华中科技大学学报（自然科学版）2016，Vol.44Issue(11)：43-47,5.DOI:10.13245/j.hust.161110
基于多模型检测的视频监控行人跟踪算法
Pedestrian tracking algorithm by video surveillance based on ensemble detection
摘要
Abstract
The pedestrian detection and tracking problem in network video surveillance were studied , and a pedestrian tracking method was presented based on ensemble detection algorithm .Firstly ,the strengths and weaknesses of state‐of‐the‐art detection algorithms were evaluated according to experi‐mental results .Meanwhile ,an ensemble model making full use of the existing algorithms was pro‐posed w hich greatly improved the detection accuracy .By using the fast feature pyramid method ,the real‐time performance of the proposed detection algorithm was further improved .Then ,by combining the Kalman filter and Hungarian method ,the pedestrian tracking algorithm was proposed .Experi‐ments on MOT challenge video data demonstrate that the proposed method can robustly detect and track multiple pedestrians simultaneously ,w hich is suitable for video surveillance scene .关键词
网络视频监控/行人检测/行人跟踪/卡尔曼滤波/匈牙利算法Key words
network video surveillance/pedestrian detection/pedestrian tracking/Kalman filter/Hungarian method分类
信息技术与安全科学引用本文复制引用
韩光星,张轩,李崇荣..基于多模型检测的视频监控行人跟踪算法[J].华中科技大学学报（自然科学版）,2016,44(11):43-47,5.基金项目
国家自然科学基金资助项目（60703052）． （）
