Proceedings of the Fourth International Conference on Trends in Electronics and Informatics (ICOEI 2020)

IEEE Xplore Part Number: CFP20J32-ART; ISBN: 978-1-7281-5518-0

Leaf Disease Detection and Classification by

Decision Tree
B. Rajesh, M. Vishnu Sai Vardhan, L. Sujihelen
Department of Computer Science and Engineering
Sathyabama Institute of Science and Technology, Chennai.

Abstract-Agricultural productivity is very dependent on which occurs when they came on the leaves of plants very
the economy. Plant diseases play an important role in early. More researchers have proposed leaf disease
agriculture because plant diseases are very natural and detection techniques. The existing systems have less
failure to care will have serious consequences for plants detection accuracy. This proposed system uses a decision
and therefore affect the quality, quantity, or productivity tree to identify and classify leaf disease and increases its
of the product. Timely and accurate diagnosis of leaf detection accuracy with less time compared with the
diseases plays a major part in preventing loss in existing system.
productivity and loss or reduction of agricultural
products. Detection of plant diseases by automated KEYWORDS: Plant disease and pest detection, Pre-
techniques is beneficial because it reduces monitoring processing, Feature Extraction, Decision Tree Method,
efforts on large plants and detects an indication of disease
Classification.

1. INTRODUCTION Brown and yellow spots, and early and late starch are
frequently observed in plants, while others are fungal,
Machine Learning is the scientific research of viral, and bacterial. Image analysis is used to assess the
statistical procedures and methods which they are used area affected, and to determine the colour variation of
by computer systems designed to perform such the area affected. Image segmentation is the mechanism
functions without specific instructions, rather than whereby an image is divided or clustered into several
trusting in the models and conclusions. This is believed sections. Today there are many ways to group images,
to be part of artificial intelligence[3][12]. Machine from simple threshold methods to sophisticated
Learning algorithms set up a mathematical model based methods for segmenting images. The process of
on data examples called "training data" to make segmentation is based on various features in the image.
predictions without the completion of a task being In proposed system leaf infection detection and
explicitly programmed. diagnosis is made through a decision tree method. This
India is an agriculture-dependent country. situation is proved using execution time calculation. In
70% of the Indian economy depends on agriculture but the literature, studies are usually carried out with
leaf infection phenomena cause the loss of major crops synthetic images of plants and pests. In this dataset, the
results in economic loss[11][13]. In plants, diseases proposed system was tested, and the results clearly
usually occur on leaves, fruit, buds, and branches. This show that it can be used in real applications. The study
situation causes plants to be damaged. For this reason, it is structured according to the following manner. While
is very important to determine the disease first and take the related work is outlined in detail in Section 2,
the necessary precautions before spreading to other Section 3 presents the model proposed. Section 4
trees[1][2]. Therefore, the fight against diseases and reports discussions and results. The conclusion is finally
plant pests is the most important issue in agriculture. set out in Section 5.
The current method for detecting plant diseases is a 2. RELATED WORKS
simple observation with the naked eye of professionals
who identify and detect plant diseases[4][5] This The authors in the paper [6] illustrate that in the
requires a large team of experts and ongoing monitoring processing flow developed, where the initial colour
from the factory, which spends a lot of money on large conversion form is made for RGB picture input,
companies. Therefore, consulting experts are even cost because RGB is used for, to create colours or change
and time taking process. In such circumstances, this images. Change RGB as a colour descriptor, using HSI.
proposed method is useful in observing the area. Also, Next, the removal of green pixels and masking is done
image processing is maintained to allow automatic using previously calculated threshold levels for the
monitoring, review of processes and robot guidance useful segments extracted at this stage, first when the
based on images. image is segmented. At last, segmentation happens. In
Malusdomestica, the authors present evidence of illness
Visual recognition of plant diseases is even using effective methods such as grouping K, texture and
riskier and less accurate and can particularly be done in colour analysis [7]. Classification and detection of
some areas. Using self-detection techniques requires different ponds using the texture and colour
very little effort, very little time and more accurate.

