Pattern Analytics of Healthy and Diseased Leaves Recognition using Genetic Algorithm

Farming techniques have been reorganized with the involvement of Artificial Intelligence. Smart and Precision agriculture has improved the yield and provided healthy crops as compared to the traditional measures involving chemicals for removal of weed. One of the steps in maintaining quality of yield is to segregate healthy crop from diseased one. We propose utilization of genetic algorithm for deploying an AI based four-step method for such segregation. Images of crop can be utilized for this purpose by employing pre-processing, followed by segmentation, extracting the useful feature and finally classification, which in this case is based on Genetic Algorithm. Groundnut, pepper, tomato and potato leaves have been utilized during the work by employing the proposed technique at two distinct datasets involving various stages of diseased. Images of size 256 × 256 have been utilized. Median Filter was deployed during pre-processing, k-means for segmentation, while Kaze and HOG (Histogram of Oriented Gradient) were chosen for extracting selected features. Classification was carried out using genetic algorithm. Results have been compared to random-forest, adaboost and naive-bayes, whereby our results show that proposed approach outperforms other methods by providing an accuracy of 81% on both datasets.