Materials Today: Proceedings xxx (xxxx) xxx

Contents lists available at ScienceDirect

Materials Today: Proceedings

journal homepage: www.elsevier.com/locate/matpr

Detection of plant leaf disease using digital image processing

Ramesh Kumar Mojjada ⇑, K. Kiran Kumar, Arvind Yadav, B.V.V. Satya Vara Prasad
Department of Computer Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, Guntur, India

a r t i c l e i n f o a b s t r a c t

Article history: Productivity in agriculture is highly economic dependent. This is one of the reasons why the diagnosis in
Received 5 October 2020 agriculture of plant diseases is important because plant disease is very natural. If proper care is not taken
Received in revised form 3 November 2020 in this area, the impact on plants and the quality, quantity or productivity of their products will be seri-
Accepted 6 November 2020
ous. For instance, in pine trees in the United States, a small leaf disease is a dangerous disease. It is helpful
Available online xxxx
to diagnose plant disease with any automated procedure, as it decreases the widespread surveillance of
farm sites, and detects the signs of diseases very early, that is when they arise on leaves of plant. A seg-
Keywords:
mentation algorithm used to automatically detect or classify plant leaf diseases is presented in this arti-
SVM
Image processing
cle. This article. It also includes surveys of the various methods used to classify diseases to detect plant
Disease detection leaf diseases. A genetic algorithm is used to segment the picture which is important for the identification
DIP of disease in leaf disease.
Support vector machine Ó 2021 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the Emerging Trends in
Materials Science, Technology and Engineering.

1. Introduction tion of plant disease by visual means is both difficult and difficult.
Less accurate than that. Whereas when Automatic disease detec-
The Indian economy is heavily dependent on productivity Agri- tion is used. It will then give more accurate results, in less time,
culture. The detection of plant disease therefore plays a role in and less effort. Segmentation of the image can be done in different
major role in the field of agriculture [1,2]. If adequate plant care ways ranging from simple threshold method to advanced Method
is required. It does not take, it creates serious plant impacts and of segmentation of the color image. This is equivalent to something
it affects the plant. Quality, quantity or productivity of the corre- that can easily be separated from the human eye and seen as this is
sponding item unhealthy region of the leaves of the plant is the an individual object. Computers are unable to recognize the
area of the leaf. Affected by disease, which reduces plant quality. objects, several techniques are being developed for the image
Automatic disease detection technique is beneficial at an initial Segmentation.
stage. Stage of disease detection. Current method of detection of
Plant disease is simply expert naked eye observation. This requires
a huge team of specialists and a continuous team of experts. Mon-
2. Literature survey
itoring of the plant, which is very expensive for large farms. Farm-
ers [3] in some countries do not have adequate facilities or the
A survey on the different classification techniques used to clas-
concept of contacting professionals.
sify plant leaf diseases is presented in Ghaiwat et al. The k-nearest-
Due to which consultants even have a high cost and it is time.
newer method seems to be the most appropriate and simple pre-
To consume, too. In such circumstances, the method suggested is
diction algorithms of all classes [4] to the test example given.
beneficial to the monitoring of large crop fields. Detection diseases
Although training data can not be linearly segregated, SVM. Intel-
by looking at the automatic way Symptoms on the leaves make it
ligent Wheat Diseases Android Phone by Y is difficult to identify
easier and more cost-effective this one.
optimal parameters. Q. Xia, Y. Li, C.
Provides support for machine vision to provide image based
Xia and Li suggested in 2015 that the device be designed to
automated process control, inspection and robot guidance. Detec-
diagnose android intelligent wheat disease. In this scenario, users
gather photographs of wheat diseases with Android phones and
⇑ Corresponding author. send images via the network. The disease diagnosis server. After
E-mail address: rameshkumar.mojjada@kluniversity.in (R.K. Mojjada). it is obtained in illness pictures [5,6], the server separates the

https://doi.org/10.1016/j.matpr.2020.11.115
2214-7853/Ó 2021 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the Emerging Trends in Materials Science, Technology and Engineering.

Please cite this article as: Ramesh Kumar Mojjada, K. Kiran Kumar, A. Yadav et al., Detection of plant leaf disease using digital image processing, Materials
Today: Proceedings, https://doi.org/10.1016/j.matpr.2020.11.115
Ramesh Kumar Mojjada, K. Kiran Kumar, A. Yadav et al. Materials Today: Proceedings xxx (xxxx) xxx

image from RGB colour to HIS space colour. Color and texture 3.3. Image segmentation
capabilities
The parameters shall be calculated by the use of a colour It the process in which dividing the input image into sub parts
moment and co-occurrence grey level matrix, the preferred charac- based on the threshold value in order to separate the background
teristics shall be the feedback from the carrier vector Recognition from the original image. So the processing will be faster for the
machine and the detection results shall be returned to the detection of the lesser region of the plant images [9,10].We can
consumer. segment the segment into more than two regions based on more
than one threshold value. The K-means algorithm is in this paper
used for segmentation of the image. It’s a genetic algorithm Opti-
3. Methodology
mization algorithm used after the k-means Segmentation to obtain
optimized results.
Fig. 1.

