Statistical bootstrap-based principal mode component analysis for dynamic background subtraction
作者：
Highlights：
• We propose a novel methodology to extract foreground objects from a video sequence with dynamic background with highly repetitive patterns.
• We propose a relaxed version of the statistical mode (which was well defined for categorical data only) that is applicable to video data.
• We incorporate the relaxed version of the statistical mode to principal component analysis and apply the statistical bootstrapping technique to capture the repetitive patterns of the dynamic background.
• We propose an effective optimization procedure that can quickly find the global optimal solution of the new statistical model.
• Experiment results show that the proposed method has superiority performance to 10 other different methods for 16 different real-world video sequences.
摘要
•We propose a novel methodology to extract foreground objects from a video sequence with dynamic background with highly repetitive patterns.•We propose a relaxed version of the statistical mode (which was well defined for categorical data only) that is applicable to video data.•We incorporate the relaxed version of the statistical mode to principal component analysis and apply the statistical bootstrapping technique to capture the repetitive patterns of the dynamic background.•We propose an effective optimization procedure that can quickly find the global optimal solution of the new statistical model.•Experiment results show that the proposed method has superiority performance to 10 other different methods for 16 different real-world video sequences.
论文关键词：Background modeling,Video surveillance,Principal Component analysis,Statistical mode
论文评审过程：Received 20 June 2019, Revised 2 October 2019, Accepted 8 December 2019, Available online 16 December 2019, Version of Record 19 December 2019.
论文官网地址：https://doi.org/10.1016/j.patcog.2019.107153
