基于HOG特征的交通信号灯实时识别算法研究  

Research on Algorithm for Real-time Recognition of Traffic Light Based on HOG Features

作　　者：周宣汝 袁家政[1] 刘宏哲[1] 杨睿[1] 

机构地区：[1]北京市信息服务工程重点实验室北京联合大学,北京100101

出　　处：《计算机科学》2014年第7期313-317,共5页Computer Science

基　　金：国家自然基金:基于超图形XGML的图像半结构化研究(61271369);国家自然基金:跨媒体社群图像语义理解(61372148);北京市教育委员会创新团队项目(IDHT20140508;CIT&TCD20130513);北京市教育委员会科技面上项目:智能车实时交通标志识别关键技术研究(SQKM201411417004)资助

摘　　要：针对无人驾驶汽车的交通信号灯识别,提出了一种基于HOG特征和SVM的交通信号灯实时识别算法。该算法首先提取视频中的红绿色区域,筛选出符合条件的区域,提取各类信号灯的HOG特征,然后运用SVM构建对应类别信号灯的分类器,之后依据判决函数对当前信号进行实时判决,从而得到准确的信号灯信息。实验结果表明该算法有良好的准确率和实时性。For the recognition of traffic lights in driverless cars, this paper proposed a real-time traffic lights detection and recognition algorithm based on HOG features and SVM. This algorithm extracts red and green areas in the video preciously, and then the eligible area will be screened. Hereafter the HOG features of all kinds of lights will be extrac- ted. Finally this work use SVM to build the classifier of corresponding category lights. We can get the accurate real-time information based on the judgment of the decision function. The experimental results show that the algorithm has good accuracy and real-time performance.

关 键 词：HOG特征 SVM 交通信号灯识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]