2004 Statistical modeling of complex backgrounds for foreground object detection教材.pdf

IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 13, NO. 11, NOVEMBER 2004 1459 Statistical Modeling of Complex Backgrounds for Foreground Object Detection Liyuan Li, Member, IEEE, Weimin Huang, Member, IEEE, Irene Yu-Hua Gu, Senior Member, IEEE, and Qi Tian, Senior Member, IEEE Abstract— This paper addresses the problem of background The appearance of background objects often undergoes various modeling for foreground object detection in complex environ- changes over time, e.g., the changes in brightness caused by ments. A Bayesian framework that incorporates spectral, spatial, changing weather conditions or the switching on/off of lights. and temporal features to characterize the background appearance is proposed. Under this framework, the background is represented The background image can be described as consisting of static by the most significant and frequent features, i.e., the principal and dynamic pixels. The static pixels belong to the stationary ob- features , at each pixel. A Bayes decision rule is derived for back- jects, and the dynamic pixels are associated with nonstationary ground and foreground classification based on the statistics of objects. The static background part can be converted to a dy- principal features. Principal feature representation for both the namic one as time advances, e.g., by turning on a computer static and dynamic background pixels is investigated. A novel learning method is proposed to adapt to both gradual and sudden screen. A dynamic background pixel can also turn to a static “once-off” background changes. The convergence of the learning one, such as a pixel in the bu