Based On The Continuous Video Stream Of Traffic Condition And Violation Of Pattern Recognition Research And Implementation

In recent years, the quantity of motor vehicle increased significantly and traffic accidents become also more frequent. However, most of the traffic accidents are caused by illegal traffic. Therefore, detecting the vehicle behavior to give the proper education and punishment to violator is one of the effective means to reduce the illegal behavior. The traditional methods of detecting vehicle violations are based on the principle of magnetic induction. Methods through the ring detection coil, traffic microwave detector, velocity radar and other equipments to extract the traffic parameter, are used commonly in the current. The methods have obvious shortcomings, such as inconvenient installation, destruction of the road, uneasy to protection, access to traffic information less etc. Intuitive, specific, lively and informative are the characteristics of video information. Therefore, the analysis of the traffic situation and the violation behavior based on the continuous video stream has become the focus on attention of researchers and scholars at home and abroad.Firstly, thesis uses background difference method to obtain the moving target by analysis and research of the common algorithms of moving target detection, and using multi-frame mean background modeling method to build the background model and realize the update of background ceaselessly. The moving vehicles are extracted by the moving objects classified according to the ratio of the length to width, the duty ratio of the moving target area.Secondly, traditional mean shift algorithm has been improved contrary to its shortcomings by research and analysis of the application of traditional mean shift algorithm in moving target tracking. Window adaptive MS algorithm based on KF is came up with, in which Kalman Filter location prediction mechanism and window adaptive size of a target mechanism are introduced into the moving target tracking algorithm. And, the thesis analyzes the performance and the effect of tracking of the improved algorithm. The results show that the improved algorithm is better than the traditional mean shift algorithm.Then, algorithm of vehicle behavior judgment and algorithm of traffic parameter extraction are designed, and they are based on the motion of vehicle. Two algorithms are combined to form based on the color and edge detection of license plate location algorithm to achieve license plate location after analyze each of algorithms based on the color of license plate location algorithm and based on edge detection of license plate location algorithm. The results show that the accuracy of license plate detection is higher than the other two algorithms used alone. And the method of the character segmentation of license plate and the method of the character recognition are also presented.Finally, thesis designs an intelligent traffic analysis system after analyze algorithms related to based on the continuous video stream of traffic condition and illegal behavior.