Picture In Picture Video Surveillance System, Based On Real-time Moving Object Tracking

Video monitor system plays an increasingly important role in today's social safeguard, which has become one of the key fields in the application of computer vision. Considering the deficiency of common video monitor system, a PIP (picture in picture) video processor with character-overlapping system was designed, in which real time tracking technology on moving objects was included to improve monitoring performance.Frame buffer addresses and screen pixels are one to one correspondence in the digital video signal process system, so the control of picture display can be realized by operating frame buffer. Video process system can obtain various sizes of pictures by scaling several input video signals, then write the data into buffer corresponding address according to control commands, so it is able to output video signal with multi-pictures overlay after coding. Character overlapping can be realized by mixing character data and video signal through the same way. Besiges, the difference comparison of continuous video frames can obtain changed pixels caused by moving object, and its clear silhouette can be extracted after connectivity analysis and edge detection etc., which is used for behavior comprehension and tracking.A PIP video processor on NVC1001 was designed, which mixed several video signals into one to achive PIP overlay, reducing the amount of video data as well as improving video security, and it also could be centralized controlled. Depending on the BMP overlay function of processor, a character library system and its related conversion model were designed, which realized overlapping any fonts on any position of screen to give more information, such as title, date-time, information of temperature and humidity. Meanwhile, system analyzed and processed the continuous video frames filmed by camera, and a modified symmetric frame difference algorithm was presented to detect then extract moving objects from video frames. A clear silhouette would be obtained after processing, which included frame pre-processing, frame difference, mathematical morphology thinning, connectivity detecting and Sobel operator edge detecting. A new tracking algorithm based on feature points was that, firstly, counted out the offset between moving objects' centroid and screen midpoint, then converted and output corresponding torques to control the rotation angles of platform together with camera, and realized object tracking by matching the two points, and the simulation results proved that the tracking algorithm programmed in VC++ 6.0 was effective and feasible. The PIP video processor had been applicated widly in highway toll booths, procuratorate interrogation rooms and so on, which was higly approved by market and the users because of its good monitoring performance.