Detection and Classification of Rice Diseases: An Automated Approach Using Textural Features

Image processing techniques are widely used for the detection and classification of diseases for various plants. The structure of the plant and appearance of the disease on the plant pose a challenge for image processing. This research implements SVM (Support Vector Machine) based image-processing approach to analyze and classify three of the rice crop diseases. The process consists of two phases, i.e. training phase and disease prediction phase. The approach identifies disease on the leaf using trained classifier. The proposed research work optimizes SVM parameters (gamma, nu) for maximum efficiency. The results show that the proposed approach achieved 94.16% accuracy with 5.83% misclassification rate, 91.6% recall rate and 90.9% precision. These findings were compared with image processing techniques discussed in review of literature. The results of comparison conclude that the proposed methodology yields high accuracy percentage as compared to the other techniques. The results obtained can help the development of an effective software solution by incorporating image processing and collaboration features. This may facilitate the farmers and other bodies in effective decision making to efficiently protect the rice crops from substantial damage. While considering the findings of this research, the presented technique may be considered as a potential solution for adding image processing techniques to KM (Knowledge Management) systems.


INTRODUCTION
considered as a main crop in this region, efforts are being carried out at government and private sector level to increase production and hazard prevention [5]. Due to lack of formal education and training, farmers struggle to prevent and cure crops from diseases such as brown spot, bacterial leaf blast, false smut, and stem rot [6]. Rice cover the entire leaf and change its appearance to yellow.
This disease is also spread through air flow and irrigation water [8,11,13].

Image Processing for Rice Leaf Disease Identification
Nowadays, research in the field of plant disease detection is gaining pace [14]. Image processing techniques are used for disease identification and classification.
Moreover, computer vision applications are making their importance in the field of agriculture [16]. Image processing provides tools and techniques for improvement and analysis of images from microscopic to telescopic level [14].
The aim of this research paper is to use image processing techniques for the detection and classification of rice leaf diseases. This section discusses the techniques used in this research work.

Support Vector Machine
SVM is a technique for the analysis and recognition of data. It is used to carry out regression analysis and classification tasks. SVM takes a group of input data and estimates classes for each input [21,[26][27]. SVM is linear model based trainer which creates a hyper plane that groups the data into separation between the hyper plane and the nearest training points based on extracted features [27-28].

Training Phase
The training phase consists of following five steps: (i) Image acquisition included the process of gathering infected rice leaf images for implementing the proposed work flow.
(ii) Image pre-processing was performed to convert leaves images to greyscale using EM GUI functions as shown in Fig. 4 The K-means clustering algorithm was used for this purpose, cluster count was set to 80 for current work and "pp centers" was used as k-mean clustering type. The grouping of features was carried out by reducing the sum of the squares of the distance between the object and the corresponding cluster [31]. Fig. 5 shows the results of K means clustering on an infected rice leaf.
BF matcher was used for matching the features which finds the closest descriptor in the set by trying each one [23]. L1 was used as parameter to reflect SIFT as descriptor ( Fig. 6).
Next, for every image to be learnt, image descriptor was made by making the histogram of clusters for that image, forming numeric descriptor for all the images. Image regions were grouped as BoW along with their vocabulary.
Training Normalized Histograms were provided for classification pr ocess. Classification task was performed using SVM. Classification is considered as a vital part of image analyzed digitally [23]. A classifier is capable of classifying an unknown image but it must be trained before. It has a number of parameters including C, gamma and nu. Nu is related to the ratio of SVM and the ratio of training error, this

Disease Prediction Phase
Used previous phase results for disease detection. Its steps were as follows: (i) Image acquisition took new image for testing purpose, this image was not used to train the system.
(ii) Image pre-processing was performed at first phase to convert leaves images to greyscale.

Image Distribution
Total of 400 diseased rice leaf images were used to validate the stated approach of disease identification. Out of these 400 images, 280 (70%) were used to train the system and 120 (30%) for testing. Testing images were again categorized into two groups, 60 (15%) images were used for parameter optimization in order to get optimal values of SVM parameters, and 60 (15%) images were used for identification of diseases. Table 2 indicates the image distribution.

Parameter Optimization
Parameter optimization was performed on 15% of the testing sample images for two SVM parameters. The concluded optimal values for the parameters are shown in

Evaluation
The proposed approach was evaluated using precision and recall method. Confusion matrix presented in Table 4 indicates the evaluation process. Images are grouped according to the results generated, 60 images were used for testing purpose. True Negative: The image was not infected by a certain disease and the system predicted it as negative.
False Positive: The image lacked a certain disease but the system predicted it positive.
False Negative: The image had a certain disease but the system predicted absence of it.
The summary of the results is given in Table 5, which depicted 113 TN, 51 TP, 9 FN and 7 FP cases.
The proposed methodology was evaluated based on features estimated in Table 6