A Roundness Detection Algorithm of Non-Standard Discontinuous Circle Based on the Curvature Change of Discrete Edge Points

Abstract

In the practical application scene of image detection, it is of great significance to detect the circle in the image and judge the roundness of the circle. Because the edges of real objects are often discontinuous due to occlusion, roundness detection of discontinuous circles is a common task. Traditional circle detection algorithms mainly include voting/clustering algorithm based on edge points, optimization algorithm and arc segment-based algorithm. For discontinuous circles, because the integrity of the circle edges is different, the voting results of edge points are difficult to be used as the basis for roundness detection. When the algorithms based on optimization and arc segment detect nonstandard circles, a large number of overlapping misjudgments are easy to occur, and the overlapping of objects before and after will lead to maximum suppression, which is difficult to achieve good results. In this paper, the geometric properties of the circle itself are used to classify the edge points, so that the roundness of the discontinuous circle can be calculated and the size of the circle can be judged. In this paper, taking the photos of mutually blocked round fruits as an example, an algorithm is constructed to detect the roundness of non-standard discontinuous round objects. The results show that this algorithm has a good effect on roundness detection of non-standard discontinuous circles.