Mark point recognition in motion tracking of a floating life jacket-man system
Fei Duan · 2006
A fast,high-precision,robust algorithm was developed for round object recognition based on the curvature scale space technique.The algorithm first extracts critical points based on the curvature from the edge map to divide the object contour into several segments.An acceptable circular arc is defined using statistical information of each edge segment.The parameters describing the round object are deduced using clustering and LS circle fitting.The algorithm has been evaluated with a number of images against the Hough transform(HT).The results show that the method offers better precision than the HT when the lengths of the locally consecutive arcs are long enough and offers a far more effective solution with images containing widely different size objects.The algorithm has been used to track the motion of a floating man in a natatorium with complex backgrounds,variable lighting conditions and occlusion of marking points with very good results.