一种多分辨率在线选择分支的单目标跟踪方法

A Single Target Tracking Method with Multi-Resolution on-Line Branch Selection

单目标跟踪作为计算机视觉领域中最基本的问题之一,可在第一帧视频给定标记目标后,在随后的帧中定位目标,而这其中有目标运动、视角变化、光照变化等干扰因素,针对此类由于目标发生剧烈外观变化而跟踪失败的问题,提出一种多分辨率在线选择分支的单目标跟踪方法。该方法基于孪生网络框架,结合Resnet-50深层网络,加入在线分支选择机制,构建出基于深层网络的具有在线分支选择的孪生跟踪框架。实验结果表明,在同等跟踪速率下,该方法具有更高的跟踪精度和更好的鲁棒性。

As one of the most basic problems in the field of computer vision,single target tracking can locate the target in the following frames after first frame of video is given a marked target,and there are interference factors such as target motion,visual angle change,illumination change,etc.To solve the problem of tracking failure due to the drastic appearance change of the target,a multi-resolution on-line branch selection single target tracking method is proposed.The method is based on twin network framework,combined with Resnet-50 deep network,adding online branch selection mechanism,and constructing twin tracking framework with online branch selection based on deep network.Experimental results show that the method has higher tracking accuracy and better robustness at the same tracking rate.