Design and optimization of the moving object detecting system based on Blackfin DSP

Based on Blackfin DSP, we design and implement a compact moving object detecting and video recording system. The background subtraction algorithm is applied to the system to detect the moving objects. And when there is any object being detected, the frames containing the moving objects will be recorded automatically. In order to improve the accuracy of the detecting results, in this paper, instead of using the median filter, we design the block filtering algorithm to filter the noise with the simply connected block region, since the block filtering algorithm needs far less amount of computation than the median filtering algorithm. Further, in order to improve the detecting speed, based on the memory configuration of the system, a new block based parallel data processing mechanism is proposed, in which the external memory data to be processed is moved into the internal memory space in advance that greatly reduces the external memory access requests of the DSP core, thus good parallel data processing and transmission is achieved. The results of the experiments demonstrate that the block filtering algorithm is more suitable for a real time system than the median filter, because the block filtering algorithm improves the filtering efficiency greatly by reducing the amount of filtering computation and also works with satisfying statistical accuracy. In addition, the results of the experiments also show that the block based parallel data processing mechanism is highly effective in improving detecting speed and contributes to a more efficient parallel detecting system.