International Journal of Scientific Research in Engineering and Management (IJSREM)

Volume: 08 Issue: 04 | April - 2024 SJIF Rating: 8.448 ISSN: 2582-3930

PLANT LEAF DISEASE DETECTION

Kaustubh Warke
Rushikesh Tharkar Department of Information Technology
Department of Information Technology VPPCOE&VA, University of Mumbai
VPPCOE&VA, University of Mumbai [email protected]
[email protected]
Pratik Pathak
Rahul Jadhav Department of Information Technology
Department of Information Technology VPPCOE&VA, University of Mumbai
VPPCOE&VA, University of Mumbai [email protected]
[email protected]

Abstract - Plant diseases affect the growth of their Diseases are a major threat to the crops.Diseases destroy the crop
respective species, therefore their early identification is very and thus reduce the yield and also the quality of the crops thereby
important. Many Machine Learning (ML) models have been decreasing the profits. The use of technology in the area of
employed for the detection and classification of plant agriculture will reduce manual labour and increase the accuracy
diseases but, after the advancements in a subset of ML, that by decreasing the amount of error. The systems developed earlier
is, Deep Learning (DL), this area of research appears to have just been able to detect diseases of leaves, this system will
have great potential in terms of increased accuracy. Many use image processing and classifiers which include CNN
developed/modified DL architectures are implemented architecture available for image classification to detect the
diseases.
along with several visualization techniques to detect and
classify the symptoms of plant diseases. Moreover, several
performance metrics are used for the evaluation of these Predictive models play a crucial role in plant disease management.
architectures/techniques. This review provides a Predicting the likelihood and timing of disease outbreaks based on
comprehensive explanation of DL models used to visualize historical data, weather conditions, and other environmental
various plant diseases. In addition, some research gaps are factors. The predictive models based on the training given to them
can provide instant and immediate results in predicting plant
identified from which to obtain greater transparency for
diseases which can help plant owners or farmers who does farming
detecting diseases in plants, even before their symptoms in reducing the damage done to plants.
appear clearly
High Accuracy in Disease Identification: The foremost objective
Keywords: Plant leaf disease detection, leaf disease is to achieve a high level of accuracy in identifying plant diseases.
detection, convolutional neural network, deep learning CNNs are powerful in image recognition tasks and can be fine-
tuned to accurately detect various types of diseases present on
I. INTRODUCTION plant leaves. The goal is to ensure that the model correctly
identifies and classifies diseases, minimizing false positives and
Agriculture plays an important role in the development of the false negatives.
economy of a country. It provides lots of opportunities to
innovate and apply the innovation. Robustness and Generalization: The model's performance
shouldn't be limited to the conditions under which it was trained.
When Plants or Crops get infected to a specific disease their Robustness is crucial. The objective is to ensure that the model
quality of life begins to deteriorate and if the disease is not can generalize well to new, unseen data, regardless of factors like
cured in early stage then the plant cannot survive anymore due different lighting conditions, angles, or variations in plant types,
to the increased rate of the disease in the plants. leaf shapes, or diseases. This could involve strategies such as data
augmentation, transfer learning, or diverse dataset collection.

© 2024, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM31382 | Page 1

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 08 Issue: 04 | April - 2024 SJIF Rating: 8.448 ISSN: 2582-3930

