Surveillance Video Based Video Analysis Method

With the development of surveillance technology, national attention on public safety, a large number of surveillance cameras being installed to ensure the public safety of the residents, but to record the massive surveillance video data. How to quickly find and analysis video content of interested in the mass surveillance data, has become the surveillance video analysis industry’s urgent needs and challenges. Video summary technology is the preferred means of resolving such problems. In recent years, although the video summary techniques have been some development, but to really put into practical use also needs to develop over time. So the video analysis method based on surveillance video has great significance.Firstly, to accurately extract moving objects in surveillance videos, an effective iteration algorithm, which combines the Vibe foreground extraction method and a simple background subtraction method, is proposed. Limitations of these two methods are analyzed, and the corresponding reasons are also studied.Secondly,in order to solve the moving object can not be synthesized by category problems. Combined SVM image classification algorithm put forward based on statistical classification method of moving objects and successful solution of the problem of moving object classification.Finally,in order to solve the video summary of during synthesis parameters difficult to configure, and foreground background fused excessive unnatural problems, the energy function for abstract video track planning process has been improved,for every little energy function sets the coefficient,and normalized the range of all the energy function,simplifies the user setting parameters. And raised the importance based moving objects trajectory planning method. In the video fusion process, synthetic background, foreground pre-fusion, fused foreground into background are presented. Successfully resolved the foreground background excessive unnatural problems.