Moving human body tracking based on small region matching

Illumination and moving objects vary dynamically in complex situations, making the tracking of moving human targets difficult. To solve this problem, a new tracking procedure is proposed. Multiple small regions of human targets are extracted through background subtraction, and their matching reliabilities are evaluated by fuzzy inference. The human target is tracked by means of tracking the small regions with high matching reliability and of neighborhood searching, position rectifying and estimating the small regions with low matching reliability. Experimental results showed that the suggested approach can solve occlusion and disturbance by other objects and is adaptive to the illumination and other environmental changes.