Research On Vehicle Detection Algorithm Based On Video Image Processing

With the development of economic and technological, the number of car ownership is rising, and the traffic problem has become more and more pressure. To build Intelligent Transportation System (ITS) has become one of the most effective methods in technological aspects to solve these problems. It owns low cost, good performance, functional diversity and so many advantages, so it can be widely used in many aspects of ITS to improve transport efficiency, reduce traffic pressure and energy consumption, ensure traffic safety.Vehicle detection technology that based on video image processing is the basic part for ITS; It is an important method to collect traffic information and it is a prerequisite to obtain good recognition results it the system, so it owns important research significance. This paper is involved in the exploration and research of the key issues and related algorithms in the vehicle detection based on image processing. This paper represents an effective image enhancement method especially for the foggy weather, an improved target detection algorithm based on background updating, including initialization and updating of the background in real time. Then do some post-processing for the detected moving target in order to get better results, such as shadow segmentation, morphology filling and non-vehicles target filtering. Use MATLAB to simulate the valiance of each algorithm. The contents of this paper mainly include:1. Image preprocessing algorithm: As the video image will be affected by the lights, weather changes; and the vehicle target detection on the video has some accuracy requirement, so image denoising and enhanced is required. This paper analyze the cause and type of image noise briefly, then from many experiments on different filter a conclusion can be drawn that: by the 5×5 window size, median filter can quickly receive a good result; and then for the actual situation that fog is very common in the south of China, this paper represents an algorithm used threshold frequency histogram to enhance the image contrast, simulation results show that processing speed of this algorithm is fast and it is simple to applied.2. Background updating algorithm for moving object detection: In order to avoid false alarms caused by changes, real-time updates on the background is needed. This paper first describes some existing common object detection algorithm and segmentation techniques. Then take consider of real-time of the system represents a simple real-time background image initialization algorithm and according to I component in HSI color space to determine whether update the background. Simulation experiments show that the real time algorithm can update the background adaptively.3. Research on post-processing of the detected moving target, including the shadow segmentation, holes filled, regional connectivity, etc. In this paper, morphological filtering is used to solve the problem of holes, an algorithm that based the color feature is used to remove shadow. After these operations the original contour and integrity of the vehicle target can be well preserved.