Pedestrian detection and matching across non-overlapping camera views

摘要

For non-overlapping views monitoring system, this paper proposed a pedestrian detection and matching approach based on the color spatial distribution model and texture pattern reconstruction model. Firstly, the target was detected and segmented based on background subtraction and color clustering;Secondly, the color spatial distribution model and texture pattern reconstruction model were established;Finally,by using spatial distribution color feature and reconstructed texture pattern, the pedestrian was matched respectively. Experimental results showed that the presented approach can effectively solve the target detection problem of incomplete segmentation. It was capable of coping with the mismatch problem due to the difference of color distribution and the reconstructed texture pattern. Besides, the proposed approach was robust to the diversity of the visual angle and the change of the object gesture.