SIA OpenIR  > 光电信息技术研究室
自适应特征融合的多尺度核相关滤波目标跟踪
Alternative TitleMulti-scale Kernel Correlation Filter Algorithm for Visual Tracking Based on Adaptive Features Fusion
陈法领1,2,3,4,5; 丁庆海1,6; 常铮1,2,4,5; 陈宏宇1,2,3,4,5; 罗海波1,2,4,5; 惠斌1,2,4,5; 刘云鹏1,2,4,5
Department光电信息技术研究室
Source Publication光学学报
ISSN0253-2239
2020
Pages1-20
Contribution Rank1
Funding Organization国家重点研发计划(No.2018YFC080192)
Keyword计算机视觉 目标跟踪 核相关滤波 自适应特征融合 多尺度估计
Abstract为提高复杂场景下目标跟踪的鲁棒性,解决因光照变化、目标形变、尺度变化和遮挡等因素导致的目标跟踪失败问题,提出了一种自适应特征融合的多尺度核相关滤波目标跟踪算法。该算法首先通过两种不同的特征分别训练两个核相关滤波器,接着利用它们响应的峰值旁瓣比和相邻两帧之间的响应一致性获得融合权重,同时采用自适应加权的融合策略将它们的响应结果进行融合,完成目标的位置估计;然后以此为中心进行多尺度采样构建尺度金字塔,并通过贝叶斯估计的方法确定目标的最优尺度;最后依据目标跟踪的置信度来进行跟踪模型更新,避免模型退化。实验选取了51组视频序列进行测试,并与近年来性能优异的目标跟踪算法进行了对比。实验结果表明,本文算法能够有效降低光照变化、目标形变、尺度变化和遮挡等因素影响,在测试视频序列上取得了较高的跟踪精度和成功率,整体性能优于对比算法。
Other AbstractIn order to promote the robustness of visual tracking in complex scenes, and tackle the tracking failure problems caused by illumination variation, target deformation, scale variation, occlusion and so on, a multi-scale kernel correlation filter algorithm for visual tracking based on adaptive features fusion is proposed. In the first place, two kernel correlation filters are trained separately with two different features, then the peak side-lobe ratio of their responses and the consistence of correlation filter responses between two consequent frames are applied as weight factor for features fusion, meanwhile an adaptive strategy is adopted to fuse the two responses for position estimation. In the next place, based on the estimated position center, multi-scale image patches are sampled to construct a scale pyramid, and the Bayesian method is employed to estimate the optimal scale of target. Finally, the tracking model is updated according to the confidence of tracking result to prevent model deterioration. 51 sequences are selected for tracking evaluation, and these visual tracking algorithms with excellent performance in recent years are compared with our algorithm. The experimental results demonstrate that the proposed algorithm reduces the interferences such as illumination variation, target deformation, scale variation and occlusion effectively. The higher tracking accuracy and success rate are achieved in these sequences, and its overall performance outperforms these comparison algorithms.
Language中文
Document Type期刊论文
Identifierhttp://ir.sia.cn/handle/173321/25733
Collection光电信息技术研究室
Corresponding Author陈法领
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院机器人与智能制造创新研究院
3.中国科学院大学
4.中国科学院光电信息处理重点实验室
5.辽宁省图像处理与视觉计算重点实验室
6.北京航天恒星科技有限公司
Recommended Citation
GB/T 7714
陈法领,丁庆海,常铮,等. 自适应特征融合的多尺度核相关滤波目标跟踪[J]. 光学学报,2020:1-20.
APA 陈法领.,丁庆海.,常铮.,陈宏宇.,罗海波.,...&刘云鹏.(2020).自适应特征融合的多尺度核相关滤波目标跟踪.光学学报,1-20.
MLA 陈法领,et al."自适应特征融合的多尺度核相关滤波目标跟踪".光学学报 (2020):1-20.
Files in This Item: Download All
File Name/Size DocType Version Access License
自适应特征融合的多尺度核相关滤波目标跟踪(1520KB)期刊论文出版稿开放获取CC BY-NC-SAView Download
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[陈法领]'s Articles
[丁庆海]'s Articles
[常铮]'s Articles
Baidu academic
Similar articles in Baidu academic
[陈法领]'s Articles
[丁庆海]'s Articles
[常铮]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[陈法领]'s Articles
[丁庆海]'s Articles
[常铮]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: 自适应特征融合的多尺度核相关滤波目标跟踪.pdf
Format: Adobe PDF
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.