| Moving object detection is all important research field in computer vision, while in the process,the background extraction is the basic work. The results of motion detection in complex environment directly impact on the higher video events understanding. The thesis is focused on background modeling of motion detection and other related problems with complicated conditions which include illumination change, shadow, occlusion, as well as target state change. The thesis describes a variety characteristics of complex environment in background modeling. Then the thesis intruduces the classical background subtraction method, such as MOG, codebook, LBP, categories the algorithms and its suitable application scenes and anlysis the strengths and weaknesses. Thirdly,based on the space neighborhood analysis, a new algorithm is presented. That is improving LBP and codebook to detect traffic accident based on image processing. This method contains the merit of LBP and codebook, eliminateing the influencement of lane marking, pale and shadow. Finally, the thesis introduce the pretreatment andpost treatment of the image.The paper’s main innovation spot:●Improved algorithm uses the texture characteristics as consideration of backgroud subtraction. As the texture of background changes little, we can use one or several code features to describe the backgroud model.●The movement of the single pixel region, not only compared the background frame and current frame pixel value at that point, it also assessed the correlation of local neighborhood. In this way, system can sharpen the operator anti-chirp ability.●By using the local texture information, only the effective goal can be extracted, so the shadow points are well suppressed, without further eliminate shadows ●as the operations all using nteger, background information can be an effective and timely updates to address the common algorithm in the slow movement of vehicles and parking problems prone to false alarms. |
PDF Full Text Request