Research Of Moving Object Detection Technology In Intelligent Transportation System

Recently, traffic detection and information collection in Intelligent Transportation System (ITS) area has become an important issue in the computer vision technology. However, the moving object detection technology based on video image processing and analysis is the most basic and important part in its vehicles detection system. In this paper, some exploration and research are presented and new ideas are proposed based on original theory on the issues. Experiment results confirm the effectiveness of the detection algorithm. The major works in this thesis include these contents:Firstly, the paper introduces the applying background and gives some research on currently development status. According to advantages and disadvantages in the object detection algorithms, the background subtraction as the main method is proposed to be the major research direction in this paper. And background modeling method is also the key content to explore the algorithm in background subtraction. By comparing kinds of background modeling methods, The Gaussian Mixture Model is used to simulate the initial background, which is better able to describe background's multimode features. However, through the experimental observation to the Gaussian Mixture Model, it still has some limitations in some particular complicated scenes.Next, an improved adaptive Gaussian Mixture Model of the algorithm combined with the frame difference is put forward to overcome the "shadow" phenomenon, which will appear in the situation when vehicles run slowly causing movement prospects going into Gaussian Mixture Model. In the initial judging stage of the background modeling, the interval frame differencing method is put forward to make a judgment, which can rapidly distinguish the movement area. Then in order to recover background covered by paused objects when they start to move again, the different update rates are given in different areas include background and moving areas, which can eliminate shadow that moving vehicles may leave. The improved algorithm is suitable for complicated scene environment of vehicles, whether fast or slow running, still have strong applicability, efficiency and good robustness compared to traditional algorithm.Lastly, when the vehicles'color is similar with scenes" pixel grayscale, it will show a outline lost phenomenon. Then a kind of edge detection based on Gaussian Mixture Model is put forward to apply in this circumstance. This method requires a canny edge detection in the results after frame differential, then do the logical "or" operation between the graph of background subtraction's targets and the frame subtraction's results, which can retain the outline of moving objects. The experiment results show that the algorithm has a good effect.