Study of the intelligent video surveillance system based on the staring ommateum model

This paper proposed a method which based on "the ommateum" staring system with multi-lens that had intelligence function and nothing blind-zone surveillance. The system used six normal lens which the field of view has 65° to encircle evenly, and made the neighboring two lens' view field have 5° overlap angles. A high speed DSP chip of TMS320DM6446 was used to central processing the six route video which for scanning method from one to six video by turns to realize the intelligence and real time processing. The system could track and recognition the moving object, calculate and record the parameter of object like speed, size and moving azimuth angle. In this paper, the improved algorithms of Auto-adapted Renewal Background Subtraction (ARBS) and the Fuzzy Auto-adapted Median Filter (FAMF) had been studied which could be detected the movement object and filter the image noise. The experiment result indicated that the staring ommateum system could be surveillance 360° scope with nothing blind-zone. The FAMF algorithm could filter the noise effectively, and the ARBS algorithm could be detection the moving object well and truly for anyone video and the parameter of object in the field of view could be calculated accurately. It had been achieved the requirement of seamless intelligence video surveillance at ultra wide range.