Robust Change Detection and Segmentation for Background Maintenance

L. Li (Singapore), I. Y.-H. Gu (Sweden), M.K.H. Leung and Q. Tian (Singapore)


Change detection, Change segmentation, Surveillance, Background extraction, Background maintenance.


Background maintenance in surveillance systems requires the adaption to both gradual illumination changes and fast background changes. Change detection and segmentation are key steps towards a successful background mainte nance. This paper introduces a robust change detection method by collecting change evidence from images in tem poral and spatial domain, and from edges and reference background images. The proposed change segmentation method exploits peaks in a vertical projected 1D curve us ing fuzzy reasoning, and then associates these peaks to the segmented 2D change regions. The advantages of such a change detection and segmentation method are that it al lows the subsequent background maintenance to use an adaptive rate to update regions in a reference background according to the types of change. Experiments on change detection and segmentation to image sequences from both indoor and outdoor environments have shown the robust ness of the method. Some results of change segmentation and adaptive background maintenance are also included.

Important Links:

Go Back