Vehicle Detection And Tracking Based On Video Technology

With the continued development of the societal economy and the sustained growth in the volume of traffic, congestion of traffic is so serious that we must pay more attention on it. Using the intelligent transportation systems to improve the utilization rate and the safety of road has become the potential direction for the future transportation. Intelligent vehicle is an important part of the intelligent transportation system and is also an important field of application of robot study. Because of the development of computer hardware and computer vision technology and the important role of visual information, as well as the advantages of accessing to information integrity and a wide range of signal detection of the mechanical vision sensors, the computer vision applications in terms of transport become wider, and has gradually become core component of the Intelligent Transportation Systems and Intelligent vehicle.In this paper, we study on the traffic scenes video in front of vehicle obtained by the camera within the vehicle. We mainly researched on choosing the interest region and pixel feature. The innovative works is following:1. The paper adopted a method combined of choosing shadow, red pixel and horizontal boundary in the HSV for detecting interest region of the original picture. We can simplify the computation process by using the new method in stead of the conventional edge detection method.2. Considering the deficiency of the traditional crossover operator of Genetic Algorithms in global searching, fitness scaling, niche technology, migration strategy, the best and adaptive mutation will be taken to improve standard Genetic Algorithm. Compared to the traditional improved Genetic Algorithm and the Partheno-Genetic Algorithm, the improved Genetic Algorithm used the partheno-crossover operator allows better stability of global convergence and some comparative results relative to the optimization of test functions has increase of 10 times, even Several decuples. The 2D Gabor filter optimized by using the PCGA allows better correct rate of vehicle detection.3. In vehicles tracking process, we can easily find tracking region at a certain area in the picture of the next frame by using the length-width ratio of vehicle region and the information about the vehicle shadow location. The author presented a method to compare the tracking region and the former vehicle region. The color of area pixel is matched in this method. It is found that the vehicles were tracked quickly by using that method.Finally, we summarized the work and pointed out the deficiency in this paper. Certainly, we should continue to make research on this topic.