Dynamic prioritization of surveillance video data in real-time automated detection systems
作者：
Highlights：
• Effective object detection algorithms for surveillance are computationally expensive.
• Sub-sampling surveillance data can reduce computation while maintaining performance.
• Methods are evaluated on three multi-camera datasets with durations >25 minutes.
• This work presents a dynamic prioritization method that fuses spatiotemporal features.
• Object detection rate increases by up to 60% versus the static subsampling baseline.
摘要
•Effective object detection algorithms for surveillance are computationally expensive.•Sub-sampling surveillance data can reduce computation while maintaining performance.•Methods are evaluated on three multi-camera datasets with durations >25 minutes.•This work presents a dynamic prioritization method that fuses spatiotemporal features.•Object detection rate increases by up to 60% versus the static subsampling baseline.
论文关键词：Video surveillance,Computer vision,Real-time systems,Object detection
论文评审过程：Received 6 February 2020, Revised 14 June 2020, Accepted 16 June 2020, Available online 2 July 2020, Version of Record 8 July 2020.
论文官网地址：https://doi.org/10.1016/j.eswa.2020.113672
