Research On Video Data Processing Technology For Observing And Recording

Observational method is an important technique used for in-depth study the interested target's behaviors. It captures changes at the important moments, and summarizes laws over a period. Video observation method has predictive applications in military training, scientific research, children care and so on, which includes the approach of video observation with the continuous, comprehensive, objective, reproducible data collecting capacity. In order to use the video information efficiently and browse it conveniently, this paper presents an in-depth research on the data processing of observation and record, aimed to make the System more intelligent. The main work and innovations are as following:(1) Based on deeply understanding of the meaning of observation and record, we compared the difference between video observation and visual surveillance. Analyzed the characteristics of requirements of Observation System, abstracted the video observation data processing flow and key methods.(2) Took people as the object of video observation, described the condition of multiple people coupling within single video data, analyzed the demand of users for video decoupling and record classification of frame images, and proposed the method to use face detection and recognition as classifying record "trigger information". Implemented the methods of face detection and face recognition, experimental results showed that taking face recognition as a record criterion is feasible. On this basis, designed and realized the video classification algorithm based on face detection and recognition, analyzed and summarized the test results of the algorithm.(3) Introduced the significance of multiple video observation to obtain comprehensive information, elaborated the task to obtain synchronous multiple observation data in data processing, analyzed the requirements of recombining multiple videos for record, proposed an approach that was to use the face angle estimation as the switching information for different video observer. Studied the University of California researchers' face angle estimation algorithm, applied this method, designed and completed a multi-video reorganization algorithm based on face angle estimation, analyzed and summarized the running results of the algorithm.(4) Analyzed the implementation of Observation System, completed the construction of hardware platform, designed and implemented software platform by using MATLAB language. The problem of insufficient memory when MATLAB platform read large amounts of video data was solved by changing the data processing flow and optimizing the program structure.