Abstract:The adjacent frame difference is a common approach used in motion detection of video surveillance, however, this method does not suit for the case of high speed dense video surveillance. Based on this case, we propose a new and simple technique to reconstructing background of high speed dense surveillance sequences in this paper. Our method is first to extract public background of video sequences, by using the property of space distribution of a frame difference sequence and the high order statistics theory, then to form a sequence of road background by removing the background interference caused by motion according to the distribution similarity of a background frame difference image and thereby to get a road background image, finally to refresh the background image via an adaptive method of the scoreboard. The experimental results demonstrate the effect of our scheme is apparent. That is to say our scheme is a new one for reconstructing background and detecting motion target in high speed video.