3.4. Feature Extraction

3.1. Image acquisition
The elimination of traits is an important part of the generous
Image acquisition means the acquisition of an image by Camera prediction of the infected area. Extraction of the feature requires
means from any real life scene. In the world of today, The com- a decrease in the amount of resources used to explain the large
monly used method is to capture photos using digital camera. dataset. Feature extraction methods are useful in various image
However, other methods can also be used. In the project, Images processing applications, e.g. character recognition. As [10] charac-
are taken from the data [7] set of the plant village through which teristics characterise an image’s behaviour, they are also important
the images will be taken and the algorithm will be trained and it in terms of storage, classification efficiency and, of course, time
was tested Process of acquiring the image using different devices. consumption.

3.2. Image preprocessing 3.5. Disease classification

Pre-processing of the image is used to enhance the image qual- Extraction and contrast of co-occurrence characteristics of the
ity required for further analysis and processing. It involves colour leaves with characteristics values is stored in the dataset function
conversion, flushing of the image and enhancing the frame. The during the classification process. The Support Vector Machine Sup-
image input quality is omitted from the image image enhancement port classifies the image. The supporting vector machinery (SVM)
to maximise the image image contrast [8]. Image cutouts are per- is a collection of linked supervised learning methods. The trains
formed in order to reach a area of interest. and test portions are broken down. 80% of images are trained.

Fig. 1. Block diagram.

2
Ramesh Kumar Mojjada, K. Kiran Kumar, A. Yadav et al. Materials Today: Proceedings xxx (xxxx) xxx

Fig. 2. Result.

The SVM testing is taken and 20% of images are taken. The aim 3. Semi Supervised Learning
[11,12] is unknown to SVM.SVM compares the features of SVM. 4. Reinforcement Learning
Enter and execute the classification based on the trained images.
The results are separated into the images. The SVM production is 3.6.1. Supervised learning
the name of the illness and the solution to it. Supervised machine learning algorithms may use labelled
examples to predict future events to apply what was already
3.6. Machine learning learned in the past to new data. The learning algorithm produces
an underlying function to predict performance values from an
Machine learning is an artificial intelligence (AI) system that interpretation of a known training dataset. The system can set tar-
lets programmes learn and improve their interactions automati- gets for any new input after sufficient training. The learning algo-
cally without being programmed directly. The goal of machine rithm can also compare its output with the right output and detect
learning is to construct computer programmes for the individual errors, so that the model can be changed if appropriate.
access and learning of data.
Machine learning is categorised into four forms 3.6.2. Unsupervised learning
The unsupervised machine learning algorithms are used where
1. Supervised Learning no data is classified or tagged. Unattended study of how systems
2. Unsupervised Learning are able to determine how the hidden structure can be described
3
Ramesh Kumar Mojjada, K. Kiran Kumar, A. Yadav et al. Materials Today: Proceedings xxx (xxxx) xxx

