CAJ | 학술논문

针对车牌字符识别中字符变形失真影响识别率的问题，文章提出一种基于切线距离的车牌字符识别算法。首先对车牌字符图像进行预处理，在先验知识的基础上确定切线向量，然后通过核主成份分析对样本图像和切线向量降维，并引入切线距离度量样本之间的相似度，最后利用K-NN分类器对字符进行识别。实验结果表明，该算法有效地保留了车牌字符的非线性信息，减少了识别时间，并且显著提高了识别率。
침대차패자부식별중자부변형실진영향식별솔적문제，문장제출일충기우절선거리적차패자부식별산법。수선대차패자부도상진행예처리，재선험지식적기출상학정절선향량，연후통과핵주성빈분석대양본도상화절선향량강유，병인입절선거리도량양본지간적상사도，최후이용K-NN분류기대자부진행식별。실험결과표명，해산법유효지보류료차패자부적비선성신식，감소료식별시간，병차현저제고료식별솔。
In order to solve the problem of the character distortions affecting the recognition rate in license plate character rec-ognition, this paper proposed a license plate character recognition algorithm based the tangent distance. The images of the li-cense plate character should be preprocessed firstly. It selected the tangent vectors by introducing a priori knowledge. And then reduced the dimensions of the sample images and the tangent vectors with the kernel principal component analysis algorithm. Tangent distance is introduced to measure the similarity between samples. Finally, the K-NN classifier recognized the character. Experimental results show that the algorithm retains the nonlinear information of the license plate characters effectively. It re-duces the recognition time and improves the recognition rate.