Research On The Method Of Vehicle Recognition Based On Video Image

With the rapid development of China’s economy, the number of cars has increased dramatically, bringing great pressure to traffic administrative departments. Intelligent transportation system is deemed to be the main solution to problems of traffic systems. Vehicle recognition system, the core part of intelligent transportation system, is widely used in highway automatic toll auto fare collection system, electronic police system, and security supervision system of parking lots.Vehicle recognition classification system was studied in this thesis by analyzing the video obtained at the traffic bayonet. Vehicle classification system was built based on digital image processing. The main work of this thesis is as follows:(1) Image preprocessing unit:In view of the characteristics of bayonet monitoring images, this thesis put forward an image preprocessing method based on the image filtering and image enhancement algorithm:median filter algorithm based on improved adaptive sliding window was proposed to eliminate the white line noise caused by strong sunlight; image enhancement algorithm based on improved histogram equalization was proposed to increase the contrast of the image when in sufficient sunlight. Experimental results showed that the image preprocessing method could eliminate the noise and increase the contrast which lays a foundation for following experiments.(2) Target vehicle detection and extraction unit:a fast and accurate background modeling method was proposed to quickly model and update the background of the video; an improved virtual coil method was proposed to take candid photographs of moving vehicles. Experimental results showed that the method proposed in this thesis was fast and accurate.(3) Vehicle recognition unit:an improved color space method was proposed to solve the shadow problem. Experimental results showed that the shadows of the cars could be detected accurately. In addition, a vehicle classification method based on area of vehicle segmentation, was proposed and proved to be effective.(4) Vehicle-logo recognition unit:a coarse-to-fine locating method was proposed to solve the logo locating problem caused by unfixed logo positions and shapes; a vehicle-logo recognition algorithm based on Euclidean distance and Hu moment was also proposed. Experimental results showed that the algorithms proposed in this thesis were effective.(5) A vehicle recognition and vehicle-logo recognition system based on C# was developed in this thesis. Experimental results showed that the algorithm used in this thesis could recognize the vehicle and vehicle-logo accurately. Firstly, the quality of the image was greatly improved after the image preprocess. Secondly, the background could be modeled accurately and provide real-time updates; Vehicle could be classified with an accuracy rate of over 90%; Thirdly, vehicle-logo could be located and recognized accurately. Finally, the accurate locating and recognition can be achieved, with an accuracy rate of over 85% and 89% respectively.