基于改进蚁群算法的行人运动特征跟踪提取方法

Tracking and Extraction Method of Pedestrian Motion Features Based on Improved Ant Colony Algorithm

为了准确关联行人运动目标与特征数据,提出基于改进蚁群算法的行人运动特征跟踪提取方法。以道路监控视频作为研究对象,检测行人运动目标,提取行人轮廓特征与肢体关节角度特征,基于改进蚁群算法求解多行人运动目标特征数据关联问题,获取行人运动目标及其特征的关联对,应用Mean Shift算法对目标进行跟踪,实时提取运动特征,实现了行人运动特征的跟踪提取。实例分析结果显示,目标与特征关联时间、关联对正确率均符合行人运动特征跟踪提取需求,充分证实了提出方法的有效性。

To accurately associate pedestrian motion targets with feature data,a pedestrian motion feature tracking and extraction method based on improved ant colony algorithm was proposed.With the road surveillance video as the research object,pedestrian moving objects were detected,and pedestrian motion features(pedestrian contour features and limb joint angle features)were extracted.Based on the improved ant colony algorithm,the data association problem of multiple pedestrian moving objects was solved,and the association pairs of pedestrian moving objects and their features were obtained.Mean Shift algorithm was applied to track the objects,and the motion features were extracted in real time.The pedestrian motion feature was tracked and extracted.The experimental results showed that the time and accuracy of target-feature association met the requirements of pedestrian motion feature tracking and extraction,and fully verified the effectiveness of the proposed method.