A Fast Randomized Detection Algorithm for Multi-Ellipse Detection Problem

In this paper, we propose a fast randomized detection algorithm for addressing the multi-ellipse detection problem. The algorithm operates by randomly sampling one edge point from the image and locally searching for two additional edge points, using neighborhood information from all three points to generate candidate ellipses. The implementation involves transforming these candidate ellipses into corresponding circles through mathematical mapping, where true ellipses are verified by confirming their circular counterparts. A key optimization in the candidate ellipse determination phase minimizes the number of random sampling points while aggressively filtering out non-elliptical points, thereby reducing无效 sampling and computational waste. The core algorithm can be implemented using edge detection functions like Canny followed by randomized point sampling and geometric transformation routines. Numerical experimental results confirm the algorithm's excellent robustness and demonstrate significantly faster detection speeds compared to existing similar algorithms.