Fungal Blast Disease Detection in Rice Seed using Machine Learning

Abstract

The economy of Pakistan mainly relies upon agriculture alongside other vital industries. Fungal blast is one of the significant plant diseases found in rice crops, leading to reduction of agricultural products and hindrance in the country's economic development. Plant disease detection is an initial step towards improving the yield and quality of agricultural products. Manual Analyzation of plant health is tiresome, time taking and costly. Machine learning offers an alternate inspection method providing benefits of automated inspection, ease of availability, and cost reduction. The visual patterns on the rice plants are processed using the machine learning classifiers such as support vector machine (SVM), logistic regression, decision tree, Naïve Bayes, random forest, linear discriminant analysis (LDA), principal component analysis (PCA), and based on classification results plants are recognized as healthy or unhealthy. For this process, a dataset containing 1000 images of rice seed crop is collected from different fields of Kashmore, and whole analysis of image acquisition, pre-processing, and feature extraction is done on the rice seed only. The dataset is annotated with healthy and unhealthy samples with the help of a plant disease expert. The algorithms used for processing data are evaluated in terms of F1-score and testing accuracy. This paper contains results from traditional classifiers, and alongside these classifiers, transfer learning has been used to compare the results. Finally, a comparative analysis is done between the results of traditional classifiers and deep learning networks.