Golf ball detection from multi-exposure images under heavy occlusion

In this paper, we propose an automatic technique for detecting golf balls from multi-exposure images under heavy occlusion. The proposed technique consists of two processes. First, an automatic region-of-interest detection process based on image thresholding and labelling techniques is employed to separate regions of interest of golf ball from each multi-exposure image. Then, a novel adaptive ball detection and selection process based on circular Hough transform and circular shape correlation is applied to each region of interest for determining the location and radius of each golf ball with high-accuracy irrespective of occlusion. Experimental results show that the proposed technique is capable of automatic golf ball detection under heavy occlusion.