3URFHHGLQJRI,QWHUQDWLRQDO&RQIHUHQFHRQ6\VWHPV&RPSXWDWLRQ$XWRPDWLRQDQG1HWZRUNLQJ

LEAF DISEASE DETECTION AND FERTILIZER SUGGESTION

Indumathi.R Saagari.N Thejuswini.V Swarnareka.R

Assistant Professor, Department of Computer Department of Computer Department of Computer
Department of Computer Science Science and Engineering, Science and Engineering, Science and Engineering,
and Engineering, Manakula Manakula Vinayagar Institute Manakula Vinayagar Institute Manakula Vinayagar
Vinayagar Institute of of Technology, of Technology, Institute of Technology,
Technology, Puducherry. Puducherry. Puducherry.
Puducherry. saagi.jasmine@gmail.com thejuswini98@gmail.com swarnareka3@gmail.com
indu5490@gmail.com

Abstract— The field of agriculture is in a provide useful information to control it. This system
great threat this includes the diseases that attack the does few image pre-processing techniques like image
plant leaf. Our system finds the area of leaf that has acquisition, image segmentation, feature extraction and
classification. Modern agricultural practices assure great
been affected and also the disease that attacked the
development of cultivation. We have many smart
leaf. This is achieved by using Image Processing; there agriculture developing models used as a real time
are systems that predict the diseases in the leaf. Our monitoring systems to monitor the temperature,
humidity, moisture content in the soil using various
system uses K-Medoid clustering and Random Forest
sensors. We have many systems that do work
algorithm to produce more accuracy in the detection of automatically but there are very few systems that detect
disease in the leaf. The image is first pre-processed and problems and provide suggestions to those problems.
One such automatic disease detection system is
then the clustering method is applied to find the
developed with the help of image processing. Image
affected area of the leaf. This is then processed to fetch processing is applied in such a way that even a spot in
13 characters like Mean, SD, Entropy, RMS, Variance, the leaf is detected. The system detects almost all
Smoothness, Kurtosis, Skewness, IDM, Contrast, possible diseases and also addresses the severity levels of
the disease.
Correlation, Energy and Homogeneity through this we
will measure the accuracy and find the disease. Mostly plant leafs are infected due to
pathogens, fungi, bacteria and viruses. In recent years,
Keywords- Image Processing, Clustering, most of the disease detection systems were developed
Classification, Detection of Leaf Disease, Accuracy, based on image processing. These systems input should
Pre-processed be the set of affected plant images that are saved in the
database. In this paper we are taking the images of the
I. INTRODUCTION leaf that are already captured. The images are given to
the system after that using image processing techniques
A system that automatically detects leaf disease the image is enhanced to increase the accuracy in finding
with the help of image processing is being developed. the disease. The techniques we are using are clustering,
This in turn helps the farmers in the identifying the classification, feature extraction and segmentation. The
diseases at an early stage and diseases will vary based on the size, color, shape and the
virus affected. System takes lots of

1
#,(((
3URFHHGLQJRI,QWHUQDWLRQDO&RQIHUHQFHRQ6\VWHPV&RPSXWDWLRQ$XWRPDWLRQDQG1HWZRUNLQJ

infected parts from the images and those images are put of leaf diseases has been introduced, this method is based
into image processing techniques. As there are various on K-means as a clustering procedure and ANNs
systems that detect leaf diseases, accuracy matters the (Artificial Neural Networks) as a classifier tool using
most. The accuracy of the system is improved by using some texture feature set .The aim of this work is
the best algorithmic pair. threefold:

