Geometric constraint-based image segmentation algorithm for concave point detection in complex backgrounds

In workpiece assembly scenarios, there is a common phenomenon of sticking between workpieces, which seriously interferes with the subsequent counting and sorting of objects. Therefore, effective segmentation of these sticking workpieces is particularly critical. To address the difficult problem of segmenting highly attached objects in complex backgrounds, a concave point segmentation algorithm based on geometric constraints has been proposed. The method first utilizes angular thresholding to detect concave points in preprocessed images. Then, pseudo-concave points generated by uneven illumination are eliminated based on the number of concave points and their connected contours within a fixed range centered on the concave points. Subsequently, the distance between the concave point and the angular bisector where the remaining concave points are located is utilized to achieve the matching of concave pairs, and the straight line segmentation is performed accordingly. The results show that compared with the traditional watershed algorithm and morphology algorithm, the proposed algorithm has a significant segmentation effect in the case of uniform illumination, with an accuracy of 96.8%. At the same time, the proposed algorithm also presents a certain segmentation effect in the case of insufficient light and has a certain degree of anti-interference. This study provides a novel and efficient solution to the problem of image segmentation in complex backgrounds.