Abstract: In order to solve the problem that it is difficult for the users to identify the disease accurately in the domestic hydroponic cucumber, a cucumber leaf spot recognition system based on image processing was designed. Adaptive wavelet was applied to the original image noise reduction processing. The ideal cucumber leaf segmentation image was obtained by threshold segmentation combined with morphological operation in HSV space. The disease spots of cucumber leaves were obtained by adaptive threshold segmentation, and the morphology, color and texture original characteristic parameters were extracted from spots. The sensitivity of original characteristic parameters was defined based on GA-BP network, and the optimal combination of features was realized by recursive elimination of some features with low sensitivity. According to the optimized combination of characteristic parameters, the support vector machine was used to identify the anthrax and powdery mildew of cucumber. The experimental results showed that the method effectively improved the recognition rate of cucumber diseases and provided a reference for other crop diseases intelligent identification.

Key words: leaf spot of cucumber, GA-BP neural network, sensitivity, feature vector optimization combination, support vector machine, disease spot recognition