基于抓拍数据时空判别的车牌检错纠错研究

摘要

针对车牌识别设备应用场合背景复杂,车牌识别率易受到光线、车牌污损与遮挡、车辆行驶速度等因素的影响,车辆抓拍识别数据中存在错误车牌号码这一问题,分析了车辆抓拍识别系统提高识别准确率的方法,提出了一种以车牌识别设备采集到的源数据为基础进行车牌号码检错与纠错的方法.首先基于抓拍数据计算路段平均行驶速度,通过对抓拍数据时空判别进行了车牌检错,然后对错误车牌进行了字符替换并作时空判别进行了车牌纠错.最后在杭州市主城区大量的抓拍数据与机动车登记信息数据中进行了实验,并以车辆抓拍识别设备采集的视频图像数据进行了校验.研究结果表明:该方法可以较好地检测出错误识别的车牌号码,进一步提高车辆抓拍数据的准确率.%Aiming at the problem that complex application background of license plate recognition equipment,the recognition rateis vulnera-ble to light,license plate fouling and blocking,vehicle speed and other factors,there are wrong license plates information in the collected da-ta,the method of improving the recognition accuracy of VLPR was analyzed. Based on the data collected from license plate recognition de-vices,a method of error detection and correction was proposed for license plate recognition. First the average travel speed of road based on the captured data was calculated and wrong license plate by time-space judgement was detected. Then character of wrong license plate was re-placed and the wrong license plate by time-space judgement was corrected. Finally this method was validated in the big data collected from li-cense plate recognition devices of Hangzhou and motor vehicle registration information data of Hangzhou.Besides,the experimental results were verified by video image data collected from license plate recognition devices.The results indicate that this method can detect the wrong license plate,improve the accuracy of data collected by VLPR system.