智能交通中違章車輛身份鑒別方法研究.pdf

1、河北工業(yè)大學(xué)碩士學(xué)位論文智能交通中違章車輛身份鑒別方法研究姓名：高章迎申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：機(jī)械電子工程指導(dǎo)教師：戴士杰20081101智能交通中違章車輛身份鑒別方法研究 iiSTUDY ON VEHICLES PECCANCY IDENTIFICATION RECOGNITION IN INTELLIGENT TRANSPORTATION ABSTRACT Vehicle recognition system is the

2、core of Intelligent Traffic System. It is very important in the peccancy detecting, stealing vehicle detecting etc. The purpose of the illegal vehicle recognition is to avoid traffic accidents and illegal, which through

3、acquisitioning variant features of vehicle in real-time, such as vehicle-license-plate, vehicle-logo and color of vehicle-body. By utilization of optimized recognition algorithms or information fusion techniques, the veh

4、icle’s type, its color and license plate can be automatically recognized. The paper put forward the recognition of vehicle-license-plate, vehicle-logo and color of vehicle-body to identify the vehicle real identity or th

5、e range of vehicle, aiming at the inverted license, false license and the wear license. Then, the research is divided into the following for aspects: (1) License plate location and recognition. The first step is to use

6、the horizontal scanning combined coarse texture characteristic to positioning roughly, and then to positioning accurately using the vertical projection; because the effect of tilt and border, the segmentation is not accu

7、rate, so the paper use Hough transform to correct, then removal the border, and then segment the character through the projection combined texture characteristic of the character; finally the paper use Support Vector Mac

8、hine to recognize the character of vehicle. This method has a high rate and speed of recognition, and satisfied the character recognition in real practical target. (2) Vehicle logo location and recognition. By full utili

9、zation of the location of vehicle-license-plate, and its topological structure relationship among the vehicle-logo, plate and head, the author get the location roughly of license-plate, at last, by an improved template m

10、atching method, the vehicle-logo can be accurately located; then the author put forward Tchebichef moment invariants and support vector machines to recognize the logo. The logo feature extraction is first obtained by Tch