Research on Detection Technique of Multiple Moving Pedestrians in Sequence Image

This paper proposes one kind of multiple moving pedestrians´ detection and recognition method in the driver assistance system which is based on the computer vision. Firstly, it uses the improved Consecutive Frames Subtraction to establish a background model for sequence images, and unifies Background Subtraction to carry on an effective detection for multiple moving objects, thus obtains the clearer and more complete prospect moving objects. Secondly, it uses the region marking algorithm to demarcate multiple moving objects. Finally, the moving objects are quickly recognized and demarcated according to pedestrian´s Length and breadth ratio character and moving velocity character. The experimental result indicates that this method can detect and recognize the multiple moving pedestrians accurately and fast in the sequence images.