A real-time and robust multi-circle detection method based on randomized Hough transform

Abstract

A real-time and robust multi-circle detection method based on randomized Hough transform (RHT) was proposed in this paper. In the method, circle detection is done by the RHT method in each connected component like [9]. We change the points sampling strategy of RHT to increase the hitting probability. To enhance the robustness, we do circle detection using the RHT method at local areas which possibly contain circles at the last step of our method. Performance analysis of the original RHT as well as our method is given, and a way to estimate sample threshold under certain detection confidence is introduced. Experiment results demonstrate that our method is much faster and more robust than the standard RHT method and can meet the requirement of real-time circle detection application.