Abstract: A method based on the structural features of binary images and the color information of the color images is proposed respectively to distinguish the real images from the mixed images of license plate character images. To meet the requirement of a real-time license plate recognition system, binary images are obtained from the gray-scale char images by a fast threshold algorithm. Most of the fake char images are removed, combined the extracted structural features and the inherent characters of vehicle plates. Moreover, the color information of the fake images which are difficult to exclude from structural features will be analysed in HSV color model, and the color images in which huge differences exist from others will be removed at last by the geometric distribution of characters. Experimental results demonstrate that the proposed algorithm is feasible and has a good result. The algorithm is suitable for the real-time license plate recognition system as the post-process part of the character segmentation.?

Key words: true characters &, fake characters distinguish, characters de-noise, characters segmentation, structure features, HSV color analysis, image segmentation, plate segmentation

摘要： 提出一种融合车牌字符切割后的二值字符图像的结构特征及对应的彩色小字符图像的颜色信息对车牌分割后的小字符图像进行真伪字符图像区分，以此达到检测字符图像目的的算法。为满足实时车牌对时间响应的要求，对车牌字符切割得到的灰度图像作快速二值化，在二值化的图像上提取结构信息，结合车牌字符分布的特点去除了大部分非颜色车牌的伪字符图像。对于难以从结构上进行字符检测的颜色车牌伪字符图像，在结构特征分析的基础上再次通过从彩色图像提取的颜色信息进行相似性分析，排除伪字符图像。对候选字符图像融合大间隔这个先验知识得到输出的字符图像。实验结果表明算法有良好的字符检测效果，可以用于实时车牌识别系统中作为字符切割后期处理一部分。

关键词: 真伪字符区分, 字符去噪, 车牌字符分割, 结构特征, HSV颜色分析, 图像分割, 车牌分割