II. RELATED WORK 1) Identifying the infected object(s) based upon

K-means clustering.
PLANT DISEASE DETECTION AND
CLASSIFICATION USING IMAGE 2) Extracting the feature set of the infected
PROCESSING AND ARTIFICIAL NEURAL Leaf images.
NETWORK, IRJET VOLUME 05, ISSUE 06,
JUNE 2018 3) Detecting and classifying the type of disease
using ANNs(Artificial Neural Networks).
In this paper a software solution for fast,
accurate and automatic detection and classification of LEAF DISEASE DETECTION USING IMAGE
plant diseases through Image Processing is presented [1]. PROCESSING IRJET VOLUME 05 ISSUE 12
Identification of the plant diseases is the key to DECEMBER 2018.
preventing losses in the quality and quantity of the
agricultural product. Health monitoring and disease Doing study on the disease severity or harshness
detection of plant is critical for sustainable agriculture. of leaf using image processing techniques. They used
The typical method of studying plant disease is to rely on feature extraction methods such as threshold and
visually observable patterns on the plant leaves. Visually triangular threshold methods. Identification of diseased
identifying plant diseases is inefficient, difficult, time leaf of brown and blast spot of leaf image processing
consuming, requires expertise in plant diseases and techniques were carried out by. They used zooming
continuous monitoring which might be expensive in algorithm, SOM neural network for disease detection.
large farms. Therefore; a fast, automatic and accurate
method to detect plant disease is of great importance.
The authors made investigation on diseases (for e.g.
Hence, image processing technique is employed for the
Early scorch, Ashen mould, Late scorch, Cottony mould
detection of plant diseases. The implementation of these
and Ting whiteness diseases) of plants using K- Means
technologies will lead to improved productivity India has
clustering, Back propagation algorithm. Made study on
a diverse agricultural sector. Agriculture plays a vital
diseases using an image processing technique which
role in India’s economy and over 58 per cent of the rural
involves morphological processing, colour clustering.
households depend on agriculture as their principal
Leaf disease detection of orchid leaf (for e.g.) such as
means of livelihood. Research in agriculture is aimed
Black leaf spot and Sun scorch was carried out by. They
towards increase of productivity and quality of food.
applied pattern classification and border segmentation
TECHNIQUES USED techniques for detection of diseased on leaf. The present
work has been carried out for the automatic disease
There are two main characteristics of plant- detection of plant leaf and provide solution about the
disease detection machine learning methods that must be diseases that solution can be provided using the deep
achieved, they are: speed and accuracy. In this study an study about the diseases and the soil moisture level
automatic detection and classification detection and also the temperature level detection
whether that crop get

2
#,(((
3URFHHGLQJRI,QWHUQDWLRQDO&RQIHUHQFHRQ6\VWHPV&RPSXWDWLRQ$XWRPDWLRQDQG1HWZRUNLQJ

affected by which reason exactly to determined A NOVEL MACHINE LEARNING BASED

solutions. APPROACH FOR DETECTION AND
CLASSIFICATION OF SUGARCANE PLANT
TECHNIQUES USED DISEASE BY USING DWT IRJET VOLUME 04
ISSUE 12 DECEMBER 2017
Classification is used in the interpretation of the
extracted diseased region in an image which helps in the Input has to be the set of affected plant images that are
identification of the type of disease infection in leaves. In saved in the database. These images are captured by the
our analysis back propagation neural network (BPNN) is camera by continuously monitoring the field. In this
used which build association between known pattern of paper we are taking already captured images .The images
input and specific output. The input layer analyses the are given to the system after that using image processing
diseased region while the output layer specifies the techniques the quality of the image is increased. And by
disease outcome of the affected region. A hidden layer increasing the accuracy for finding the disease is the task
occurs in between the input and output layer which to be improved. The techniques we are using are
provides connecting link between the input and output classification, feature extraction, segmentation. The
images. It is applied to obtain least error in the symptoms of the diseases will vary with respect to size,
classification of disease of the affected region. colour, shape based on the virus affected. Here in this
paper we are solving the problem of feature extraction
LEAF DISEASE DETECTION AND SELECTION OF
and pre-processing by enhancing the image quality.
FERTILIZERS USING ARTIFICIAL NEURAL
System takes lots of infected part images from crops are
NETWORK IRJET VOLUME 04 ISSUE 06 JUNE
else downloaded from the Flickr and Google and those
2017
images are going through the image processing
Plant disease especially on leaves is the one of the major techniques .In the recent times improving the accuracy is
reagent of reduction in both quality and quantity of the main thing to detect the disease faster. Image
food crops. The quality and quantity of food production segmentation is main technique to detect the affected
become reduced only because of pest’s presence in the area using threshold value. Here according to the pixel of
crops and leaves. Thus it leads to increase in difficulty, the image the threshold value have to be assigned
food insecurity and fatality rate. In modern years in order according to the pixel the system will compare with
to identify the plant disease, so many different concepts images that saved in the dataset. By using DWT
of image process technology have been adapted. One of algorithm the pixel of the image is to be compared and
the dominant dilemmas for agronomist is to reduce and remove the unused pixels using Fourier transform by
destroy the progress of pests affecting the crop harvest. forming a matrix over the image. Most of the plants will
The common pests like aphids, fungus, gnats, flies, be in green pixels if its changed means that part is
slugs, snails, caterpillars, etc these pests are most affected by some virus or fungus. The proposed method
commonly determined in the plant disease. Almost all is the improvement of the approach which will increase
agriculturists are used to identify pests systemically over the accuracy and quality of the image. According to the
examination by using their eyes but this access is high severity of the situation the system will give the simple
and it takes some time. The techniques of digital image recommendations and notifications to suggest some
processing have been organized in the agriculture field in pesticides and mention the fungus that is affected.
order to analyze the purposes in different agricultural
applications like plant recognition, crop yield estimation,
soil quality estimation etc. Feature extraction method of
neural network with GLCM approach was also
established in order to detect the crop disease. The
extraction of texture feature will be carried out by both
using GLCM.

