车牌自动识别系统设计

Design of Automatic License Plate Recognition System

针对地下停车场找车难的问题,设计了一种适用于地下停车场的车牌自动识别系统,为自动寻车功能提供数据支持。首先,采用HSV颜色分割与SVM-Sobel边缘检测相结合的方法获取车牌区域,提高了车牌定位的可靠性;而后采用一种基于Radon变换与中心投影相结合的方法,实现倾斜和畸变车牌的校正;最后,采用机器学习训练字符模板来识别字符。对设计的车牌识别系统进行了不同拍摄角度、拍摄距离、光线强度等条件的测试以及效率测试,车牌定位、校正、字符识别成功率分别可达98.5%,99.0%,98.1%,系统识别车牌平均耗时4s,可较好实现地下停车场的车牌识别功能。

To solve the problem of finding the owner's car difficultly in an underground parking lot,designed an automatic license plate recognition system suitable for underground parking lots,which could provide data support for the function of finding the owner's car automatically.Firstly,used the method of combining HSV color space and SVM-Sobel edge detection to obtain the license plate area,which improved the reliability of locating license plate;Secondly,a method,combining the Radon Transform and center projection,was used to regulate the oblique and distorted license plate;Finally,the machine learning was used to train character templates for recognizing characters of license plate.Not only being done the efficiency tests,this license plate recognition system was also tested under the circumstances of different shooting angles,different shooting distances,insufficient light and sufficient light.These experiments show that the accuracy of locating,correcting license plate and recognizing its characters can reach 98.5%,99.0%and 98.1%respectively and the average time to recognize a license plate is 4 seconds,which can recognize license plates better on underground parking lots.