Neuro-kNN classification system for detecting fungal disease found on vegetable crops using local binary patterns

Abstract

This paper describes the behavior of classifiers for identification and classification of fungal disease symptoms found on vegetable crops. Symptoms of fungal disease, namely, anthracnose, powdery mildew, rust, downey mildew, early blight, and late blight found on specific type of vegetable crop are considered for recognition and classification. The way the disease analysis is done considering both sides (front and back portions) of the leaves has been addressed. The analysis of the fungal disease present on the leaves of vegetable crops is detected in the early stage before it damages the whole leaf and subsequently the plant. The Local Binary Patterns(LBP) extracted from disease affected leaves are used as input to the classifiers. An integrated classification system “Neuro-kNN” has been proposed, of which multilayer BPNN classifer is used for training purpose and k-Nearest Neighbor(k-NN) classifier for testing purpose. The recognition accuracy is observed using Artificial Neural network (ANN) and Neuro-kNN classifier methods. The average classification accuracy is found to be 84.11% for the test samples using ANN. The average classification accuracy has increased to 91.54% using Neuro-kNN classifier. The work finds application in automatic recognition fungal disease found on vegetable crops by the service robots in the real world.