Unsupervised flow-based motion analysis for an autonomous moving system

作者：

Highlights：

• This paper focuses on segmenting the motion from dense optical flow fields.

• Two unsupervised clustering methods are presented and a model selection is proposed.

• A comparison between the proposed techniques with the K-means and EM is made.

• Experiments are conducted in a surveillance scenario with an autonomous mobile.

• The proposed techniques are superior in terms of robustness and computational demands

摘要

•This paper focuses on segmenting the motion from dense optical flow fields.•Two unsupervised clustering methods are presented and a model selection is proposed.•A comparison between the proposed techniques with the K-means and EM is made.•Experiments are conducted in a surveillance scenario with an autonomous mobile.•The proposed techniques are superior in terms of robustness and computational demands

论文关键词：Motion segmentation,Optical flow,Moving observer,Active surveillance,Mobile robot

论文评审过程：Received 28 June 2013, Revised 14 February 2014, Accepted 2 April 2014, Available online 13 April 2014.

论文官网地址：https://doi.org/10.1016/j.imavis.2014.04.003