Detection Of Abnormal Aggregation Of People Based On Video

This paper is mainly focused on the research of the detection of abnormal aggregation of people and aimed on crowd density estimation. Firstly, the research and development of the digital image processing technology as well as the merits and demerits of the technology are introduced in detail. Then the author has discussed and analyzed the digital image processing technology applications in the field of pattern recognition,which lays the foundation for the subsequent processing.Secondly, the foreground is extracted by different methods. From the aspects of speed, practicality and robustness, an adaptive updating background construction method is selected, and combining with the background subtraction method it is successful to extract the foreground target crowd. According to their own needs, the existing multiple classification support vector machine classification method is improved, the results show that the method can effectively describe the population density.Next, non-parametric estimation of the distribution function method is applied to pattern recognition, providing a new idea for crowd abnormal aggregation detection.Non-parameter estimation on the population density of each point in time is subject to the distribution function, and thus can achieve monitoring population density of the target area for surveillance camera for real-time, and people are reminded automatically when abnormal aggregation occurs, to reduce and prevent the occurrence of risk events.Finally, based on the MATLAB platform, the algorithm and theory are programmed, and the video samples are used to test the program. Tests show that the algorithm in the population abnormal accumulation detection can get good results.