The Study, Based On Video Images Of Traffic Background Extraction Method

Motion detection is a key step in the traffic video surveillance, while in the process, the background extraction is the basic work. Together with the increasingly serious traffic situation, most experts take their focus on the background extraction method which can detect road accurately and rapidly.This paper analyzes the current traffic background extraction algorithms, and puts forward a new method of background extraction. In this method, we divided the road into linear load and curved road according to the road alignment structure firstly; and then designed different algorithms according to the road alignment structure. For linear road image, enhance the contrast of image through morphological algorithm at first; secondly, extract the edge of road based on Hough transform, at last determine the region of road. While in the case of curved road, first of all, extract the texture characteristics of fuzzy area which are caused by the moving objects, secondly make use of the regional growth which can ensure the position of the curved road. All the algorithms had been implemented on MATLAB 7.1 platform and the results were compared with the following steps.After the localization of the region of the road, we need to update the background because of the transformations of background caused by the changes of environment such as light. Based on the model of interval distribution to tracking the transformations of background in this paper, the testing time is shortened nearly two-third because the volume of data is reduced based on this algorithm.The experiments show that the background extraction algorithm proposed in this paper reduces the test data, and carry out the background extraction and change the background in real-time. So through this algorithm, it can provide reliable information and data for the further motion detect and traffic monitoring.