License Plate Positioning Recognition System Design And Implementation

The rapid increasing of the number of vehicles makes it difficult for traficmanagement. To alleviate the conflicts between vehicles, pedestrians, and tranficmanagers, it is becoming an inevitable trend to improve traffic management withinformation technology and intelligent technology. Vehicles is a basic element oftransportation systems, which can be uniquely identified by the registered code on thelicense plate. So, automatic license plate recognition is an important research topic forintelligent transportation.This dissertation introduces the development of a license plate recognitionsystem. It detailly describes the algorithms of image pre-processing, license platelocation, character segmentation and recognition,and finally gives the implementationof a prototype system.In the image pre-processing step, color image is transferred to grayscale image bya weighted average method, and enhanced by a segmented algorithm. The graysaleimage is then transferred to a binary image by combination of Bernsen local thresholdalgorithm and a fuzzy enhancement algorithm. In plate location step, with the edgefeatures and a projection method, the plate area is roughly located, and then preciselylocated by erasing the frame of plate and slant correction. After that, the characterareas in the plate are segmented by vertical projection. Finally, each character isrepresented by the pixels in the character area, and sent to a character recognizor forrecognition. Our character recognizor is based on BP neural network trained on thecollected samples, and is very adaptive and robust.Based on the developed algorithms, we developed a prototype system of licenceplate recognition with VS2005 environment and Opencv toolkit. Experiments indicatesthat the system is easy-use, effective, and can accomplish license plate location,recognition and other major functions.