基于改进TLD算法的视频目标跟踪

摘要

视频目标跟踪受到姿态、相似背景和遮挡等因素干扰,导致传统跟踪学习检测(TLD)算法无法实现视频目标准确跟踪.为解决传统TLD算法的局限性,获得理想的视频目标跟踪结果,提出基于改进TLD的视频目标跟踪算法.采用卡尔曼滤波器对目标区域进行预估,减少目标检测范围,采用L1跟踪器代替传统TLD算法的跟踪器,在Matlab 2014平台上进行视频目标跟踪仿真实验,实验结果表明,相对于传统TLD跟踪算法,改进TLD算法解决了目标遮挡和漂移难题,提高了视频目标跟踪精度,加快了视频目标跟踪的速度,增强了相似背景的视频目标跟踪性能.%In the video target tracking process,factors such as attitude,similar backgrounds and occlusion influence the tracking performances,so the traditional TLD algorithm is difficult to realize accurate video target tracking.To solve the limitations of traditional TLD algorithms and obtain ideal results of video target tracking,a video target tracking algorithm based on improved TLD was proposed.Kalman filter was used to estimate target area to narrow down the scope of target detection,L1 tracker was used to replace original tracker in traditional TLD algorithm,and video target tracking simulation was carried out on Matlab 2014.Results of simulation show that compared with the traditional TLD,the proposed algorithm solves the object occlusion and drift problem.It improves the tracking precision,accelerates the speed of video target tracking,and enhances video target tracking performances in similar background.