Kernel Correlation Filter Tracking based on Spatial Constraint
首发时间:2018-05-02
Abstract:Correlation filter (CF) based trackers have become quite popular in video tracking because of their impressive performance and high frame rates. A large amount of recent research focuses on the improvement of training model of correlation filter to get a tracker with better discriminative power. However, this only helps the tracker to discriminate the target object from background within a small neighborhood, which is not suitable for fast motion scenes. In this paper, we propose a new detection model to dig out the potential of the correlation filter to deal with the challenge of fast motion. The model performs detection operations on multiple small search areas within a large one. Thus, our tracker can accurately localize the target object in a larger search area. In addition, we also added space constraints to boost the tracking performance of the model. The extensive experimental results demonstrate that the proposed tracker outperforms several state-of-the-art trackers on the challenging benchmark dataset with 51 sequences.
keywords: Visual tracking correlation filter fast motion detection model space constraints.
点击查看论文中文信息
基于空间约束的核相关滤波跟踪算法
摘要:基于相关滤波的目标跟踪算法因为其精确的跟踪性能和较高的帧率,成为视频跟踪领域的主流方法。目前,大量的研究聚焦于相关滤波器的训练模型,以此来提升跟踪器的判别性能。然而,此类方法只能在一个小的范围内区分目标与背景,从而不能很好的适应目标的快速运动情况。在这篇文章中,我们提出一种新的检测模型来提升相关滤波器处理快速运动的潜力。该检测模型在几个小的搜索区域内进行操作。因此,我们的跟踪器能在一个更大的搜索区域内对目标进行精确的定位。另外,我们增加空间约束来提升跟踪器的性能。实验结果显示,我们的跟踪器在51个视频集上展现了较几种前沿跟踪算法更优的性能。
基金:
引用
No.****
同行评议
共计0人参与
勘误表
基于空间约束的核相关滤波跟踪算法
评论
全部评论0/1000