基于多算法融合的抗遮挡目标跟踪算法  

Anti Occlusion Target Tracking Algorithm Based on Multi Algorithm Fusion

作　　者：林颖 汤文兵[1] Lin Ying;Tang Wenbing(College of Computer Science and Engineering,Anhui University of Science and Technology,Anhui 232001)

机构地区：[1]安徽理工大学计算机科学与工程学院,安徽232001

出　　处：《现代计算机》2021年第31期40-45,共6页Modern Computer

摘　　要：针对目标跟踪过程中的遮挡引起的精度下降、跟踪漂移、目标丢失等问题,提出一种基于多算法融合思想的抗遮挡算法,将卡尔曼滤波、MeanShift算法和SVM分类器融合。当目标正常运动或局部遮挡时,采用卡尔曼滤波结合前一帧目标位置信息预测目标当前帧目标所在位置,作为均值漂移的迭代初始点,可以缩短迭代初始点与目标实际位置之间的距离,从而减小迭代次数,提高算法的计算量。用SVM检测器在目标被严重遮挡后辅助算法定位,可以在脱离长时间严重遮挡时或者目标丢失后捕捉到目标位置,将重检测到的目标位置作为MeanShift算法的起点继续迭代跟踪。Aiming at the problems of accuracy degradation,tracking drift and target loss caused by occlusion in the process of tar⁃get tracking,an anti occlusion algorithm based on multi algorithm fusion is proposed,which integrates Kalman filter,mean shift algo⁃rithm and SVM classifier.When the target is moving normally or partially occluded,Kalman filter is used to predict the target position in the current frame combined with the target position information of the previous frame.As the iterative initial point of mean shift,the distance between the iterative initial point and the actual position of the target can be shortened,so as to reduce the number of itera⁃tions and improve the amount of calculation of the algorithm.SVM detector is used to assist the algorithm location after the target is seri⁃ously occluded.The target position can be captured when it is out of serious occlusion for a long time or after the target is lost.The re detected target position is used as the starting point of mean shift algorithm to continue iterative tracking.

关 键 词：目标跟踪 遮挡 SVM 目标检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN713[自动化与计算机技术—计算机科学与技术]