基于几何和颜色综合特征的车牌自动定位及识别

License Plate Automatic Detection and Identification Based on Geometric and Color Comprehensive Features

研究设计了在普通的监控条件下,基于几何和颜色综合特征的车牌自动定位及识别方法.该方法综合运用车牌的垂直边缘和车牌照颜色定位出若干候选车牌区域,根据车牌纵横比等条件去掉不符车牌几何特征的选区,然后采用支持向量机(SVM)分类器对候选区做进一步的分类筛选.将车牌选区内的字符自动分割出来后,调用训练好的BP神经网络对单个字符进行识别,最后将识别结果按序排列后即可得到车牌号.该方法不需要给汽车增加射频卡等特殊的设备条件,成本更低、可以更加友好地应用于不同的需求.最后在实际开发的基于车牌识别的车辆智能门禁系统中验证了本文方法的有效性.

The automatic detection and recognition methods for license plates under common monitoring conditions are studied and designed. The method uses a combination of the vertical edges of the license plate and the color of the license plate to locate a number of candidate license plate areas, removes the constituencies that do not match the geometric characteristics of the license plate according to the license plate aspect ratio, and then uses the trained support vector machine(SVM) to further the candidate area by automatically segmenting the characters in the license plate selection area, and employing the trained BP neural network to identify individual characters, and to arrange the recognition results in order to obtain the license plate number. This method does not need to add special equipment such as radio frequency cards to the car, and it is more cost-effective and friendlier to meet different needs. Finally, the validity of the proposed method is validated in the actual development of the vehicle intelligent access control system based on license plate recognition.