Plant Disease Diagnosis with Novel Segmentation and Multiple Feature Selection Based on Machine Learning

Smart agriculture has expanded as a result of artificial intelligence, benefitting our country’s economy. Automated plant disease detection systems are attracting a lot of attention in the recent years. With the help of efficient plant disease detection systems, the onset of the disease can be detected thereby reducing the spread and damage caused to the crops. The paper’s major contribution is to develop and construct a plant disease detection system that includes a unique segmentation, multi-feature selection and feature extraction approach. An efficient segmentation process is first developed using the statistical measures to extract the diseased area and a feature selection process to extract the corresponding features. In this research a combination of features are used for the feature selection process which makes it efficient to improve the classification accuracy. For classification machine learning based classifier is used. The performance of the proposed solution was tested on banana dataset using metrics such as detection accuracy and classification accuracy. From the results it was observed that the proposed solution achieves around 95.4% overall classification accuracy proving it to be a feasible solution for practical plant disease detection application.