Knowledge-based Fuzzy Reasoning for Maintenance of Moderate-to-Fast Background Changes in Video Surveillance

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


Background maintenance, Fuzzy reasoning, Change detection, Video surveillance.


Background maintenance is an essential part in a video surveillance system. Background maintenance requires timely updating to gradual illumination changes as well as to moderate-to-suddenchanges in the background. It is also essential that slow moving foreground objects would not be absorbed during the background maintenance. While the gradual background changes can be absorbed by pixel level maintenance [1], this paper introduces a novel method for adaptive maintenance of moderate-to-fast background changes. Using the 2nd-order moment-based Mahalanobis distance to prior specified foreground object categories, and employing motion edge and shape features, regions containing moderate-to-fast changes are first assigned as fuzzy members of foreground and background objects. Up dating background changes is then followed at the rate according to the membership grade of each detected re gion. Experimental results to indoor real image sequences showed that the proposed background maintenance method has well updated to these background changes without ab sorbing foreground objects. The proposed method has been tested for real-time processing using a camera operated at 8 frames/second.

Important Links:

Go Back