computational noisy inputs, competitive with the
Ghaiwat, Parul classification complexity was classification best available
reduced to of high
Arora techniques that can machine learning
quadratic dimensional
be used for plant optimization datasets. algorithms in
leaf disease problem and it classifying high-
classification, such was easy to dimensional data
as k-Nearest control sets.
Neighbour complexity of
Classifier, decision rule and
frequency of
Probabilistic
error.
Neural Network,
Genetic Algorithm,
Support Vector
Machine, and
Principal
Component
Analysis, Artificial
neural network,
Fuzzy logic.
Hiteshwari 2015 This paper It performs image Accuracy of Results shows that
sabrol,Satish proposed algorithm pre-processing, disease highest accuracy for
kumar for digital image differentiate detection plant disease
disease infected
acquisition of recognition and
region from a
infected and non- non-infected classification done
infected plants; region using under three
segmentation, categories i.e.
extract features Feature Analysis
from segmented using colour-based
images for classification,
recognition and
Neural Networks
classification.
using PCA and
SVM with 100%
accuracy

© 2019, IJCSMC All Rights Reserved 66

 Ufaq Khan et al, International Journal of Computer Science and Mobile Computing, Vol.8 Issue.4, April- 2019, pg. 59-68

Prajakta 2016 In this paperwork is Here, the image is Crop SVM algorithm
Mitkal,Priyanka done for sugarcane segmented useful protection and gives the better
Pawar, Mira leaf disease segment used for amount of result as compare to
Nagane et al. extraction finally production.
detection in which other algorithms.
texture statistics is
green pixels are completed and This approach can
removed according to also
analysis disease be developed using
prevention is normal techniques
provided. like JAVA, but
using Matlab gives
the efficient and
effective
result
Table 1: Table of Comparison

CONCLUSION
In this work, it is been concluded that plant disease detection is the technique to detect infected portion from the
leaf. The plant disease detection consist of two steps, in the first step the image segmentation is done and in the
second step technique of feature extraction and classification is applied which will classify diseases and normal
portion in the image. In this paper, various techniques of plant disease detection is reviewed and discussed in
terms of various parameters.

REFERENCES
1. Tejal Deshpande, Sharmila Sengupta, and K.S.Raghuvanshi, “Grading & Identification of Disease in
Pomegranate Leaf and Fruit,” IJCSIT, vol. 5 (3), pp 4638-4645, 2014.
2. Jitesh p.shah,Harshadkumar B.Prajapati.Vipul.K.Dhabhi” A Survey on Detection and Classification of Rice
Plant Diseases” IEEE,2016
3. P.Revathi and 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 (INCOSET), Tiruchirappalli, pp 169-173, 2012.
4. Vaibhavi S. Bharwad, Kruti J. Dangarwala “Recent Research Trends of Plant Disease Detection”
International Journal of Science and Research (IJSR) ISSN (Online),ImpactFactor (2014)
5. Amanpreet kaur “Comparative Analysis of PCA+BPA with Novel PCA+BPA+SVD Technique in Neural
Network to Improve Plant Disease Detection Rate” International Journal of Advance Research in Computer
Science and Management Studies ,Volume 4, Issue 7, July 2016
6. Saradhambal, G., Dhivya, R., Latha, S., Rajesh, R., „Plant Disease Detection And Its Solution Using Image
Classification‟, International Journal of Pure and Applied Mathematics, 2018,Volume 119, No. 14, pp. 879-
884
7. Aravind, K.R., Raja, P., Mukesh, K.V., Aniirudh, R., Ashiwin, R., „ Disease Classification in Maize Crop
using Bag of Features and Multiclass Support Vector Machine‟, Proceedings of the Second International
Conference on Inventive Systems and Control (ICISC), 2018, pp. 1191-1196
8. Santanu Phadikar and Jaya Sil, “Rice Disease Identification using Pattern Recognition Technique”,
Proceedings of 11th International Conference on Computer and Information Technology (ICCIT 2008) , 25-
27 December, 2008
9. H. Al-Hiary, S.Bani-Ahmad, M.Reyalat et.al, “Fast and Accurate Detection and Classification of Plant
Diseases”, International Journal of Computer Applications (0975-8887) Volume 17 – No. 1, March 2011

© 2019, IJCSMC All Rights Reserved 67

 Ufaq Khan et al, International Journal of Computer Science and Mobile Computing, Vol.8 Issue.4, April- 2019, pg. 59-68

10. Sabah Bashir, Navdeep Sharma, “Remote Area Plant Disease Detection Using Image Processing”, IOSR
Journal of Electronics and Communication Engineering (IOSRJECE), ISSN: 2278-2834 Volume 2, Issue 6
(Sep-Oct 2012), PP 31-34
11. Piyush Chaudhary, Anand K. Chaudhari et.al, “Color Transform Based Approach for Diseases Spot
Detection on Plant Leaf”, International Journal of Computer Science and Telecommunications, Vol 3, Issue
6, June 2012
12. Haiguang Wang, Guanlin Li et.al, “Image Recognition of Plant Diseases Based on Backpropagation
Networks‟, 5th International Congress on Image and Signal Processing (CISP 2012), IEEE, 2012
13. Sagar Patil, Anjali Chandavale, “A Survey on Methods of Plant Disease Detection”, International Journal of
Science and Research (IJSR), ISSN: 2319-7064, 2013
14. Shital Bankar, Ajita Dube et.al, “Plant Disease Detection Technique using Canny Edge Detection & Color
Histogram in Image Processing”, International Journal of Computer Science and Information
Technologies, Vol. 5(2), 2014, 1165-1168
15. Savita N. Ghaiwat, Parul Arora, “Detection and Classification of Plant Leaf Diseases Using Image
Processing Techniques: A Review”, International Journal of Recent Advances in Engineering and
Technology (IJRAET), ISSN: 2347-2812, Volume-2, Issue – 3, 2014
16. Hiteshwari Sabrol, Satish Kumar, “Recent Studies of Image and Soft Computing Techniques for Plant
Disease Recognition and Classification”, International Journal of Computer Applications (0975-8887),
Volume 126 – No.1, September 2015
17. Prajakta Mitkal, Priyanka Pawar et.al, “Leaf Disease Detection and Prevention Using Image Processing
using MATLAB”, International Journal of Recent Trends in Engineering and Research (IJRTER), Volume
02, Issue 02, February 2016, ISSN: 2455-1457

© 2019, IJCSMC All Rights Reserved 68

View publication stats