A Spherical Codebook in YUV Color Space for Moving Object Detection

Codebook background model is widely applied to moving object detection in surveillance video. The traditional codebook background model realizes moving object detection via measuring the brightness range of RGB color space but it can not measure color distance accurately. In addition, for inherent disadvantages of ROB color space, detected objects are provided with noise or detection results are incomplete. This paper proposes a moving object detection algorithm with sphere codeword model, which replaces RGB Color space with YUV Color space to reduce sensibility of illumination in codebook background. Meanwhile, spherical codeword in YUV color space is adopted to improve cylinder codeword model in RGB Color space. Experiments and comparative analysis with known algorithms for moving object detection show that the proposed algorithms in this paper improve correctness of moving object detection with good performance in timelessness and robustness.