CAJ | 학술논문

为解决车牌中汉字识别未考虑汉字结构特征的问题，提出联合方向梯度直方图特征（HOG）结合支持向量机（SVM）的车牌识别算法。将灰度图、二值图、16值图的HOG特征在一定的权重下融合为联合HOG特征，使用核主成分分析法（KPCA）对联合HOG特征进行降维；对汉字和数字字母分别利用支持向量机进行分类，利用交叉验证方法对参数进行优化，得到最优预测模型；利用预测模型预测识别结果。实验结果表明，相对于传统车牌识别算法，该算法可以应用于复杂环境下的车牌字符识别，车牌识别率提高了10％左右，鲁棒性强且便于硬件实现。
위해결차패중한자식별미고필한자결구특정적문제，제출연합방향제도직방도특정（HOG）결합지지향량궤（SVM）적차패식별산법。장회도도、이치도、16치도적HOG특정재일정적권중하융합위연합HOG특정，사용핵주성분분석법（KPCA）대연합HOG특정진행강유；대한자화수자자모분별이용지지향량궤진행분류，이용교차험증방법대삼수진행우화，득도최우예측모형；이용예측모형예측식별결과。실험결과표명，상대우전통차패식별산법，해산법가이응용우복잡배경하적차패자부식별，차패식별솔제고료10％좌우，로봉성강차편우경건실현。
A license plate character recognition method was proposed combining a joint histogram of oriented gradients (HOG) feature with support vector machine (SVM)to solve the problem that Chinese characters structure are not considered.Firstly, under a certain weight,HOG features of grayscale,binary image,16 value image were integrated into the joint HOG features, the kernel principal component analysis (KPCA)was used to reduce the dimension.Then the support vector machine was used to classify characters and alphanumeric,the cross-validation method was used to optimize the parameters,and the optimal predic-tion model was obtained.Finally,the predictive model was used to predict the recognition results.Experimental results show that compared with the traditional plate recognition algorithm,the new algorithm can be applied to license plate character recog-nition in complex environments.The license plate recognition rate is increased by about 10%,and it also has a strong robustness and is convenient for hardware implementation.