Study Of The Independent Moving Objects Detection And Tracking In Video Surveillance

Intelligent video surveillance system arose much more attention in the field of video data mining, which has significance for theories and application. To analyze each object independently is the base of the video data mining. Detection, segmentation and tracking of object are the key technology of analyze. This paper focuses on the three aspects.Firstly, we discuss the foreground object region robust detection technology in the static surveillance scene. Improving background Gaussian mixture model to conquer some typical problems like mixture model superposed question, the hole question and the model untruth question. Secondly, at the base of the region detection, author design a segment method with measuring the similarity of distance and color character, which can segment the foreground region to independent objects. Thirdly, using a double differentiation center of mass tracking model, which translate the common differentiation objects tracking to the low differentiation blob objects center of mass tracking to reduce the complexity. We also optimize the SSDA matching algorithms in the Kalman filter tracking course to accelerate. Experimental results show that the new algorithms can run in a robust and reliable way.Finally, we integrate algorithm software modules with the hardware equipment, and achieve a realtime application system, which can deal with real-time camera video data and correct interleaved saw-tooth effect. In order to ensure the repetition of experiments, the system also support the standard record video data and the H.264 format record video data.