Identification and Classification of Mango Leaf Disease Using Wavelet Transform based Segmentation and Wavelet Neural Network Model
Satyasis Mishra1, Ellappan V.2, Sunita Satapathy3*, Gemechu Dengia4, BayisaTaye Mulatu5,Ferew Tadele6
1,2,4,5,6Dept. of ECE, SoEEC, Adama Science and Technology University, Adama, Ethiopia
3 Dept.of Zoology, SoAS, Centurion University of Technology & Management, Odisha, India Email:3[email protected]
Automatic detection of plant diseases is very ample beneficial as it reduces the workload of farmers. To improve the quality and quantity of crop yield, identification of plant diseases is important in agricultural field.
Leaves are considered as the food source for plants and the early and accurate recognition of leaf diseases is very much essential. This research work presents an wavelet transform image segmentation technique for automatic detection and a WNN(Wavelet neural network) based approach presented that classifies leaf diseases in Mango plant species. It is proposed to identify and detect the disease from the mango leaf by taking high resolution images. The plant Village dataset which is consisting of 1130 images of diseased and healthy mango leaves is considered for segmentation and classification. The proposed WNN model achieves an accuracy of 98.93% for identifying the leaf diseases in mango leaves thereby showing the feasibility of its usage in real time applications.
Wavelet Transform, Radial Basis function Neural Network, Mango leaf diseases, wavelet neural network
The king fruit “Mango” acquires a great importance in business in agriculture as agriculture is the mother of all cultures. The plant leaf diseases destroying the crops and the farmers are facing losses in crops. Deprived of knowing regarding the diseases affected in the plant, excessive pesticides are applied by farmers for plant disease treatment. So, proper care is required for the advance detection of mango plant leaf disease. Image segmentation, utilized for mango plant leaf disease detection. It is the process of splitting or grouping an image into different parts of the affected areas.. The segmentation process is intelligent, so, we propose wavelet based algorithm for color image segmentation.
The orthodox means of diagnosis of plant disease needs larger amount of pesticide which is time-consuming and challenging, for farmers .With the introduction of “Computer Vision (CV), Machine Learning (ML), and Artificial Intelligence (AI)” technologies, “Deep Learning (DL)” has dominated the detection process in agriculture [2,3]. R. Nikam et al.  presented a methodology to determine the severity level of mango disease from leaf images. In their work they have used a model dataset of 150 color images and applied disease region segmentation, Sobel and Laplacian filters to detect the severity level. L. Dutta in . Authors have utilized the different morphological & statistical features for feature extraction and Artificial Neural Network (ANN) for classification. Savita N. Ghaiwat et al., reviewed “ ANN, SVM, PNN, SELF ORG MAPS and fuzzy logic” for classification of plant diseases, Sanjay B. et al., utilized Vision-based identification algorithm with masking for classification process.Mrunalini R. et al.,
proposed “K-means clustering”algorithm to classify the crop diseases. S. Arivazhagan et al.,
proposed (2013) classification by utilizing SVM. Kulkarni et al., proposed ANN classifier, Sabah Bashir et al., proposed (2012) Texture segmentation and K-means clustering
technique to classify various plant diseases. Naikwadi et al., (2013) proposed detection with color extraction. Sanjay B. Patil et al., utilized thresholding segmentation methods.
PiyushChaudhary et al., “Color transform” based approach for calculating dimensions of disease spot. Arti N. Rathod et al., Surveyed different techniques for leaf disease detection.
Therefore, Barbedo in  identified challenges in the classification of plant diseases. Kaur et al.
in  proposed computer vision methods for the detection and classification of the plant leaves.
To improve the drawback of the automatic classification of plant disease, the wavelet neural network is proposed. The proposed wavelet neural network model employs different layers to smooth images in order to improve the noise-immunity. Therefore, the proposed wavelet neural network is more robust than these algorithms for images corrupted by different types of noise and the proposed model is suitable for good classification results.
The rest of the paper is organized as follows: section-2 presents material and methods which includes research block diagram, proposed WNN model, section -3 presents the graphical results of segmentation and classification, section-4 presents segmentation and classification detailed discussion, section -5 presents the conclusion followed by related references.
2. MATERIAL AND METHOD
2.1 Research flow diagram
The research flow diagram indicates the step by step accomplishment of the research work.
Further the block diagram shows the flow of algorithm application for detection and classification of plant disease and presented in Fig.1.
2.1.1 General Objective:
The objective of the research work to identify of mango plant leaf disease using wavelet transform  based segmentation and diseases classification using wavelet neural network model.
