Research on Human Abnormal Behavior Detection and Recognition in Intelligent Video Surveillance
|School||Zhejiang University of Technology|
|Course||Applied Computer Technology|
|Keywords||Intelligent Video Surveillance Abnormal behavior detection Hovering trajectory Recognition of human action Non - negative matrix factorization Hidden Markov Models|
In recent years, the intelligent video surveillance is becoming an emerging application in the field of computer vision, direction, and with the monitoring system in the traditional sense, the main difference lies in its intelligence, intelligent video surveillance not only with a camera instead of the human eye, and instead of using a computer people, to assist people to complete the task of monitoring or control and alleviate the burden on the people. This will not only save a lot of manpower, material and financial resources, the more important is its ability to detect abnormal conditions to avoid the occurrence of various types of abnormal events in surveillance scene. Intelligent video surveillance for its broad application prospects and great potential economic value is beginning to attract the attention of domestic and foreign scholars and research institutions. Intelligent Video Surveillance System is a cover image processing, pattern recognition, artificial intelligence, and many other technology applications, this article is mainly for wandering trajectory of intelligent video surveillance in target detection as well as human abnormal behavior recognition applications direction key issues study, the main research content and results are as follows: briefly the current development of intelligent video surveillance systems, detection and behavior recognition method and the existing body hovering trajectory learning and research, including some wandering trajectory method to judge, and Hu moments Zernike moments and the R transform human behavior feature extraction method and analysis of its shortcomings. In response to the current lack of track detection method, this paper proposed a wandering trajectory based on the angle of the human body detection and analysis methods. A generic algorithm, the method according to the behavior of the moving target trajectory in real time all kinds of wandering behavior judgment, the experimental results show that this method does not require any training sample, the greatest extent possible to reduce the time complexity of the algorithm and space complexity. 3 a recognizable human behavior based on non-negative matrix factorization (NMF) and hidden Markov model (HMM). Non-negative matrix factorization method is applied to the extraction of human behavioral characteristics to determine each video sequence based matrix and the number of basis vectors, and eventually get its characteristic matrix; behavior recognition and classification of the extracted features using Hidden Markov Models through the Baum-Welch algorithm the estimated HMM optimum parameters, and compare the likelihood value of each component of each type of behavior to complete the identification process. Hu moments and R transform human behavior feature extraction method, compared to the experimental results show that this method can human behavior recognition, the recognition rate is significantly higher than the other two methods, intelligent video surveillance system automatic analysis of human behavior has important significance. 4. Designed with a small intelligent video surveillance system, the proposed two algorithms and some other basic functions integrated coordination unify comprehensive monitoring scene. The system validation and implementation, the proposed algorithm prove its feasibility in practical applications.