An Efficient Machine Learning Approach for Apple Leaf Disease Detection

Abstract

Apples are one of the most popular agricultural products. Despite being one of the most widely grown commodities, apple demand is on the rise. As a result, this crop, which was formerly only grown in temperate climates, is now being grown in tropical climates. Pest and disease infestations are a major issue that affects apple output each year. In this paper, an approach has been made which combines machine learning and image processing concepts to identify diseases from infected apple leaves. This method effectively differentiates between diseased and non-diseased apple leaves. Pre-processing of the image is done using grab cut segmentation which is the primary stage in the disease identification process. The infected type from the original leaf image is recognized by 96% using the segmentation of the diseased portion, and multiclass SVM detects the infected type from 500 images using the feature extraction.