II. LITERATURE REVIEW Discuss the limitations of traditional methods, such as time-
consuming and subjective nature.
1. Validation: The model's performance is evaluated on a Explore the application of computer vision techniques for plant
separate validation dataset to tune hyperparameters and leaf disease detection.
monitor overfitting. Review image processing algorithms used for feature extraction
and classification.
2. Testing: The trained CNN is tested on a different set of
images not seen during training to assess its generalization
performance.
Discuss the advantages of computer vision techniques, such as
3. Deployment: Once the model performs well, it can be
speed and scalability.
deployed in real-world scenarios. This can be on-site using
Discuss the use of machine learning algorithms for plant leaf
cameras, drones, or smartphone apps to detect diseases in
disease detection.
plant leaves.
Review different types of machine learning algorithms employed,
4. CNN Model: A CNN architecture is designed for plant leaf such as supervised, unsupervised, and deep learning.
disease detection. Common architectures include VGG, Highlight the importance of datasets in training machine learning
ResNet, Inception, or custom designs. The model is trained on models.
the dataset to learn to classify images as healthy or diseased. Focus specifically on the application of deep learning techniques,
particularly CNNs, for plant leaf disease detection.
As of late, image processing techniques are utilized in Discuss the architecture of CNNs and their effectiveness in feature
different fields such as automation, medical etc. Indeed, even learning from images.
the distinguishing proof of plant contamination utilizing Review state-of-the-art CNN models used for plant leaf disease
customary strategy is supplanted by image processing. The detection.
image processing frameworks requires camera, PC and Identify challenges faced in plant leaf disease detection, such as
essential software. Steps associated with plant disease dataset availability, model generalization, and robustness to
detection are image acquisition, pre-processing, environmental variations.
segmentation, feature extraction and classification. Discuss potential future directions for research, such as multi-
Performing image enhancement, improves the quality of the modal approaches combining image analysis with other sensor
picture just as the clearness. Basic primary colours red, green data.
and blue combinations produce many varieties of colours. Summarize the key findings of the literature survey.
Hence, implementing image processing using RGB Highlight the significance of automated plant leaf disease
components is difficult and its range is very high. Changing detection methods.
over RGB picture into its proportionate grey image is Suggest areas for further research and development.
accomplished for simpler implementation. Automated plant Data Collection: Gather a diverse dataset of plant images,
disease using image processing technique is beneficial for the including both diseased and healthy samples, from various
farmers as it reduces large human labours and can help to sources.
detected by symptoms at early stage MATLAB software’s Data Preprocessing: Clean and standardize the dataset, including
tools are utilized for recognizing the disease of the plants. image resizing, contrast adjustment, and quality enhancement to
Image acquisition is performed using digital cameras. K-mean ensure consistency.
clustering algorithm used Euclidean distance metric method Feature Extraction: Implement feature extraction techniques to
and clusters the image based on the specified number of capture relevant information from the images, such as textures and
groups [4][5]. Gray-Level Cooccurrence Matrix (GLCM) is shapes.
one of the most popular methods for texture analysis. It Model Development: Design and implement machine learning or
creates an element based gray level matrix for the shading deep learning models, often using techniques like convolutional
picture and measures the spatial separation between the neural networks (CNNs), to analyze the extracted features.
pixels. GLCM represents the distance and angular spatial Training and Validation: Train the models using a portion of the
relationship of an image in a specific size. GLCM calculates dataset, validate their performance using a separate dataset, and
how regularly the pixel with gray level power happens. fine-tune their parameters.
Horizontally values are represented as „i‟ and vertically or Testing and Evaluation: Evaluate the models' performance using
diagonally values to adjacent pixels are labelled as „j‟. a different, unseen dataset to assess their ability to accurately
Define the importance of detecting plant leaf diseases early. recognize plant diseases.
Highlight the significance of automated detection methods. Interpretability and Transparency: Implement methods to provide
Introduce the scope of the literature survey. explanations for the models' decisions, ensuring transparency in
the recognition process.
Review traditional methods used for plant leaf disease
detection, such as visual inspection by experts and laboratory
testing.

