Differential evolution-based parameter optimization for circle detection by Hough transform

We have researched image processing to inspect the presence of chips remaining inside a hole drilled by a drill. In our developed system, the center position and radius of the hole are identified by circle detection based on the Hough transform as a pre-processing step after capturing an image of the hole using a camera. Based on the detection result, the image is transformed into polar coordinates. Then, an inspection algorithm called KIZKI is applied to detect chips. An optimal set of parameters for the Hough transform should be specified for accurate circle detection. However, it is labor-intensive to search for them manually. Therefore, differential evolution that is one of the evolutionary computation algorithms is used to automatically optimize the set of parameters in this study. By using a custom-designed fitness function, the optimal set of parameters for accurate circle detection was successfully obtained.