Research And Application On Video Moving Target Detection And Tracking Technology

Intelligent video surveillance is a multi-disciplinary frontier research subject in the field of computer vision, it has a very important of scientific meaning development and broad application prospects. Moving target detection and tracking is the most basic core technology in the intelligent video surveillance. In the whole monitoring process, moving target detection, classification and target tracking is the basic level of visual problems, describe and understanding the behavior is the higher level of visual problems. Although target detection and tracking is the lower level of the computer vision problems, but they are the prerequisite for do the research of the video monitoring technology better. Intelligent video surveillance technology is widely used, it reflect the inevitable trend of the future video monitoring system to be digital, intelligent, diversified.In this paper we do some research on the video surveillance technology in some safe places. And we focus on the static background motion object detection and target tracking, to hope to get a more reliable and practical algorithm in the intelligent video surveillance. The main work are:In the moving target detection aspect, we do a in-depth research on the current moving target detection algorithms, and detailed introduces the basic principles of the background difference method, the neighboring frame difference method and the edge detection method, moreover we do some related experimental analysis and use a background difference method combined with edge detection method which are based on a mixture Gaussian model. This method overcomes the interference effects which are the noise produced on the target information in the different scenes video images, it also have a good robustness to the background noise and accurately realize the moving target detection.In the moving target tracking aspect, this paper mainly studies two algorithms: the moving target tracking algorithm based on the kalman filter and particle filter. We do some discussion and analysis separately of both of the algorithms, furthermore we do the study on tracking algorithm based on the integration of multiple cues method. In the frame work of the particle filter, we proposed a new tracking algorithm adaptive fusion of color clues, corner clues, and edge clues. Through the simulation experiments, in the target tracking respect, the method based on the fusion of multiple cues is better than the single clue method. And it is easy to deal with the light mutations and mixed background brought the interference to the target tracking.