一种比赛运动目标自适应跟踪算法的GPU实现  

GPU Realization of Active Game Moving Target Tracking Algorithm

作　　者：马艳军[1] 于雷 MA Yan-jun;YU Lei(Sports Institute,Ningxia University,Yinchuan 750021;Network ＆ Information Management Center,Ningxia University,Yinchuan 750021 China)

机构地区：[1]宁夏大学体育学院,宁夏银川750021 [2]宁夏大学网络与信息管理中心,宁夏银川750021

出　　处：《湘潭大学学报（自然科学版）》2018年第4期71-74,共4页Journal of Xiangtan University(Natural Science Edition)

基　　金：2017年宁夏回族自治区哲学社会科学规划年度项目(17NXBTY01)

摘　　要：运动目标跟踪是计算机视觉领域核心技术之一,其软、硬件的实现方法对于促进视频和图像处理具有重要意义.以足球比赛为例,针对移动目标的大量、实时、高精度跟踪任务,提出一种基于特征融合与粒子滤波的新型自适应目标跟踪算法,并设计基于图像处理单元(graphics processing unit,GPU)的算法硬件平台.实验结果表明,所提算法及硬件平台的高效整合可以实现运动目标的可靠、快速、高精度跟踪.Moving object tracking is one of the key techniques in computer vision, and the realization of software and hardware is very important for video and image processing. A novel adaptive target tracking algorithm based on feature fusion and particle filter is proposed to deal with massive, real-time and high precision missions in football game, and the hardware platform is also designed based on （graphics processing unit, GPU）. The experimental results show the proposed algorithm and hardware platform can track the moving targets reliably, fast, and precisely.

关 键 词：目标跟踪 特征融合 粒子滤波 图像处理单元 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]