Learning spatio-temporal context via hierarchical features for visual tracking
作者：
Highlights：
•
• An improved approach to Spatio-temporal context based visual tracking algorithm.
• A mapping neural network is used to acquire dynamic training confidence map.
• Hierarchical features is exploited for the construction of context prior models.
• Training confidence index is resorted to guide updating process.
• Experiments of both ordinary and aerial tracking show excellent tracking results.
摘要
•An improved approach to Spatio-temporal context based visual tracking algorithm.•A mapping neural network is used to acquire dynamic training confidence map.•Hierarchical features is exploited for the construction of context prior models.•Training confidence index is resorted to guide updating process.•Experiments of both ordinary and aerial tracking show excellent tracking results.
论文关键词：Visual tracking,Convolutional neural network,Transfer learning,Spatio-temporal context,Dynamic training confidence map,Training confidence index
论文评审过程：Received 4 November 2017, Revised 17 April 2018, Accepted 17 April 2018, Available online 25 April 2018, Version of Record 14 May 2018.
论文官网地址：https://doi.org/10.1016/j.image.2018.04.010
