基于相关滤波器的抗遮挡目标跟踪算法
首发时间:2020-04-01
摘要:传统核相关滤波目标跟踪算法在尺度变化、遮挡等场景中无法准确跟踪。本文提出改进,结合自适应核相关滤波器算法,尺度变换和平移滤波环节均采用特征降维方法提升速度。融合了判断遮挡更新机制,若存在遮挡,停止对模型的更新并预测目标新位置,利用预测信息修正目标位置。通过对OTB50测试视频集跟踪实验分析,结果表明本文算法在抗遮挡和速度方面优于传统算法。
For information in English, please click here
Anti-occlusion target tracking algorithm based on correlation filter
Abstract:The traditional kernel correlation filtering target tracking algorithms can not accuratelytrack in some scenes,such as scale changes and occlusions. This article proposes improvements, combinate with the adaptive kernel correlation filter algorithm, the scale transformation and the parallel-shift filter all adopt the adopt feature dimensionality reduction to increase the speed. Integrate the judgment occlusion update mechanism,if there is occlusion, stop updating the model and predict the new target position, usethe predicted information to correct the target postion. The experimental analysis of the OTB50 test video set tracking results show that the algorithm in this paper is superior to traditional algorithms in terms of anti-occlusion and speed.
Keywords: Target Tracking Occlude Correlation filtering Feature reduction
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
基于相关滤波器的抗遮挡目标跟踪算法
评论
全部评论