2.1.2 Specific Objectives:
Objective to localize the disease infected areas using wavelet transform segmentation technique. It will help the farmers at the rural belt to know in advance the effect of disease.
Objective to develop a new wavelet based neural network model for classification of mango leaf disease.
2.2 Proposed Wavelet Neural Network Model
The wavelet neural network and presented in Fig.2 are a new class of networks which have been used for classification of mango diseases. In this research the wavelet function is modified and a new wavelet function has been utilized in hidden node to improve the performance of the neural network.
Fig.1 Work flow diagram of research
MLP: Multilayered neural network
In this model the data points x1,x2,...xn
are inputs (features) and 1,2...n
are the wavelet activation function in the hidden units. Where the wavelet function proposed as
n N iN i
n w w x w x x
(1) The objective function is to minimize the error and the mean square error is given by
N d e MSE
(2) Where “d” is the desired vector.
Radial basis functional neural network, MLP
Input Image(Leaf sample)
Feed Forward Neural Network
Statistical Feature Extraction
Wavelet Neural Network
Wavelet algorithm for Image Segmentation
Classify Diseased plants
Fig.2 wavelet neural network model
In the proposed work, the Mango leaves dataset  repository having leaves of multiple plants are considered for segmentation and classification. A total of 1130 images were taken from the plant Village dataset. Out of these 80% of data has been considered for training and remaining 20% are taken for testing.
2.4 The mango tree diseases
Alternariatenuissima (Kuntze: Pers) Wiltshire,the disease appears in the small, circular and brownish spots, which enlarge and become irregular to form big Alternariatenuissima (Kuntze:
Pers) Wiltshire. The disease appears in the small, circular and brownish spots, which enlarge and become irregular to form big water-soaked patches. Reddish patches develop on the flesh below the spotted area of fruit.MangoMalformation,Anthracnose is also known as blossom blight, leaf spot, fruit rot and twig blight. This disease is severe both in field and storage. The disease is present all mango area of India The verities neelam and bangalora are highly susceptible to this disease.
Fig.3Image samples (a) Anthracnose (b) Golmich (c) Red Rust
The literature survey shows different classification techniques, segmentation process for leaf disease detection and classification.
2.5 Feature extraction
The input leaf images will undergo the process of gray image conversion, for computation of disease location detection, and segmentation. A total of seven statistical features are extracted for purpose of classification.
Table -1 Normalized Feature Extraction Table Feature values
Correlation 0.6469 Entropy 0.9018 Skewness 0.5785 Mean 0.3425 Kurtosis 0.8422 Energy 0.8335
3.1 Segmentation results
Fig.4(a) Image-1Wavelet transform segmentation
Fig.4(a) Image-2Wavelet transform segmentation
3.2 Classification Results
Fig.5 Classification comparison results
Table 2 Classification Accuracy of the models
Model No. of
Computatio nal time
RBFNN 1130 27.5487 93.28
FFNN 1200 21. 2548 95.85
MLP 1200 18.2547 97.89
WNN 1200 12.6587 98.93
Table 3 Segmentation Accuracy and Performance Evaluation Algorithm Accuracy in% Computational time
FCM 96.31 21.54
WT 98.87 11.27
WT: Wavelet Transform FCM: Fuzzy c means
Fig. 4(a) and Fig.4(b) presents the localization of infected areas of leaf diseases. The wavelet transform segmentation shows prominent identification of diseases. After segmentation the images are under gone for statistical feature extraction and the feature values are presented in Table-1.The segmentation accuracy has been shown in Table-3. The extracted features are fed as input to the proposed wavelet neural network (WNN)model for classification. The features are grouped into batches and applied for classification. It is found from the Fig.5 that, the wavelet neural network outperforms than the Radial basis function neural network , multi layered neural network and feed forward neural network. RBFNN,MLP took nearly 850 and 900 iterations to converge. The proposed WNN took nearly 400 iterations for convergence which is lowest in comparison to other models which can be observer visually from Fig.5.
This research work proposes a novel wavelet neural network by introducing a new wavelet function in the hidden nodes. The WNN is preferred for classification due to faster convergence speed and needs lesser number of hidden nodes in its structure. The leaf images are segmented by wavelet transform and applied for feature extraction. There are seven statistical features are obtained from 1030 images. Further, the features are given as input to the proposed WNN model for classification. It is found the accuracy of classification is 98.93 % which is higher than the other models presented in Table-2. The computational time taken by proposed model is 12.6587 seconds whereas the other models took higher computational time for convergence. The proposed WNN model can also be applied for other medical imaging data bases in future.
