CAJ | 학술논문

针对智能交通系统中中文车牌图像中字符识别准确率不高速率低的问题，根据中文车牌字符图像纹理特点改进经典的局部二值模式( LBP)，并在此基础上提出一种中文车牌字符识别高效算法。该方法采用改进的局部纹理算子LBP描述车牌字符，对于中文、字母、数字这三种类型字符分别使用不同维数扩展的LBP特征描述，并通过多层感知器( MLP)分类算法识别字符，因此同时结合了LBP和MLP算法的优势。实验结果表明，与工业上常用车牌字符识别算法相比，所提方法字符识别更准确，准确率约96.5%，同时识别时间比其他常用算法缩短了24%~62%，可满足智能交通系统实时性与准确性的要求。
침대지능교통계통중중문차패도상중자부식별준학솔불고속솔저적문제，근거중문차패자부도상문리특점개진경전적국부이치모식( LBP)，병재차기출상제출일충중문차패자부식별고효산법。해방법채용개진적국부문리산자LBP묘술차패자부，대우중문、자모、수자저삼충류형자부분별사용불동유수확전적LBP특정묘술，병통과다층감지기( MLP)분류산법식별자부，인차동시결합료LBP화MLP산법적우세。실험결과표명，여공업상상용차패자부식별산법상비，소제방법자부식별경준학，준학솔약96.5%，동시식별시간비기타상용산법축단료24%~62%，가만족지능교통계통실시성여준학성적요구。
Since Chinese license plate is not easy to recognize fast and accurately in Intelligent Transportation System ( ITS) , the typical Local Binary Pattern ( LBP) was improved based on the image characteristics of Chinese license plate character, and then an effective Chinese license plate recognitiosn approach was proposed. The improved LBP was applied to describe the characters in the license plate, in which different dimension of LBP features were extended for Chinese characters, English letters and digits, the three different types of characters in Chinese license plate. And the Multi-Layer Perceptron ( MLP) classification method was applied to recognize the characters. So it combined the advantages of both LBP and MLP. Experimental Results show that compared with other common state-of-the-art algorithms, the proposed algorithm is more accurate, with the recognition accuracy of about 96. 5%, and the recognition time reduces by 24% -62%, thus the approach could satisfy the real-time and accurate demands of the ITS.