基于HOG特征及支持向量机的车辆品牌识别方法

发布时间：2018-05-12 23:17

本文选题：车辆品牌 + HOG特征　；参考：《东南大学学报(自然科学版)》2013年S2期

【摘要】：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3 000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.
[Abstract]:In order to solve the problem of identification between license set and illegal vehicle, a method of vehicle brand recognition is proposed. The method firstly detects the front face area of the vehicle based on symmetric features, then extracts the HOG features of the front face area of the vehicle. Finally, the support vector machine is used to classify the vehicle brand. According to the images of road bayonet provided by Suzhou Public Security Bureau, the database of Che face is constructed. The database includes 15 vehicle brands, such as Audi, Changan, Nissan and so on, with a total of 3,000 images. Based on the Che face database, the experiments are carried out using the proposed method of vehicle brand recognition, and the linear kernel function of support vector machine is compared and analyzed. Performance of polynomial kernel function and radial basis kernel function the global classification accuracy of the three kernel functions is 89.27% and 89.89% respectively. Theoretical analysis and experimental results show that the proposed vehicle brand recognition method based on HOG feature and support vector machine is feasible, and the performance of SVM classifier based on radial basis function kernel function is optimal.
【作者单位】：东南大学交通学院;
【分类号】：U495

【共引文献】