CAJ | 학술논문

为了解决套牌车与违章车的身份确认问题，提出了一种车辆品牌识别方法．该方法首先基于对称特征检测车辆前脸区域，然后提取车辆前脸区域的HOG特征，最后采用支持向量机对车辆品牌进行分类．实验根据苏州市公安局提供的道路卡口图片，构建了车脸数据库，该数据库包括奥迪、长安、日产等15种车辆品牌，共3000张图片．基于构建的车脸数据库，采用所提出的车辆品牌识别方法进行了实验，并对比分析了支持向量机（ support vector machine， SVM）线性核函数、多项式核函数和径向基核函数的性能，3种核函数的整体分类精度分别为89．27％，89．74％和89．89％．理论分析和实验结果表明，所提出的基于HOG特征及支持向量机的车辆品牌识别方法是可行的，并且基于径向基核函数的SVM分类器的性能最优．
위료해결투패차여위장차적신빈학인문제，제출료일충차량품패식별방법．해방법수선기우대칭특정검측차량전검구역，연후제취차량전검구역적HOG특정，최후채용지지향량궤대차량품패진행분류．실험근거소주시공안국제공적도로잡구도편，구건료차검수거고，해수거고포괄오적、장안、일산등15충차량품패，공3000장도편．기우구건적차검수거고，채용소제출적차량품패식별방법진행료실험，병대비분석료지지향량궤（ support vector machine， SVM）선성핵함수、다항식핵함수화경향기핵함수적성능，3충핵함수적정체분류정도분별위89．27％，89．74％화89．89％．이론분석화실험결과표명，소제출적기우HOG특정급지지향량궤적차량품패식별방법시가행적，병차기우경향기핵함수적SVM분류기적성능최우．
In order to solve the problem of fake-licensed car and illegal car identification, a vehicle brand recognition method is proposed.It detects the car front face based on symmetric feature, ex-tracts the HOG ( histogar of driented gradient) feature of the car face and then uses support vector machine ( SVM) to classify the vehicle brand.According to the road bayonet pictures provided by the Suzhou Municipal Public Security Bureau, a car face database is built, which includes 3000 pic-tures of 15 kinds of vehicle brands such as Audi, Changan and Nissan.Based on the built database, experiments are conducted using the proposed method for vehicle brand recognition.The perform-ance of three kinds of kernel functions ( linear kernel function, polynomial kernel function and radial basis kernel function) of SVM is compared and analyzed.The overall classification accuracy of the three kernel functions are 89.27% , 89.74% and 89.89%.Theoretical analysis and experimental results show that the proposed recognition method based on HOG features and SVM is feasible, and the SVM classifier based on the radial basis function performs optimal.