机动车车牌自动识别系统与VMLA定位算法  

A License Plate Auto-recognition System and the Vector Map Location Algorithm

作　　者：朱风云[1] 曹晓光[1] Ahmed Hassanien 胡爱明[1] 

机构地区：[1]北京航空航天大学图象处理中心,北京100083

出　　处：《中国图象图形学报（A辑）》2003年第6期679-682,共4页Journal of Image and Graphics

基　　金：航空科学基金项目 ( 99A5 1113 )

摘　　要：介绍了一种机动车车牌自动识别系统及其机动车车牌自动定位的 VML A (Vector Map L ocationAlgorithm)新算法 .该算法基于机动车车牌字符笔画两个边缘互相关值最大这一特征 ,先利用粒子图象测速原理得出位移矢量图 ,然后在位移矢量图中定位机动车车牌 .VML A算法具有实时、机动车车牌大小在较大范围内自适应的特点 ,同时还能得到机动车车牌字符与背底对比情况、机动车车牌大小的粗略估计等信息 ,这些信息对后续车牌的精确定位、分割、识别非常有用 .5 4 0幅机动车车牌图象的初步实验结果为定位速度 0 .2 s/个 ,定位正确率98.9% .这表明该方法速度快、一次定位正确率高 。A license plate recognition system and a new license plate location algorithm, which is called Vector Map Location Algorithm (VMLA), are introduced in this paper. The algorithm is based on the feature that the stroke of license plate characters and numbers have two edges which have the maximum cross correlation values. Thus by using the same method of flow PIV (Particle Image Velocimetry), a kind of vector map can be calculated and showed. VMLA is a real time and plate size adaptivity algorithm at a large range. At the same time, the contrast condition about the character to the background and a coarse estimation to the width of the stroke of license plate characters and numbers also can be derived from the map. Those above information are important to location precisely, character segmentation and recognition. The testing result of 540 images is that it costs about 0 2 second to process an image and 98 9 percent of the picture have good result, which shows that VMLA is a fast, high accuracy rate and promising location algorithm.

关 键 词：机动车 牌照自动识别系统 VMLA定位算法 智能交通控制系统 互相关矢量图 

分 类 号：U491.116[交通运输工程—交通运输规划与管理] TP391.41[交通运输工程—道路与铁道工程]