from unmarked data. The system does not find the right output, grapes, peach, bell pepper. The proposed algorithm is tested in
but explores the data and can derive data from datasets to describe these five classes above of the images of the plant leaf.
unmarked structures.
CRediT authorship contribution statement
3.6.3. Semi supervised learning
Somewhere in supervised and unattended learning is a semi- Ramesh Kumar Mojjada: Conceptualization, Methodology,
supervised machine learning algorithm, as both identified and Software, Data curation. K. Kiran Kumar: Writing - original draft,
unmarked training data – a limited amount usually of marked data Validation, Visualization. Arvind Yadav: Writing - original draft,
and a large amount of unmarked data. The programmes using this Validation, Visualization. B.V.V. Satya Vara Prasad: Investigation,
approach will increase learning accuracy dramatically. Semi- Supervision, Software, Writing - review & editing.
controlled learning is typically selected where acquired marked
data requires professional and relevant teaching tools. Otherwise, Declaration of Competing Interest
it usually takes no further resources to collect unmarked details.
The authors declare that they have no known competing finan-
3.6.4. Reinforcement learning cial interests or personal relationships that could have appeared
It is a learning approach that communicates with the surround- to influence the work reported in this paper.
ings by making behaviours and identifying faults or incentives. The
key features of enhancement learning are the analysis, error check References
and delayed incentives. This approach helps machines and soft-
ware agents to evaluate optimal actions automatically in a partic- [1] D. Radhika, K.D. Aruna, The smart triad: Big data analytics, cloud computing
and internet of things to shape the smart home, smart city, smart business &
ular situation to improve their efficiency [13]. A quick feedback is
smart country, Int. J. Recent Technol. Eng. 8 (2 Special Issue 11) (2019) 3594–
expected in order to let the officer know if the action is best. 3600.
[2] S.S. Kalavala, S. Sakhamuri, B.B.V.S.V. Prasad, An efficient classification model
3.6.5. Support vector machine for plant disease detection, Int. J. Innov. Technol. Explor. Eng. 8 (7) (2019) 126–
129.
They will work in linear as well as non-linear situations, dra- [3] G.V. Sai Prasanna, K.G. Vijay, Controlling and monitoring the plant growth
matically enhancing efficiency in many contexts. SVMs are proba- conditions using embedded systems, Int. J. Innov. Technol. Explor. Eng. 8 (6)
bly the best choice in combination with neural networks for many (2019) 1552–1555.
[4] J. Keerthi, S. Maloji, P.G. Krishna, An approach of tomato leaf disease detection
tasks when a good separating hyperplane is not easy to find. For based on SVM classifier, Int. J. Recent Technol. Eng. 7 (6) (2019) 697–704.
example, for a long time SVMs were the best option for classifying [5] G. Balram, K.K. Kiran, Smart farming: Disease detection in crops, Int. J. Eng.
MNIST datasets. Thanks to their absence from the use of a mathe- Technol. (UAE) 7 (2) (2018) 33–36.
[6] B.V.V.S. Narayana, K.S. Ravi, N.V.K. Ramesh, A review on advanced crop field
matical trick, they can capture very strong nonlinear dynamics monitoring system in agriculture field through top notch sensors, J. Adv. Res.
without having a mathematical truck. Help Vector Machine Dyn. Control Syst. 10 (6 Special Issue) (2018) 1572–1578.
(SVM) is a monitored learning machine algorithm for classification [7] S. Kiran, S.B. Sriramoju, A study on the applications of IOT, Indian J. Public
Health Res. Dev. 9 (11) (2018) 1173–1175.
and regression problems. But it is primarily used in issues of [8] S. Sridevi, P.M. Bindu, P.V.S.J. Krishna Teja, User behavior analysis on
grouping. Per data object is seen as a point in n-dimensional space agriculture mining system, Int. J. Eng. Technol. (UAE) 7 (2) (2018) 37–40.
in the SVM algorithm (where n is your number of features). The [9] J.K.R. Sastry, A.K. Sai, R. Samuel, D.B.K. Kamesh, Architectural models for fault
tolerance within clouds at infrastructure level, ARPN J. Eng. Appl. Sci. 12 (11)
value of each function is the co-ordinate value. Then we differenti-
(2017) 3463–3469.
ate by finding the hyperplane which varies greatly in the two [10] K. Likitha, P.S.G. Aruna Sri, K. Ratna Phanitha. (2017),’A secure data sharing and
groups (Fig. 2). revocation in cloud using IDE’, Int. J. Appl. Eng. Res.,12(Special Issue 1),PP.228-
235.
[11] J.G. Barbedo, Arnal,, A new automatic method for disease symptom
4. Results segmentation in digital photographs of plant leaves, Eur. J. Plant Pathol.
(2016) 1–16.
[12] Bhong, S. Vijay, B.V. Pawar, Study and analysis of cotton leaf disease detection
This paper uses the K-means algorithm for the image Classifica- using image processing, Int. J. Adv. Res. Sci. Eng. Technol., 3 (2), 2016.
tion segmentation and SVM. Results obtained is the name of the [13] N. Belsha, N. Hariprasad, ‘‘An approach for identification of infections in
disease solution for the disease [14]. The following table shows vegetables using image processing techniques”, Innovations in Information
Embedded and Communication Systems (ICIIECS) 2017 International
the system output. This is the system accepts the input images
Conference on, pp. 1-6, 2017.
from the test folder, The images are unknown to the SVM classifier. [14] Loyce Selwyn Pinto, Argha Ray, M. Udhayeswar Reddy, Pavithra Perumal, P
Features of images based on previously [15] trained image features Aishwarya, ‘‘Crop disease classification using texture analysis”, Recent Trends
in Electronics Information & Communication Technology (RTEICT) IEEE
and it produces the images of the input leaf for the system used is
International Conference on, pp. 825-828, 2016.
Corn cercospora-leaf-spot, common corn rust, black grape rot, [15] Aparajita, Rudransh Sharma, Anushikha Singh, Malay Kishore Dutta, Kamil
blight of grape leaf, bacterial peach spot, pepper-bell bacterial spot, Riha, Petr Kriz, ‘‘Image processing based automated identification of late blight
bacterial spot tomatoes and tomatoes it’s late blight. The result for disease from leaf images of potato crops”, Telecommunications and Signal
Processing (TSP) 2017 40th International Conference on, pp. 758-762, 2017.
the above mentioned plants shall be produced by System with 75%
accuracy.
Further Reading

5. Conclusion [1] S. Sakhamuri, A. Virupakshi, V. Pushpalatha, D. Nagamani, Elimination of

redundant data in cloud with secured access control, Int. J. Innov. Technol.
Explor. Eng. 8 (6) (2019) 1344–1348.
The identification by image analysis of plant leaf diseases helps [2] K. Indumathi, R. Hemalatha, S. Aasha Nandhini, S. Radha, ‘‘Intelligent plant
to find the disease at an early stage. Automated disease detection disease detection system using wireless multimedia sensor networks”,
reduces the work of monitoring and identifies an early stage ill- Wireless Communications Signal Processing and Networking (WiSPNET)
2017 International Conference on, pp. 1607-1611, 2017.
ness. Infected Leaf Image Dataset identified for tomatoes, corn,