 J. Ma, K. Du, F. Zheng, L. Zhang, Z. Gong, and Z. Sun, ``A recognition method for cucumber diseases using leaf symptom images based on deep convolutional neural network,'' Comput. Electron. Agricult., vol. 154, pp. 18_24, Nov. 2018. doi:
 K. P. Ferentinos, ``Deep learning models for plant disease detection and diagnosis,'' Comput. Electron. Agricult., vol. 145, pp. 311_318, Feb. 2018. doi:
 E. C. Too, L. Yujian, S. Njuki, and L. Yingchun, ``A comparative study of _ne-tuning deep learning models for plant disease identi_cation,'' Comput. Electron. Agricult., to be published. doi: 10.1016/j.compag.2018.03.032.
 R.Nikam, M. Sadavarte, “Application of Image Processing Technique in Mango Leaves Disease Severity Measurement”, National Conference on Emerging Trands in Computer, electrical and Electronics (ETCEE-2015), International Journal of Advance Engineering and Research Development (IJAERD), 2015.
 K. Muthukannan, P. Latha, P. Nisha, R. PonSelvi, “An Assessment on Detection of Plant Leaf Diseases and Its severity using image segmentation”, International Journal of Computer Science and Information Technology Research (IJCSITR), January-March 2015.
 GhaiwatSavita N, AroraParul. Detection and classification of plant leaf diseases using image processing techniques: a review. Int J Recent AdvEngTechnol 2014;2(3):2347–
812. ISSN (Online).
 Dhaygude Sanjay B, KumbharNitin P. Agricultural plant leaf disease detection using image processing. Int J Adv Res Electr Electron InstrumEng 2013;2(1).
 Mrunalini R Badnakhe, DeshmukhPrashant R. An application of K-means clustering and artificial intelligence in pattern recognition for crop diseases. IntConfAdvInfTechnol 2011;20. 2011 IPCSIT.
 Arivazhagan S, NewlinShebiah R, Ananthi S, Vishnu Varthini S. Detection of unhealthy region of plant leaves and classification of plant leaf diseases using texture features.
AgricEngInt CIGR 2013;15(1):211–7.
 Kulkarni Anand H, AshwinPatil RK. Applying image processing technique to detect plant diseases. Int J Mod Eng Res 2012;2(5):3661–4.
 Bashir Sabah, Sharma Navdeep. Remote area plant disease detection using image processing. IOSR J Electron CommunEng 2012;2(6):31–4. ISSN: 2278-2834.
 NaikwadiSmita, AmodaNiket. Advances in image processing for detection of plant diseases. Int J ApplInnovEng Manage 2013;2(11).
 Patil Sanjay B et al. Leaf disease severity measurement using image processing. Int J EngTechnol 2011;3(5):297–301.
 ChaudharyPiyush et al. Color transform based approach for disease spot detection on plant leaf. IntComputSciTelecommun 2012;3(6).
 RathodArti N, TanawalBhavesh, Shah Vatsal. Image processing techniques for detection of leaf disease. Int J Adv Res ComputSciSoftwEng 2013;3(11).
 J. G. A. Barbedo, ``A review on the main challenges in automatic plant disease identification based on visible range images,'' Biosyst. Eng., vol. 144, pp. 52-60, Apr.
2016. doi: 10.1016/j.biosystemseng.2016.01.017.
 S. Kaur, S. Pandey, and S. Goel, ``Plants disease identification and classification through leaf images: A survey,'' Arch. Comput. Methods Eng.,vol. 26, no. 2, pp. 507530, 2019.
 Yuehui Chen, Jiwen Dong , Bo Yang and Yong Zhang, A Local Linear Wavelet Neural Network, Proceedings of the 5th World Congress on Intelligent Control and Automation, June 15-19, 2004, Hangzhou, P.R. China
 Gu X., Du JX., Wang XF. (2005) Leaf Recognition Based on the Combination of avelet Transform and Gaussian Interpolation. In: Huang DS., Zhang XP., Huang GB. (eds) Advances in Intelligent Computing. ICIC 2005. Lecture Notes in Computer Science, vol 3644. Springer, Berlin, Heidelberg. https://doi.org/10.1007/11538059_27
 D. P. Hughes and M. Salathe. (2015). ``An open access repository of images on plant health to enable the development of mobile disease diagnostics.'' [Online]. Available:
 A. Patra, S. Das, S. N. Mishra, M. R. Senapati , “An adaptive local linear optimized radial basis functional neural network model for financial time series prediction” Neural Computing and Applications, vol.28, Issue 1, January 2017 ,Pages 101-110, doi:10.1007/s00521-015-2039-0