Abstract:

By using digital image processing technology, the disease image of Hemerocallis citrina leaf was taken as the recognition object. The disease area was segmented based on Lab space and K-means clustering algorithm, and the color characteristics, histogram of oriented gradient (HOG) and shape characteristics of the target areas were extracted from the images. The single-feature model and multi-feature model were established respectively based on the extracted features. Particle swarm optimization (PSO) algorithm was used to optimize the penalty factor and kernel parameter of the support vector machine (SVM) model through cross validation. Multi-feature classification model based on PSO-SVM was established to identify diseases of H. citrina leaves. The recognition rate of SVM based multi-feature classification model was higher than that of single-feature classification model, and the recognition rate could reach 81.67%. The recognition rate of multi-feature classification model based on PSO-SVM was as high as 92.39%. The multi-feature classification model based on PSO-SVM has high recognition rate and can identify the disease of H. citrina leaf timely, conveniently and efficiently.

Key words: image processing, Hemerocallis citrina, disease recognition, support vector machine (SVM), particle swarm optimization (PSO), multi-feature fusion