Abstract: Rice particle adhesion in rice detection process brings great difficulties to subsequent rice detection and classification. In this paper, a method of segmentation of adhesion rice based on concave point matching is presented. Firstly, the original image is preprocessed based on the feature of Rice image combined with the thought of morphology. Second, the original image is preprocessed by using chain code tracking to calculate the curvature of the image boundary after segmentation, and then search the concave point according to the curvature result and chain code difference. At last, the concave point is matched, and the concave point is connected with the dividing line. The concave point matching segmentation is realized. The experimental results show that the method has better segmentation effect than traditional segmentation method, and can effectively restrain the segmentation phenomenon.

Key words: concave point matching, chain code, rice detection, image segmentation