A foreground detection system for automatic surveillance

computer vision;background subtraction;sparse representation;linear decoding;automatic surveillance

Dikmen, Mert ; Huang, Thomas S. ; Huang ; Thomas S.

关键词: computer vision;    background subtraction;    sparse representation;    linear decoding;    automatic surveillance;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/14760/Dikmen_Mert.pdf?sequence=2&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Automated surveillance has long been an application goal of computer vision.An integral part of such surveillance systems is concerned with accuratelysegmenting foreground objects from the static background in the videos. Inthis thesis we introduce a novel system for background subtraction, whichtakes a di erent approach than the conventional background subtraction systems.We make the assumption that the video background is stationary andthe foreground objects take up only a small portion of the entire frame at anygiven time. This speci c assumption allows us to formulate the foregroundsignal as a sparse additive error introduced to otherwise clean background signal.We outline the algorithm for performing background subtraction usinglinear programming, and demonstrate accurate segmentations of foregroundobjects under realistic surveillance scenarios. The proposed method is on parwith the state of the art approaches for accurately segmenting the foregroundunder challenging conditions. Furthermore we propose several methods forbuilding a set of bases to represent the background and provide empiricaljusti cation of their e ectiveness.