Adaptive model updating for robust object tracking

作者：

Highlights：

• We propose to use the features from hierarchies of CNNs as object appearance representations for visual tracking.

• The Hedge algorithm is utilized to predict the similarity between the new image patch and old one.

• Evaluation on the benchmark datasets demonstrates that the proposed tracking algorithm well handles various challenge scenarios and is comparable to state-of-the-art methods.

摘要

•We propose to use the features from hierarchies of CNNs as object appearance representations for visual tracking.•The Hedge algorithm is utilized to predict the similarity between the new image patch and old one.•Evaluation on the benchmark datasets demonstrates that the proposed tracking algorithm well handles various challenge scenarios and is comparable to state-of-the-art methods.

论文关键词：Hierarchical convolutional feature,Correlation filter,Object tracking,Adaptive model updating

论文评审过程：Received 31 January 2019, Revised 1 October 2019, Accepted 1 October 2019, Available online 17 October 2019, Version of Record 26 October 2019.

论文官网地址：https://doi.org/10.1016/j.image.2019.115656