| Along with the rapid economic development, highway-building projects are growing steadily. How to realize their scientific, efficient and suitable operation management is put on the agenda. Vehicle recognition is the first concern. Using vehicle detectors to recognize vehicles is a common vehicle-detect method. But this method has many imperfections. There is also problem in manual recognition. This problem is mainly the economic loss caused by the toll collector's neglect, misjudgment and embezzlement. For the variety of vehicles, either method is used, the workload is heavy. At the same time, for each parking toll station, manual vehicle recognition is liable to cause traffic jam, further to lead to commercial vehicles' delay in the transportation and increase their transportation cost. Manual vehicle recognition has become the bottleneck for toll stations to enhance their work efficiency.A convenient and practical vehicle recognition system based on image processing is introduced in this paper. Three goals are targeted. The first goal is to improve the existing vehicle recognition methods and offer technical support for non-parking toll collection. The second goal is to offer a modernized technical way of traffic survey. The third goal is to offer a new thinking pattern to realize a unified vehicle classification standard. The probability of instant, scientific and reasonable vehicle recognition by a minimum system through image processing is explored. If this probability can be achieved, auto management of urban road and highway traffic flow can be effectively assisted. Safe operation management of the whole system can also be achieved. The function of vehicle recognition is accomplished by a simple set and use of software to achieve image processing arithmetic.In this paper, vehicle classification standards both at home and abroad are analyzed first. Vehicle's length, width, height and passenger car or van characteristic value are made as vehicle recognition parameters. Then image processing involving in the system is explored. Grounded on this theory, template-matching based characteristic point auto search method is put forward. The arithmetic model of vehicle's length, width and height is built on the foundation of integral idea. Passenger car or van characteristic value arithmetic is built on the base of passenger car and van image characteristics analysis. Next classification methods of pattern recognition are comprehensively analyzed. Vehicle recognition model is built by using ideas such as data clustering, neural network and statistical analysis. At last, in order to meet the requirements of specific application, the whole system is deigned and the system's hardware and software modules are exemplified. Vehicle recognition system is validated by the examination platform Visual C++6.0. Further improvement of relative arithmetic model is also made. |
