一种基于监控视频的出租车识别方法  

Surveillance Video Based Taxi-Car Identifying

作　　者：朱伟[1,2] 叶飞[2] 马超[2] 张重阳[2,3] 

机构地区：[1]上海市公安局青浦分局科技科,上海201700 [2]上海交通大学图像通信与网络工程研究所,上海200240 [3]上海市数字媒体处理与传输重点实验室,上海200240

出　　处：《电视技术》2013年第7期154-156,166,共4页Video Engineering

基　　金：国家"十二五"科技支撑计划项目(2012BAH07B01)

摘　　要：研究了一种基于视频监控的出租车识别算法。对已经完成跟踪的车辆,通过提取车辆的方向梯度直方图(HOG)特征,作为支持向量机(SVM)分类检测的输入,进行车辆是否为出租车的分类识别。通过多窗口投票机制,增强了分类识别算法的准确性与鲁棒性。实验证明,该方法能准确进行出租车的分类识别,基于实际的标清监控视频,出租车的分类准确率达到90%左右。A taxi-car identifying system based on surveillance video is presented. One vehicle is tracked in a surveillance video, and its HOG （Histo- gram of Direction Gradient） features are extracted firstly, which are taken as the input of SVM （Support Vector Machine） based classifier to indentify its catalog： taxi-car or not. One multi-window voting mechanism is developed in this system to improve the accuracy and robustness of the classifying. The experimental results show that this method can identify the taxi in the surveillance effectively.

关 键 词：出租车识别 方向梯度 直方图 支持向量机 多窗口投票机制 

分 类 号：TN949.6[电子电信—信号与信息处理]