#,(((
3URFHHGLQJRI,QWHUQDWLRQDO&RQIHUHQFHRQ6\VWHPV&RPSXWDWLRQ$XWRPDWLRQDQG1HWZRUNLQJ

III. EXISTING
JJJ. SYSTEM
The existing system uses K-Means clustering and
Support Vector Machine classification algorithm to Input
detect the disease in the leaf of the image uploaded to the
Image Pre-
Image Processing
system. The system does usual preprocessing steps
followed by clustering and classification. This system
gives an accuracy of about 80%. This approach
significantly does not support an accurate detection of
leaf disease since this system is very time consuming.
The system has failed in improving the accuracy of leaf
disease detection. The existing system uses 11 features
such as Mean, SD,Variance, Smoothness, Kurtosis,
Skewness, IDM, Contrast, Correlation, Energy and
Clustering
Homogeneity to determine the accuracy.

Classification

Output

Fig.3 Architecture Diagram of Existing System

IV. PROPOSED SYSTEM

Our system uses K-Medoid clustering and
Random Forest classification algorithms to detect
the leaf diseases. This system is found to efficient
in finding the disease accurately. The working of
the system is as follows: The system
maintains an image dataset of infected and
healthy leaves. The entire dataset is divided
into train and test dataset
on a random basis. With this data the system is
trained to find the diseases in the leaf.

#,(((
4
3URFHHGLQJRI,QWHUQDWLRQDO&RQIHUHQFHRQ6\VWHPV&RPSXWDWLRQ$XWRPDWLRQDQG1HWZRUNLQJ

Feature Extraction

The system takes the input of an infected leaf Image

and the contrast of the image is enhanced followed by Acquisition
few pre processing steps. The reason for enhancing the
contrast of the image is to get clarity of the image in
detail. Then the R, G, B colors of an image is plotted as a
bar graph. Image Pre-
Processing
After enhancing the image, the pixels of the
image is clustered randomly. We have to select the best
clustered image that shows us the infected parts of the
leaf clearly. On choosing the best clustered image we can
find the percentage of affected area.

Followed by this, 13 features like Mean, SD, Segmentation

Entropy, RMS, Variance, Smoothness, Kurtosis,
Skewness, IDM, Contrast, Correlation, Energy and
Homogeneity of the leaf is calculated. Then the image is
classified by comparing the image with the existing
dataset. This results in the type of disease that has Feature Extraction
infected the leaf.

As an enhancement our system also prescribes

the fertilizer to be used to treat the leaf in order to save it.
This is done with the help of dataset that holds the leaf
Clustering
disease name along with the fertilizers that can be used
to treat it. This dataset is used to provide us a suggestion
for the fertilizer that can be used.

This is a normal search process which acts a

dictionary with key value pair. The name of the disease Classification
is treated as a key by searching the key we get the
equivalent value which is the fertilizer as a result.

