Study On Detection Algorithm Of Moving Objects Based On DSP

It has been a key technology that to detect and identify the moving objects from the motion image sequence. This paper processes the video image sequence, and researches on the methods in this field.First, this thesis introduces and analyzes several the mature existing methods of detect the moving objects. For the problems in the classic moving target detection algorithm, combined with video surveillance features, then uses the method combined background subtraction with difference intersection movement to detect moving objects.The algorithms extract the background model from the motion image sequence, hold and liveupdate the background of it.Then use the background difference and difference intersection movement to detect moving objects. And then using morphology processing method to deal with the result of the detect objects, there be a more clear and accurate moving target. At last, transplant the algorithm to DM642 EVM board; optimize the code according to the structural characteristics of DSP. Thus the detect algorithm meet the characteristics of real-time video surveillance of the moving target test results.In order to verify the superiority of this algorithm, this paper uses several different types of video surveillance scenes video image sequence to detect the moving target. The test results show that the algorithm applies to the following scene:High speed motion, complexity scene and moving object; Low speed motion, simply scene, intensity changed light, etc. The algorithm may abstract the moving object in the same video sequence.This algorithm can not influenced by light, the type of moving target and the speed of movement, using the high-speed multimedia processors to make the algorithm higher processing speed for video surveillance system.