Online growing neural gas for anomaly detection in changing surveillance scenes

作者：

Highlights：

• Online GNG employs neighbor-related strategies to adjust dominant learning parameters in a fully automated way.

• Online GNG can estimate the learning efficiency and adjust the network size to fit the unstable data space dynamically.

• Online GNG effectively reduces the false alarms and leak detections caused by model aging which frequently happens in changing surveillance scenes.

摘要

Highlights•Online GNG employs neighbor-related strategies to adjust dominant learning parameters in a fully automated way.•Online GNG can estimate the learning efficiency and adjust the network size to fit the unstable data space dynamically.•Online GNG effectively reduces the false alarms and leak detections caused by model aging which frequently happens in changing surveillance scenes.

论文关键词：Anomaly detection,Video surveillance,Unsupervised learning

论文评审过程：Received 20 October 2015, Revised 18 September 2016, Accepted 19 September 2016, Available online 10 October 2016, Version of Record 23 November 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.09.016