CAJ | 학술논문

针对国内车牌字符的多样性和识别效率不高的现状，文中利用模板匹配和集成学习思想设计了一种新颖的识别算法．该算法由特征加权模板的方法构建弱分类器，经AdaBoost快速提升成强分类器，利用图像的整体灰度信息，缩短大量Haarlike特征的训练时间，克服单一特征弱分类器的不稳定性．仿真实验表明，该算法能够获得较好的字符识别率和稳定性．
침대국내차패자부적다양성화식별효솔불고적현상，문중이용모판필배화집성학습사상설계료일충신영적식별산법．해산법유특정가권모판적방법구건약분류기，경AdaBoost쾌속제승성강분류기，이용도상적정체회도신식，축단대량Haarlike특정적훈련시간，극복단일특정약분류기적불은정성．방진실험표명，해산법능구획득교호적자부식별솔화은정성．
In order to solve the license plate characters' diversity and improve the recognition rate,a novel algorithm was presented. Weak classifier is firstly built by feature-weighted template matching, and then strong classifier is formed by AdaBoost. It shortens Haarlike feature's training time and overcomes the instability of weak classifier based on single feature. Simulation experiment result shows that it owns good character recognition rate and stability.