Measuring human actions using sequential boundary contour pairs

摘要

In this paper, we present a method for measuring human actions using sequential boundary contour pairs, extracted from images obtained using a binocular sensor system with gaze control. Continually the differences between the current and previous pairs of contour images are used to reconstruct a human body model, which is represented by a set of cue spheres (CSs). For the reconstruction, first 2-D features called cue circles (CCs) are calculated for each of the contour differences. Then stereo matching is carried out by finding pairs of CCs in the pair of contour images under consideration: a CS is projected on the two image planes as its corresponding CCs. The time variation of the relative spatial arrangements of the recovered CSs can be used for the analysis and understanding of human motions by machines.