一种车牌字符位置侦测的新方法

提出一种快速可行的鲁棒性车牌图像二值化算法,实现了对不同质量车牌图像有普遍适用性的车牌图像二值化方法;结合设计的快速连通体标记方法、二值图像简化算法及动态车牌模板搜索算法,实现了从车牌图像中快速定位车牌字符及车牌本身并同时获得车牌字符轮廓信息的新方法.实验证明,该方法对车牌及其字符定位准确率高,且对不同条件下获得的不同质量的车牌图像有较好的适应性.

A New Approach to Location of Characters in License Plate Recognition

In this paper,a new method to detect the position of characters in license plate is studied.The aim of this method is to improve license plate recognition system to have a better robustness,a quicker executive speed and a better adaptation.A robust and rapid algorithm to obtain binary image from color license plate is first proposed,which is a region based approach and can overcome many common image flaws such as poor illumination conditions,dust,rust,mud,peeling paint,and fading color.Based on binary image converted,a component-labeling algorithm using contour tracing technique is used to get all connected components and their bounding boxes.According to the features of license plate,a clean algorithm is proposed to remove those obvious noncharacter connected-component candidates.Finally,a dynamic template of specific license plate is presented to search all character connected-components in a license plate.Therefore the positions of all characters and their contours can be known.The test results show that this approach is very efficient and has a success rate up to 99.5% in location of characters of license plate.