MMFS

Abstract

At present, pest and disease recognition methods based on machine vision generally have the problems of complicated image pre-processing steps and high difficulty in image processing technology, resulting in the technology is not practical in production. In this paper, the main disease pictures of grape leaves are taken as the research object, and we propose a grape disease recognition method based on multi-feature fusion and SVM (MMFS). First extract the four features of color, HSV, texture, and Histogram of Oriented Gradients(HOG) for each leaf image, then reduce the dimension by principal component analysis(PCA), and use multi-feature fusion method based on feature merging to form feature vectors. Finally, they are put into the SVM classifier for training, and the trained classifier is used for the automatic recognition of lesions in the entire image. The experimental results show that the accuracy of disease recognition based on the new features formed after fusion has reached 92.5%, which verifies the effectiveness of the classifier based on MMFS in improving the accuracy of grape leaf disease recognition, and it has positive significance for real-time monitoring of grape growth, providing timely pest prevention and early warning, and reducing labor cost.