© 2024, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM31382 | Page 2

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 08 Issue: 04 | April - 2024 SJIF Rating: 8.448 ISSN: 2582-3930

[1] Revathi, P., Hemalatha, M. (2012 “Classification of cotton S. S. Sannakki and V. S. Rajpurohit [4] suggested a
leaf spot diseases using image processing edge detection Backpropagation Neural Network based mostly classifier
techniques”.169-173.10.1109/ INCOSET.2012.6513900J. (BPNN) for sleuthing the malady in Pomegranate leaf. Features
Clerk Maxwell, A Treatiseon Electricity and Magnetism, 3rd have been designated as color and texture. BPNN detects and
ed., vol.2. Oxford: Clarendon, 1892, pp.68– 73. classifies the diseases with a precision of around ninety seven.30
%. Dr. K. Thangadurai and K. Padmavathi [5] recommended pc
[2] Arivazhagan, S., Newlin S., Ananthi, S., Vishnu V (2013)
vision image improvement for leaf malady identification. It
“Detection of unhealthy region of plant leaves and
includes color conversion and Histogram deed. Histogram deed
classification of plant leaf diseases using texture features”,
Agric Eng. Int CIGR, 15, pp.211-217 will increase the image clarity. RGB to Grayscale conversion is
used to retain the luminance data instead of Hue and Saturation
[3] Gavhale, K., R., and Gawande, U. (2014) “An overview of data. For encoding of linear intensity values, Gamma expansions
the research of plant leaves disease detection using image are used. Cumulative Gaussian distribution operate distributes
processing, IOSRJ, Comut.Eng, pp.10-16. the intensity worth of the image. Histogram deed provides the
[4] Singh, V., Misra, A. (2015) “Detection of unhealthy region higher quality image in Grayscale. [6] proposed an SVM-based
of plant leaves using Image Processing and Genetic Algorithm Multiple Classifier System (MCS)[25] for wheat leaf diseases. It
in Computer Engineering and Applications (ICACEA). uses a stacked generalization structure to join the classification
[5] Arti.N.R, Bhavesh, T., Vatsal, S., (2013), “Image choices obtained from 3 sorts of support vector machines (SVM)
processing Techniques for Detection of Leaf Disease”, based mostly classifiers. The features like color, texture and
International Journal of Advanced Research in Computer shape options area unit used as coaching sets for classifiers.
Science and Software Engineering, vol 3, no.11. Firstly, features area unit classified mistreatment a classifier in
low-level of MCS to corresponding mid-level classes, which will
S. S. Sannakki and V. S. Rajpurohit, proposed a
part notice the symptom of crop diseases in line with the
“Classification of Pomegranate Diseases Based on Back
Propagation Neural Network” which mainly works on the information of plant pathology. Then the mid-level features area
method of Segment the defected area and color and texture are unit generated from these mid-categories generated from low-
used as the features. Here they used neural network classifier level classifiers.
for the classification. The main advantage is it Converts to
L*a*b to extract chromaticity layers of the image and III. PROPOSED METHODOLOGY
Categorisation is found to be 97.30% accurate. The main Fig. 2 displays the Convolutional Neural Network (CNN) is a type
disadvantage is that it is used only for the limited crops. P. R. of deep learning algorithm specifically designed for image
Rothe and R. V. Kshirsagar introduced a” Cotton Leaf Disease processing and recognition tasks. Compared to alternative
Identification using Pattern Recognition Techniques” which classification models, CNNs require less preprocessing as they can
Uses snake segmentation, here Hu’s moments are used as automatically learn hierarchical feature representations from raw
distinctive attribute. Active contour model used to limit the input images. They excel at assigning importance to various objects
vitality inside the infection spot, BPNN classifier tackles the and features within the images through convolutional layers, which
numerous class problems. The average classification is found apply filters to detect local patterns
to be 85.52%. [3] Aakanksha Rastogi, Ritika Arora and Shanu
Sharma,” Leaf Disease Detection and Grading using
Computer Vision Technology &Fuzzy Logic”. K-means
clustering used to segment the defected area; GLCM is used
for the extraction of texture features, Fuzzy logic is used for
disease grading. They used artificial neural network (ANN) as
a classifier which mainly helps to check the severity of the
diseased leaf.
Godliver Owomugisha, John A. Quinn, Ernest Mwebaze and
James Lwasa, proposed” Automated Vision-Based Diagnosis
of Banana Bacterial Wilt Disease and Black Sigatoka Disease
“Color histograms are extracted and transformed from RGB to
HSV, RGB to L*a*b.Peak components are used to create max
tree, five shape attributes are used and area under the curve
analysis is used for classification. They used nearest
neighbors, Decision tree, random forest, extremely
randomized tree, Naïve bayes and SV classifier. In seven
classifiers extremely, randomized trees yield a very high score,
provide real time information provide flexibility to the
application.

