Research On Human Behavior Recognition And Application In Video Monitoring Of Hydropower Station

As an important component of computer vision and pattern recognition, vision-based human activity analysis is widely used in intelligent surveillance and advanced human-computer interaction, and becomes one of a future direction of prospective research currently. In recent years, accidents which is caused by illegal behaviors of staff have occured frequently in the power system. To detect illegal behaviors, an intelligent warning system is necessary in the working environment of power system, therefore intelligent surveillance technology becomes more and more important.Generally, a human behavior recognition system requires to analysis and detect the abnormal behavior of the crowd in a specific environment. In view of the actual situation of human behavior recognition system in video monitoring of hydropower station, this paper proposed a human behavior recognition algorithm, which combining rule-based and discriminant common vector.In this paper, the main work and achievements are as follows:1. In view of the human behavior need to identify, this paper adopted the method of characteristic data statistical analysis, and established the initial classification rule tree according with the actual demand of the system.2. In view of the limitations of established rules artificially, this paper optimized determining threshold in the rules using genetic algorithm, and proposed a evolution of human behavior recognition method for generating rules based on genetic algorithm. The recognition accuracy and scope of application has improved significantly.3. Aiming at the problems of small sample in behavior recognition and rule-based method may not integrity, this paper improved the traditional discriminant common vector classification method, and applied the improved discriminant common vectors classification method to human behavior recognition to improve the identification effect of the system.4. The paper systematically analysis the human behavior recognition process in detail and provides sufficient experimental verification to the algorithms proposed in this paper, and put forward a design scheme for human behavior recognition in video monitoring system of hydropower station.