Grape leaf image classification based on machine learning technique for accurate leaf disease detection

Grape leaf diseases have a major impact on the growth of grape industry and grape crop yield. Thus, there is a need of a grape disease detection in early stages of disease so that disease spread and their impact could be controlled and development and production of grape industry remain continuous and active. However, the detection of grape leaf disease in initial stages is highly critical and challenging. Therefore, in this article, machine learning technique is adopted for the early detection of grape leaf disease and accurately distinguish between various classes of disease. Furthermore, Convolutional Neural Network based Classification (CNNC) model and improvised K- Nearest Neighbor (IKNN) model are introduced for classification of leaf diseases. High quality histogram and extended histogram features are obtained to provide structural, pattern, boundary and discriminative information. Then, classification process is performed on the obtained high quality gradient based features. Classification accuracy is improved to a great extent using proposed CNNC and IKNN model. The accuracy of the proposed CNNC and IKKN model is tested with the help of public dataset named as Plant-Village Dataset. The performance of proposed CNNC and IKKN model is compared with various traditional classification models considering classification accuracy.