An algorithm of moving object detection based on texture and color model
Zhenhong Shang, Zhenping Qiang, Hui Liu · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2013
In this paper, we propose an algorithm for Moving Object Detecting which can remove influence of shadow and illumination change. The algorithm is based on background subtraction using color and texture information, we establish a texture model based on LBP (local binary pattern) for each pixel, and adopt a newly developed photometric invariant color measurement to description color information, Use a similarly pixel-based models update algorithm that proposed by Stauffer et al, but the difference is that we use a novel ‘hysteresis’ scheme for update of the weight. We use two layer process in foreground detecting, at the pixel layer, through the texture and color model we mentioned above to divide the each pixel to background or foreground, at the another layer, calculate the LBP texture information for the foreground regions boundaries which come out by color model subtraction, through comparing them to texture information come out by texture model for the foreground regions boundaries to remove fault detect of foreground.