The Research And Actualization Of The Automatic Recognition Of Vehicle License Plates

The increasingly serious problem of the transportation is pressing our country to devote great efforts to develop an Intelligent Transport System (ITS) so as to relieve the unceasingly enhanced pressure of the traffic control. The first problem of ITS to be resolved is how to effectively obtain the information of the vehicle which would be direct basis in the Intelligent Traffic control. So the automatic recognition of the vehicle license plate is not only an important method and indispensable process but also a difficult problem urgent to be resolved.This work is the sub-project of a co-operative project of Tianjin University and Tianjin Tiandy Digital Technology Co, Ltd. The aims of the study include research on the key algorithms of the Vehicle License Plate Recognition (VLPR), performing the function of automatic location, division and recognition of vehicle license plate in software platform.This article introduces the major algorithms and processing steps for the vehicle plate recognition, compares the existing theoretical algorithms and domestic situation of the vehicle license plate, and experiments location, division and recognition combining with own thoughts. In the course of the experiments, new problems arise continuously and have been thought and analyzed, most of which have been resolved, and finally, automatic recognition of the vehicle license plate has been achieved successfully. We pay more attention to the accurate location of the effective vehicle local area and pretreatment of characters division and recognition, in order to provide a good basis for license plate recognition afterward and improve the accuracy of the recognition ultimately.The main tasks of the experiment include:1. In the process of image pretreatment, a new algorithm of contrast deviation coefficient for gray-scale images auto-estimation has been researched and put forward, and the gray-scale images have been gray stretched so as to improve the images contrast.2. In the process of fast image target edge extraction with a first-order finite difference method, threshold filter has been set up to eliminate disturbing 3. In the process of localization for the vehicle plate with projection, the method of dynamic projection scanning has been brought forward to improve the accuracy and shorten the search time.4. The problem of slanted and distorted vehicle license plate in original image has been studied and analyzed in depth. The resolution has been found and the rectification of the slant and distortion is modified.5. The feature of the noise in vehicle license plate zone location has been summarized and analyzed and therefore a noise elimination algorithm has been concluded to eliminate all noise in vehicle license plate zone location successfully.6. The procedures from obtaining information of vehicle license plate to automatically recognizing and outputting characters of vehicle license plate have been actualized in the software MATLAB.