Case Assistant System Based On The Intelligent Video Analysis Technology

With the growing urban monitoring networks, video image detection has become an important way in the criminal investigation. However, there’re some difficult problems during video image detection, such as too long time about video, too much invalid case information, too hard for searching suspected target in the video and the large cost of human resource. Through studying video summarization algorithms, aiming at the shortcomings of the classical algorithms and combining with the practical characteristics of research, the paper proposes a new video summarization algorithm based on temporal and spatial relationship. It provides a set of algorithm in the absence of artificial participation which can automatically extract and describe the characteristics and contents of the video to achieve quick search and positioning related target, and accomplish image fusion and fuzzy image processing to generate video summarization. In this paper, the main work is embodied in the following two aspects.Creatively proposed a new video summarization algorithm based on the temporal and spatial relationship. The algorithm uses marker analysis method for video structure analysis, able to quickly position the effective information from the monitoring video segments. Meanwhile, the algorithm by combining the actual characteristics of subject research, in terms of background modeling, makes further improvement on Gaussian mixture background model, greatly reducing the workload of the algorithm. In video fusion part, it also creatively puts forward the timeline reconstruction algorithm based on temporal and spatial relationship, considering the video motion density, to make the video summarization better.And introduce the blur image automatic identification and processing algorithm. Because of the possible displacement of camera surveillance video or environmental noise interference, video images may produce a certain degree of blur. As for case surveillance video, its clarity directly affects the credibility of the video, and indirectly influences the working efficiency. Therefore, the quality of video image analysis and processing is of great significance. Through studying and researching the classic image de-blur processing technology, I propose an algorithm which can automatically identify and process blur image using DFT, texture and color feature information.During the research period, I also design and implement the case assistant system based on the algorithms that I have mentioned before. And we can see from the view point of engineering practice, the above algorithms are verified, and combining with the characteristics of the subject, the improvement measures are put forward. According to the experimental results and analysis, the algorithms can effectively achieve the system main functions to solve practical problems in real-time environment.