Fig. 2. Working of Plant leaf disease system

© 2024, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM31382 | Page 3

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 08 Issue: 04 | April - 2024 SJIF Rating: 8.448 ISSN: 2582-3930

IX. Configuring Disease Reorganization and Pest

Initially, the digital images are acquired from the Recommendation .
circumstances using a digital mobile camera. Then
imageprocessing techniques are applied to the acquired
images to extract RGB Pixel counting features that are
necessary for further analysis. After that, some analytical IV. CONCLUSION
perceptive techniques are used to classify the images
according to the specific problem at hand. In this work Today, the technology is used in every aspect of life.
farmers can take decision immediately at the time. They Farmers are facing difficulties in manually identifying the
want to get the best solution to diseases and pest diseases in plants. Also, expert advice is not available easily
recommendation is 3 languages Tamil, English, Hindi, to the farmers. It is beneficiary to have an automated
Production can be improved, the yield loss can be reduced, mechanism to identify the plant diseases. The proposed
they minimum cost of ultimate system very useful to farmers
method used machine -learning techniques to detect the leaf
and we can increase the economic of the country. Main
farmers life protects and reduces their burden. diseases in plants the plants related sector will grow more
effectively and efficiently due to availability of such type of
automated systems to detect plant diseases.

V. FUTURE WORK
In neural network it’s difficult to understand structure of
algorithm and to determine optimal parameters when
training data is not linearly separable.
Artificial Neural Network and Fuzzy Logic with other soft
computing technique can be used to classify the crop
diseases.
Disease spot area can be computed for assessment of loss
in agriculture crop. Disease can be classified by calculating
dimensions of disease spot.

Fig. 3. Plant Leaf Diseases Identification System

Architecture REFERENCES

Users acquire images of the leaves from the field using [1] Revathi, P., Hemalatha, M. (2012 “Classification of
sensors and pass it to the computer system which analyze the cotton leaf spot diseases using image processing edge
input images using the homogenous edge detection detection techniques”.169-173.10.1109/
algorithms and diseases wise the pixels call function logic INCOSET.2012.6513900J. Clerk Maxwell, A
used for diseases wise detect the affected parts of the leaves Treatiseon Electricity and Magnetism, 3rd ed., vol.2.
to recognize the diseases and then result (recognition of the Oxford: Clarendon, 1892, pp.68– 73.
diseases and pest recommended) is given as output to the
farmers in three languages.
[2] Arivazhagan, S., Newlin S., Ananthi, S., Vishnu V
(2013) “Detection of unhealthy region of plant leaves
I. RGB image acquisition and classification of plant leaf diseases using texture
features”, Agric Eng. Int CIGR, 15, pp.211-217
II. Create the color transformation structure [3] Gavhale, K., R., and Gawande, U. (2014) “An
III. Convert the color values in RGB to the space overview of the research of plant leaves disease
specified in the color transformation structure. detection using image processing, IOSRJ, Comut.Eng,
pp.10-16.
IV. Apply Color Filtering [4] Singh, V., Misra, A. (2015) “Detection of unhealthy
V. Masking green-pixels region of plant leaves using Image Processing and
Genetic Algorithm in Computer Engineering and
VI. Remove the masked cells inside the boundaries of Applications (ICACEA).
the infected clusters [5] Arti.N.R, Bhavesh, T., Vatsal, S., (2013), “Image
processing Techniques for Detection of Leaf Disease”,
VII. Calling the pixel Ranging function to calculate the
RGB features (each and every disease) International Journal of Advanced Research in
Computer Science and Software Engineering, vol 3,
VIII. Texture Statistics Computation no.11.

© 2024, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM31382 | Page 4