【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。更多还原

【Abstract】 A taxi-car identifying system based on surveillance video is presented.One vehicle is tracked in a surveillance video,and its HOG(Histogram 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.更多还原