Learning Adaptively Context-Weight-Aware Correlation Filters for UAV Tracking with Robust Spatial-Temporal Regularization | |
Dongze, Hao1,2; Yinghao, Cai2; Yiping, Yang2; Jixiang, Zhang2 | |
2021-06 | |
会议名称 | 2021 The 4th International Conference on Image and Graphics Processing |
会议日期 | 2021-01 |
会议地点 | Sanya, China(线上) |
摘要 | Recently, Discriminative Correlation Filter (DCF) based methods have been widely applied in tracking for unmanned aerial vehicles (UAVs) because of their promising performance and efficiency. However, boundary effect, filter corruption, lack of context information and the poor representation of the object lead to the decrease in discriminability. In this paper, a novel learning adaptively contextweight-aware correlation filters with robust spatial-temporal regularization method (ACRST) is proposed. Both convolutional features and hand-crafted features are employed to improve representations for object appearances. Then the ACRST tracker extracts context samples around the object to help the filter be aware of the background information and adaptively learns the weights of these context patches. Thus, the tracker can improve the robustness against background noises especially for similar samples. Meanwhile, the tracker merges a robust spatial-temporal regularization to prevent the filter corruption and boundary effect. We design a center-attention spatial regularizer to focus on the valid information of the object better and we propose a method to obtain the value of the parameter of the temporal regularization adaptively. Extensive experiments have been conducted on 123 challenging UAV tracking sequences. The results prove that our tracker performs better than other state-of-the-art trackers |
DOI | https://doi.org/10.1145/3447587.3447599 |
收录类别 | EI |
语种 | 英语 |
七大方向——子方向分类 | 机器人感知与决策 |
引用统计 | |
文献类型 | 会议论文 |
条目标识符 | http://ir.ia.ac.cn/handle/173211/44985 |
专题 | 综合信息系统研究中心_视知觉融合及其应用 |
通讯作者 | Jixiang, Zhang |
作者单位 | 1.School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China 2.Institute of Automation, Chinese Academy of Sciences, Beijing, China |
第一作者单位 | 中国科学院自动化研究所 |
通讯作者单位 | 中国科学院自动化研究所 |
推荐引用方式 GB/T 7714 | Dongze, Hao,Yinghao, Cai,Yiping, Yang,et al. Learning Adaptively Context-Weight-Aware Correlation Filters for UAV Tracking with Robust Spatial-Temporal Regularization[C],2021. |
条目包含的文件 | ||||||
文件名称/大小 | 文献类型 | 版本类型 | 开放类型 | 使用许可 | ||
ICIGP2021-12.pdf(6102KB) | 会议论文 | 开放获取 | CC BY-NC-SA | 浏览 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论