Description: Moving object segmentation based on background modeling is an important task of intelligent video surveillance, quality models directly affects the detection, tracking, recognition accuracy motion analysis. The current modeling methods are mostly single, ignoring the characteristics of pixels on the temporal and spatial linkages, model description is not accurate enough, the background disturbances, global illumination changes and complex indoor and outdoor scenes and other conditions are not robust strong, leading to a hollow-point division and noise appear. To solve these problems presents a double-modeling method, the pixel brightness feature extraction using time domain modeling codebook in the first layer, the second layer of texture feature extraction using neighborhood center symmetric local binary pattern based construction mode. Experiments show that the method used for motion segmentation when compared with the conventional method has better accuracy and robustness.
To Search:
File list (Check if you may need any files):
融合码本和纹理的双层视频背景建模方法_李峰.pdf
Download