978-1-7281-5518-0/20/$31.00 ©2020 IEEE 705

Authorized licensed use limited to: University of Vermont Libraries. Downloaded on July 26,2020 at 04:58:33 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Fourth International Conference on Trends in Electronics and Informatics (ICOEI 2020)
IEEE Xplore Part Number: CFP20J32-ART; ISBN: 978-1-7281-5518-0

characteristics normally found in normal and affected The segmentation algorithm takes a picture as a source
areas. The authors describe an algorithm that uses and presents a specific collection. Feature extraction is a
image processing techniques to segment disease spots type of size reduction that effectively shows interesting
on plant leaves [8]. Images use a media filter that calms parts of the image as a concise vector. This approach is
them down. In the final step, threshold calculations can useful when large image sizes and reduced performance
be made using the Otsu method for colour components are needed for fast tasks such as matching and taking
to find the location of the disease. To eliminate that pictures. Decision Tree Algorithms are used in this
noise, the CIELAB colour model is used. Kulkarni et al. proposed method. This algorithm serves to identify
introduce early detection methods and precise detection diseases of the plant leaf. The knowledgebase is a
of plant infections using neural networks and multiple collection of data organized in a form that facilitates
optimization techniques. The classification based on analysis through an automatic deductive process. The
ANN classifies different plant diseases and uses a mix Machine Learning Classifier used here is a decision tree
of texture, shape, and properties to correctly identify algorithm.
this disease [9]. Mrunalini et al. [10] develop methods
for the diagnosis and detection of various diseases Decision tree classifiers are used to identify healthy and
involving plants. The neural network is often used to unhealthy leaves. To remove branches where the
detect disease automatically within the leaves. The algorithm is 25 percent or safer, the standard 25 percent
suggested solution will contribute significantly to the pruning reliability is used to prevent the presence of
accurate identification of leaves and seems to be an certain branches that will not represent aggregation.
effective alternative to resolving vapor and root disease, This prevents data from overflowing. The minimum
which needs fewer analytical effort. case parameter is the stop criterion for further data
division into particular decision nodes. Trees growing at
3. PROPOSED SYSTEM a given node of decision are stopped when at least one
class has the same or fewer cases than the minimum
The main purpose of this proposed system is to detect case. As a result, trees cannot be divided into nodes that
the leaf disease and identify what type of disease. are too specific and contain little supporting data. A
mean receiver operating curve (ROC) was created to
analyze the sensitivity and specificity of the solution
algorithm. The area under the curve (AUC) acts as a
measure of the algorithm's overall performance. Along
with the high AUC, the algorithm with maximum
sensitiveness was chosen.
In results, a decision tree with 1,000 leaves was built for
leaf detection. The resultant diagnostic algorithm is
demonstrated in Fig.2 of Healthy and Unhealthy leaves.

Fig. 1 Process for the proposed system

The proposed work uses a decision tree to identify the
leaf disease. The proposed work is based on the
morphological characteristics of plant leaves. Fig. 1
describes the step by step procedure for the proposed
system. Fig.2 Classification
In the first step, where images are taken as 4. RESULTS AND DISCUSSION
input. Image input is pre-processed to improve image
quality and to avoid unwanted image distortion. To classify leaf diseases a collection of leaf
Trimming the sheet image is done to get an interesting images is required. The images are downloaded from
area of the image, and then the image is smoothed using the kagle.com. The images from the dataset are taken
a refinement filter. Image enhancement is also done to for training the model. It contains a collection of images
increase contrast. Segmentation involves dividing an taken in different environments. The dataset contains
image according to several criteria into groups of pixels. more than 1000 images of the leaf. The samples of the

978-1-7281-5518-0/20/$31.00 ©2020 IEEE 706

Authorized licensed use limited to: University of Vermont Libraries. Downloaded on July 26,2020 at 04:58:33 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Fourth International Conference on Trends in Electronics and Informatics (ICOEI 2020)
IEEE Xplore Part Number: CFP20J32-ART; ISBN: 978-1-7281-5518-0

dataset are shown in table 1. The model is trained by

healthy images in fig.3 and unhealthy images in fig.4.
The proposed system will detect unhealthy leaf and
report the remedies to the farmer. The proposed system
detection accuracy is above 95%.
Table 1 Dataset for Classification

Fig.6 Analyse an Image

Table 2 Sample Output

Fig.3 Sample Image of Healthy Leaves

The sample outcome of the given work is shown in

Fig.5. To check the leaf diseases, first, the model should
be trained with healthy and unhealthy leaf images. The
leaf should be tested by taking the real-time image of
Fig.4 Sample Images of Unhealthy Leaves the leaf is shown in fig.5. Then the leaf image is
analyzed for the healthy or unhealthy leaf. If it is an
unhealthy leaf then it will find what type of disease and
remedies for the unhealthy leaf in Fig.6 and Table 2.
4.1 Experimental Results
The Detection Accuracy is measured by taking
5 different types of leaves and 100 images that are taken
for training and a few images which are taken for
testing. The proposed method compares with two
existing algorithms K-Mean and SVM shown in Table
3. The accuracy of detecting healthy and un-healthy
leaves is shown in fig.7 and fig.8.
Table 3 Detection Accuracy

