Accurate and Robust Circular Object Detection Using Conditional Probability Searching

Circular object detection is very important in image processing. In this paper, accurate and robust circular object detection using probability searching is presented. The main contributions are threefold. We first redefine the gradient line with direction, which is robust against noise. Then we randomly select two pixels to determine a candidate center and radius by the intersection of gradient lines in a connected region instead of the whole image to improve time speed. After the candidate circle is determined, we search the other points using the conditional probability, and we revise accurate center and radius quickly in the searching process. Three tests demonstrate that the proposed method outperforms single-circle and multi-circle detection methods in the robust, accuracy and real-time.