基于运动和外形度量的多目标行人跟踪
Multi-object pedestrian tracking based on motion and appearance measurements

单目视觉的行人检测跟踪是计算机视觉领域研究的热点和难点之一。现有的检测跟踪算法对于遮挡情况的处理仍然不够理想，尤其是未经校正的缺少深度信息的单目场景。文中面向普通单目视频，针对行人的短暂遮挡情形，提出了一种基于运动外形信息融合的多目标行人跟踪方法。首先对视频序列中检测出的行人多目标提取其运动信息和外形信息，并根据信息的分布情况调整相应系数，将行人目标与各轨迹的距离度量融合为距离代价，构成代价矩阵，最终使用匈牙利算法实现任务分配，以此进行关联跟踪。PETS2009监控视频行人跟踪数据库的实验结果表明了文中算法对于行人目标短暂遮挡处理的可行性和有效性。
Monocular vision-based multi-objects pedestrian tracking is one of the hottest and most difficult research areas in computer vision.In pedestrian occlusion situations,existing tracking algorithms are not ideal,especially in an uncalibrated monocular camera setting losting depth information.Aimed at temporary pedestrian occlusion situations in ordinary monocular videos,a multi-object pedestrian tracking method based on motion and appearance information fusion is proposed.Firstly,motion and appearance distance measures are calculated between each detected pedestrian object and each track.Then,the weight coefficients are adjusted according to the distribution of the calculated distances,so that all the distances are fused to a single cost value,forming the cost matrix.Finally,Hungarian algorithm in assign problem is used to associate each pedestrian object to its track using the cost matrix above.Tracking results in performance evaluation of tracking and surveillance (PETS) 2009 dataset indicate the feasibility and the effectiveness of the proposed algorithm in temporary occlusion situations