Leaf Images which Images which K S Decisi

Samp are used for are used for me V on
les training testing an M tree
Fig.5 Input of tested leaf Tomat
o 100 45 80 90 96
Lemo
n 100 50 82 85 95

978-1-7281-5518-0/20/$31.00 ©2020 IEEE 707

Authorized licensed use limited to: University of Vermont Libraries. Downloaded on July 26,2020 at 04:58:33 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Fourth International Conference on Trends in Electronics and Informatics (ICOEI 2020)
IEEE Xplore Part Number: CFP20J32-ART; ISBN: 978-1-7281-5518-0

[3] Chalapathy R., Chawla S. (2019). In-depth learning to detect

Rose 100 40 81 87 94.5 anomalies: survey. arXiv: 1901.03407.
Papay [4] Ching T., Beaulieu-Jones B.K., Kalinin A.A., Do B.T.,
a 100 56 83 89 95.2 Himmelstein D.S., Way G.P. et al. , (2018). Opportunities and
Banan obstacles for comprehensive education in biology and medicine.
a 100 64 75 90 95.6 J. R. Soc. The interface. 15: 20170387. 10.1098 / rsif.2017.0387
[5] Knillmann S., Liess M. (2019). Effects of pesticides on
ecosystems. Cham: Springer International Publishing.
[6] Dhaygude Sanjay B, Kumbhar Nitin P (2013), "Detection of leaf
plant diseases through image processing", Int J Adv Res
Electron Electro Instrum, Volume 2 (1).
[7] Bashir Sabah, Sharma Navadepe (2012). Remote detection of
plant diseases through image processing. IOSR J Electron
Commun Eng; 2 (6): 31-4. ISSN: 2278-2834.
[8] Chaudhary Piyush et al. (2012). Color transformation based
approach for detecting diseases in the leaves of IntComput
SciTel Telecommun; 3 (6).
[9] Kulkarni Anand H., Ashwin Patil RK (2012). The use of image
processing technology to detect plant diseases. Input J Mod Eng
Res; 2 (5): 3661-4.
[10] Mrunalini R Badnakhe, Deshmukh Prashant R (2011). The use of
K-means grouping and artificial intelligence to identify patterns
of cultural disease. Int ConfAdvInfTechnol; IPCSIT.
Fig. 7 Detection Accuracy [11] Christina, Ruth, Greeshma Liz Shajan, and B. Ankayarkanni.
"CART-A Statistical Model for Predicting QoE using Machine
Learning in Smartphones." In IOP Conference Series: Materials
Science and Engineering, vol. 590, no. 1, p. 012001. IOP
Publishing, 2019.
[12] M. J. N. V. S. K. Asrith, K. P. Reddy and Sujihelen, "Face
Recognition and Weapon Detection from Very Low Resolution
Image," 2018 International Conference on Emerging Trends and
Innovations In Engineering And Technological Research
(ICETIETR), Ernakulam, 2018, pp. 1-5.
[13] Chandy, Abraham. "RGBD ANALYSIS FOR FINDING THE
DIFFERENT STAGES OF MATURITY OF FRUITS IN
FARMING." Journal of Innovative Image Processing (JIIP) 1,
no. 02 (2019): 111-121.

Fig. 8 Detection Accuracy for detecting the exact

disease
5. CONCLUSION
This proposed system provides information on
classification technique which is used to detect diseases
in plant leaves, as well as algorithms for image partition
techniques which is to be used for detection and
categorization of diseases in leaves finally. Another
benefit of this process is that plant infections can be
detected in the early stage. This proposed work
performs better than the existing system when
compared with the detection accuracy. The extension of
this research would concentrate on the development of
improved algorithms for the rapid and effective
identification of diseased leaves.
REFERENCES:
[1] Barbedo J.G.A. (2018a). Factors that influence the use of
comprehensive training in detecting plant diseases. Biosyst. 172,
84-91. 10.1016 / j.biosystemseng.2018.05.013
[2] Barbedo J.G.A. (2018b). Effect of size and variation of data sets
on the effectiveness of transfer training and depth for
classification of plant diseases. Comput Electron. Agric 153, 46-
53. 10.1016 / j.compag.2018.08.013

978-1-7281-5518-0/20/$31.00 ©2020 IEEE 708

Authorized licensed use limited to: University of Vermont Libraries. Downloaded on July 26,2020 at 04:58:33 UTC from IEEE Xplore. Restrictions apply.