电力监控场景下基于光流特征点的目标跟踪算法  

Object Tracking Algorithm Based on Optical Flow Feature Points in Power System Monitoring Scene

作　　者：金玥佟 杨耀权[1] 杜永昂 JIN Yuetong;YANG Yaoquan;DU Yongang(Department of Automation,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学自动化系,河北保定071003

出　　处：《电力科学与工程》2020年第5期40-47,共8页Electric Power Science and Engineering

摘　　要：针对在特定电力系统监控场景下的目标跟踪问题,提出了一种基于光流特征点的目标跟踪算法。首先,对Kanade-Lucas-Tomasi(KLT)跟踪算法提取到的特征点进行背景特征点滤除,分离出关键特征点;其次,利用Density Based Spatial Clustering of Applications with Noise(DBSCAN)聚类方法对关键光流特征点进行聚类处理,区分出不同运动目标;最后,在KLT跟踪算法中引入Kalman滤波器对因遮挡导致的跟踪目标识别不全甚至目标丢失进行了优化。仿真实验结果表明:提出的算法能够在电力系统监控视频中实现对多目标的有效跟踪,并对跟踪目标遮挡情况有较高的鲁棒性。Aiming at solving the problem of object tracking in specific power system monitoring scene,an object tracking algorithm based on optical flow feature points is proposed.First of all,all the background feature points were removed in order to refine the feature points,which were detected and tracked by the Kanade-Lucas-Tomasi(KLT)algorithm.Secondly,the key optical flow feature points were clustered by the Density Based Spatial Clustering of Applications with Noise(DBSCAN),which distinguished different moving objects.Finally,the Kalman filter is introduced into the KLT algorithm to optimize the incomplete or even lost tracking object caused by occlusion.The simulation results show that the algorithm can effectively track multiple objects in surveillance video of electric power system.This method is robust under partial occlusion.

关 键 词：电力系统 目标跟踪 光流特征 KLT算法 DBSCAN聚类 KALMAN滤波 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]