【摘要】 目标跟踪在智能监控、无人驾驶、航空航天等领域有广泛的应用,其目的是在视频每一帧中找到运动目标并用目标框将其定位出来,但由于运动模糊、外观变化、遮挡、光照变化和背景混杂等原因,跟踪器在跟踪过程中极易丢失跟踪目标。由于SiamRPN模型搜索目标区域面积较小,模型有丢失目标的风险,为了提高跟踪准确率和成功率,论文提出了一种扩大搜索区域的改进SiamRPN模型ACSiamRPN,利用目标在图像前后帧间的运动信息进行目标预定位的方法,扩大目标搜索区域,借助强化学习中的Actor-Critic方法,训练预定位网络来回归目标位置,并利用预定位结果来校正SiamRPN模型搜索区域中心,从而提高跟踪准确率和成功率。在OTB2013、OTB2015、DTB70、NFS30以及VOT2016数据集上,论文提出的改进SiamRPN模型ACSiamRPN的跟踪准确率和成功率均超越了SiamRPN,运行速度达到65fps,仍然保持良好的实时性能,与当今较为先进的一些跟踪方法相比具有明显优势。更多还原

【Abstract】 Object tracking has a wide range of applications in the fields of intelligent monitoring,unmanned driving,aerospace,etc. The purpose is to find the moving object in each frame of the video and locate it with the bounding box. However,due to challenges like motion blur,appearance change,occlusion,illumination and background cluster,the tracker may easily lose the object during tracking. Since SiamRPN has the risk of losing the target out of its small searching area,this paper proposes an improved SiamRPN method named ACSiamRPN with larger searching area,using motion information of the object between two sequential frames to perform the object pre-positioning. The pre-positioning network expands the object searching area,and was trained by means of the Actor-Critic method in reinforcement learning. Then,it regresses the object pre-position and modifies the searching area center of SiamRPN,thereby improving the precision rate and success rate of the SiamRPN tracker. On the OTB2013,OTB2015,DTB70,NFS30 and VOT2016 datasets,our proposed ACSiamRPN tracker,an improved SiamRPN model,exceeds SiamRPN in terms of precision rate and success rate,running at 65 fps,which still maintains good real-time performance against several advanced object tracking methods.更多还原