变电站场景下高置信度策略的DSST目标检测及跟踪  

DSST Target Detection and Tracking for High Confidence Strategy in Substation Scene

作　　者：扆梦楠 白一帆 邓红霞[1] 李海芳[1] 李杨 武淑红[1] YI Meng-nan;BAI Yi-fan;DENG Hong-xia;LI Hai-fang;LI Yang;WU Shu-hong(College of Computer Science and Technology,Taiyuan University of Technology,Jinzhong 030600,China)

机构地区：[1]太原理工大学计算机科学与技术学院,晋中030600

出　　处：《科学技术与工程》2018年第23期257-264,共8页Science Technology and Engineering

基　　金：国家自然科学基金(61472270)资助

摘　　要：目标跟踪在视频监控、人机交互及图像压缩等领域发挥着重要作用。由于DSST(discriminative scale space tracking)跟踪算法需要手动标记初始帧的位置,并且其实时速度和更新策略有待改善。因此,提出了一种改进的DSST跟踪算法;该算法在DSST算法的基础上融合了检测算法和图像缩放算法,同时引入了APCE(average peak-to correlation energy)置信度指标。使用变电站和benchmark数据集对其进行了测试,并与FCT(fast compressive tracking)及Staple等算法相比。结果表明该算法提高了跟踪精度,缩短了每帧的运行时间,且在背景、尺度变化及遮挡等方面表现出较强的鲁棒性。Target tracking plays an important role in video surveillance,human-computer interaction,and image compression.Because the DSST(discriminative scale space tracking)algorithm needs to manually mark the location of the initial frame,and its real-time speed and update strategy need to be improved.An improved DSST tracking algorithm was proposed.The algorithm combines the detection algorithm and the image scaling algorithm based on the DSST algorithm,and introduces the APCE(average peak-to correlation energy)confidence indicator.It was tested using substation and benchmark datasets and compared with algorithms such as FCT(fast compressive tracking)and Staple.The results show that the algorithm improves the tracking accuracy,shortens the running time of each frame,and shows stronger robustness in terms of background,scale changes,and occlusion.

关 键 词：目标跟踪 图像缩放 目标检测 视频监控 变电站场景 置信度 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]