Vehicle Detection And Research Based On Traffic Video

Because of the rapid boom of the national economy, the pressure of traffic increases day by day, so the Intelligent Traffic become the issue of the current traffic engineering domain. And with the rapid development of the computer technology, the application value of computer vision acts. It is used in video monitoring system, on purpose of vehicle navigation and traffic management.My article aims at vehicle type detection, counting of traffic flow and the classification of vehicle size. The vehicle type detection is now realized by the recognition of plate number, it demands the short distance for detecting and definition of high level, so it is not able to achieve the classified statistic of vehicle type in a large zone. In this paper the algorithm of vehicle type detection is proposed:Extract the standard featured vector of different vehicle types using Fourier descriptors and generate the image library, then the vehicle type detection is realized by matching the template. The statistician of traffic flow will be influenced by the sheltering phenomenon, the current solution is to use video tracking to recognize the sheltered vehicles, this requires the apparent relative movement between vehicles. Once a method of separating sheltered vehicles by the operation of morphology was put forward, but it still cannot separate the over-sheltered vehicles; The algorithm based on local texture and space filtering scanning was proposed. Firstly extract the skeleton map of vehicles, then calculate out the texture parameters of all local regions, in the end take advantage of a series of correcting algorithm to accomplish the recognition and judging. The classification of vehicle size is the reference index for charging in electronic toll collection, but currently toll gate do this work by transceiver installed in vehicles, which causes correcting error. My proposal is to establish the space model of lane based on the method of integration, and design correcting algorithm in accord with parallax principle. It is able to calculate out the length and width of the vehicle, accomplish the detection and classification of the vehicle size.In this article the simulation experiment was done on VC and MATLAB to verify the validity of my algorithm of vehicle detection. Experiment shows that the algorithm based on Fourier descriptors could perfectly get with the translation and flipping of the target, we can correctly judge the vehicle type by matching the detected vehicle and the template. The recognition method based on texture is able to adequately recognize the sheltered vehicles, thus raise the accuracy of the traffic flow statistic. In the aspect of measuring the size, I have optimized the space model algorithm according to the parallax error, so improved the precision. My work has optimized the image processing method on the vehicle type detection, calculation of traffic flow and the classification of vehicle size, and my algorithm has reach the application value.