Research And Application Of Crowd Orientation And Analysis Method Based On Camera And Zigbee Linkage

In recent years, more and more interests have been focused on the crowd semantic mining under video surveillance in the computer vision field. Getting accurate semantic information for each staff position of the crowd plays an important role in the surveillance scene understanding. These semantic information of locations can be used to further analyze the crowd, thus to predicate and predict the scene content, therefore, the key is how to improve the accuracy of the crowd orientation.In order to solve the high-accuracy crowd orientation and analysis issues, this paper combines the traditional visual analysis method and method based on Zigbee, proposing a new solution.In this paper, the linkage analysis of video information captured by camera and accurate crowd population returned by active Zigbee, can dig out information correlation to improve the orientation accuracy. This paper presents an adaptive detection parameters optimization model in view of the detection and location parameters. This model is obtained from the statistic mining of the large sample data based on particular scene in training phase, characterizing the relationship between the key parameters of personnel detection and the accurate number of present crowd. Large numbers of experimental results show that this method improves the accuracy of the crowd orientation, and provides a new guideline to solve the crowd orientation and analysis issues in combination of camera and signal acquisition device like Zigbee under complex environment. Crowd analysis method is to research the crowd through the information including distribution of locating points and residence time after getting orientation information. This paper implements a prototype system of crowd orientation and analysis based on the proposed method. The test results show that the system has strong robustness and can easily complete the crowd orientation and analysis function in specific scenes. As the video monitor is usually used in specific scenes and the camera is generally fixed, this method has a bright prospect in practical applications.