A Machine Vision Method for Automatic Circular Parts Detection Based on Optimization Algorithm

Circle Hough transform (CHT) is the most commonly used method to inspect circular shapes for its advantage in strong robustness. However, it requires large amounts of storage and computing power, which cannot meet the requirements of real-time processing. To overcome this deficiency, this paper presents a novel circle detection method based on adaptive artificial fish swarm algorithm (AAFSA) by determining the circle center and the radius of circular parts. A new fitness function had been developed to evaluate the similarity of a candidate circle with a real circle. Based on the fitness values, a batch of encoded candidate circles is modified through the AAFSA in order that they can match with the actual circles on the edge map. Experiments results show our proposed method can accurately detect circular parts by search the optimum values in the parameter space. Compared to other popular approaches, i.e., the CHT, the least square method (LS) and the random sample consensus (RANSAC), the proposed method achieved a remarkable improvement in both accuracy and speed of circular parts detection.