Online anomaly detection in surveillance videos with asymptotic bound on false alarm rate

作者：

Highlights：

• Deep learning approaches to video anomaly detection lack performance analysis.

• Future frame prediction is an intuitive approach for detecting anomalies.

• Sequential decision making is suitable for quick and reliable anomaly detection.

• Online decision making is a critical factor in video surveillance.

摘要

•Deep learning approaches to video anomaly detection lack performance analysis.•Future frame prediction is an intuitive approach for detecting anomalies.•Sequential decision making is suitable for quick and reliable anomaly detection.•Online decision making is a critical factor in video surveillance.

论文关键词：Computer vision,Video surveillance,Anomaly detection,Asymptotic performance analysis,Deep learning,Online detection

论文评审过程：Received 13 June 2020, Revised 7 October 2020, Accepted 24 January 2021, Available online 1 February 2021, Version of Record 16 February 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.107865