Efficient circular object detection with hypothesis filtering strategy and Hough transform

A Hough-like technique for circular object detection is reported. The proposed technique is simple in implementation, efficient in computation and robust to noise. In general, to evaluate circle parameters for all possible point triplets in an edge image containing n points, _nC3 enumerations of the points have to be examined. However, if specific relations of the circle points are sought, the required number of enumerations can be reduced. In this paper, we propose one such technique with point triplets possessing right angle property and the required enumerations can be reduced to _nC2. In addition, a novel hypothesis processing strategy known as hypothesis filtering is introduced. This includes two novel constraints: consistency checking with gradient angles and neighbouring points validation, which are used to filter non-circular feature point sets possessing right angle property. It is found that the use of hypothesis filtering as a preprocessing step improves significantly the speed of detection. Experimental results demonstrate the effectiveness of the method to detect circles in both synthetic drawings and real images