Detected Leaf
Disease
O
u
t
p
u Fertilizer
t Suggestion

Fig.4 Architecture Diagram of Proposed System

#,((( 5
3URFHHGLQJRI,QWHUQDWLRQDO&RQIHUHQFHRQ6\VWHPV&RPSXWDWLRQ$XWRPDWLRQDQG1HWZRUNLQJ

Journal of Science and Technology,December

V.CONCLUSION 2016

With the algorithms used and the dataset [5] Wenjiang Huang, Qingsong Guan, JuhuaLuo,
provided the systems makes the best attempt to find the JingchengZhang,VJinling Zhao, Dong Liang,
disease of the input leaf image and the fertilizer that can Linsheng Huang, and Dongyan Zhang, “New
be used to treat it. The accuracy of the system stands Optimized Spectral Indices for Identifying and
high with the ability of detecting the disease. The time Monitoring Winter Wheat Diseases”, IEEE
taken for computing the disease in the infected leaf is journal of selected topics in applied earth
reduced in this system and the memory consumption is observation and remote sensing,Vol. 7, No. 6,
also manageable. As the trained dataset grows the June 2014
efficiency of the system increases which in turn increases
the accuracy in finding the disease. Thus the system [6] Dr.K.Thangadurai, K.Padmavathi,
stands for the betterment of farmer’s welfare thereby “Computer Visionimage Enhancement For Plant
increasing the production and the economy of the Leaves Disease Detection”, 2014 World
country. This system can be further enhanced by Congress on Computing and Communication
changing different algorithmic pair to find the disease in Technologies.
a more accurate way.
[7] Monica Jhuria, Ashwani Kumar, and
REFERENCES: RushikeshBorse, “Image Processing For Smart
Farming: Detection Of Disease And Fruit
[1] Guiling Sun, Xinglong Jia, and TianyuGeng, Grading”, Proceedings of the 2013 IEEE Second
“Plant Disease Recognition Based on Image International Conference on Image Information
Processing Technology”,Journal of Electrical Processing (ICIIP-2013)
and Computer Engineering Volume, 2018.
[8] Zulkifli Bin Husin, Abdul Hallis Bin Abdul
[2] Sujatha R, Sravan Kumar Y, Garine Uma Aziz, Ali Yeon Bin MdShakaffRohaniBinti S
Akhil, “Leaf Disease Detection using Image Mohamed Farook, “Feasibility Study on Plant
Processing”, Journal of Chemical and Chili Disease Detection Using Image Processing
Pharmaceutical Sciences, volume 10, Issue 1, Techniques”,2012ThirdInternational
March 2017. Conference on Intelligent Systems Modelling and
Simulation.
[3] Rajneet Kaur, Manjeet Kaur, “A Brief
Review on Plant Disease Detection using in [9] Mrunalini R. Badnakhe, Prashant R.
Image Processing” International Journal of Deshmukh, “Infected Leaf Analysis and
Computer Science and Mobile Computing, Comparison by Otsu Threshold and k-Means
February 2017. Clustering”, International Journal of Advanced
Research in Computer Science and Software
[4] Iqbaldeep Kaur, Gigty Aggarwal and Amit Engineering, Volume 2, Issue 3, March 2012.
Verma, “Detection and classification of Disease
Affected Region of Plant Leaves using Image [10] H. Al-Hiary, S. Bani-Ahmad, M. Reyalat,
Processing Technique”, Indian M. Braik and Z. ALRahamneh, “Fast and
Accurate Detection and Classification of Plant
Diseases”,

6
#,(((
3URFHHGLQJRI,QWHUQDWLRQDO&RQIHUHQFHRQ6\VWHPV&RPSXWDWLRQ$XWRPDWLRQDQG1HWZRUNLQJ

International Journal of Computer

Applications (0975 – 8887)Volume 17–
No.1, March 2011

[11] Chunxia Zhang, Xiuqing Wang, Xudong Li,

“Design of Monitoring and Control Plant Disease
System Based on DSP&FPGA”, 2010 Second
International Conference on Networks Security,
Wireless Communications and Trusted
Computing.

[12] A. Meunkaewjinda, P. Kumsawat, K.

Attakitmongcol and A. Srikaew, “Grape leaf
disease detection from color imagery using hybrid
intelligent system”, Proceedings of ECTI-CON
2008.

[13] SantanuPhadikar and Jaya Sil, “Rice

Disease Identification using Pattern Recognition”,
Proceedings of 11th International Conference on
Computer and Information Technology

[14] WeizhengS.,Yachun W.,Zhanliang C.&Hongda

W , ”Grading Method Of Leaf Spot Disease Based
OnImageProcessing”

#,((( 7