基于神经网络和支持矢量机的多机动车车牌在线检测方法

An On-line Method for Multi-license Plates Recognition Based on Neural Network and Support Vector Machine

针对道路交通多车牌识别问题,提出了一种快速鲁棒的多车牌检测识别方法,包括多车牌检测和车牌字符识别两部分:构造BP (Back-Propagation)神经网络模型用于颜色识别,结合图像形态学运算方法,筛选候选车牌目标,基于支持矢量机从候选车牌目标中判别真正的车牌目标;通过轮廓尺寸判断,并结合车牌尺寸特征,依次分割提取城市代码字符块、省份代码字符块及5位机动车编码字符块,最后基于BP神经网络识别字符块内容.基于上述原理,开发了鲁棒的多机动车车牌自动检测识别系统,并在真实场景中进行了实验测试,结果表明:1)车辆在正常速度行驶条件下,系统依然可以保证90%以上的车牌检测识别正确率;2)系统可实现同时多车牌检测识别;3)文中实验硬件配置下,系统单幅图像检测识别平均时间低于130 ms,处理频率约8 Hz.

Aiming at the problem of multi-license plate recognition,a fast and robust method is proposed in this paper,including multi-license plate location and character recognition.A back-propagation neural network is built to identify colors,which combines image morphology to detect candidate plates.Based on support vector machine(SVM),the real license plate can be distinguished from candidate license plates.By combining the judgment of contour size and license plate size feature,the character blocks can be segmented and extracted in turns.Finally,back-propagation neural network is used to recognize all the text blocks.With the above principle,a robust system for automatic multi-license plate recognition is developed and its performance has been validated by experiments.The results indicate that:1)The system can still get a high correct rate more than 90%under the condition that cars are traveling at normal speed;2)The system can recognize multi-license plates simultaneously;3)The average time cost of processing single frame is less than 130 ms and the frequency of the system can reach 8 Hz under the current configuration.