Moving Object Detection under Various Illumination Conditions for PTZ Cameras

Video surveillance has become an increasing research field now a day. The fundamental step in video surveillance is the Moving object detection. Most of the works focused on background modeling in PTZ camera but still lacking under different positions and various illumination conditions. While the camera is on pan and sudden zoom, the pixel intensity of each position may vary and it cannot adapt the motions when the target is faraway or closer. This issues cause major problem in Background Modeling (BM). Objectives: To solve this problem a texture based method adapted to handle grey-scale variation, rotation variation and various illumination conditions of the moving objects. Methodology/Analysis: Modified version of LBP, that combines the advantages of LBP and SIFT descriptor known as eXtended Centre Symmetric Local Binary Patterns XCS-LBP. Finally GMM (Gaussian Mixture Model) is used for segmenting the foreground Extraction by the XCS-LBP descriptor with similarity measure. Findings: Experimental result shows that the proposed method is robust to obtain foreground extraction with outstanding performance under various lighting conditions. Applications/Improvements: In this paper, proposed method can be used in variety of applications such as detection of objects under some climatic conditions like fog, smoke, dew, snow falling areas. Further improvements are made to remove shadows.