Robust people counting using sparse representation and random projection

作者：

Highlights：

• We propose a robust and scalable people counting method based on sparse representation.

• The complexity of solving l1-minimization is significantly reduced by using random projection.

• The features obtained from pre-trained deep Convolutional neural network are exploited in people counting task.

• A semi-supervised elastic net is employed to automatically annotate unlabelled data with only a handful of frames.

摘要

Highlights•We propose a robust and scalable people counting method based on sparse representation.•The complexity of solving l1-minimization is significantly reduced by using random projection.•The features obtained from pre-trained deep Convolutional neural network are exploited in people counting task.•A semi-supervised elastic net is employed to automatically annotate unlabelled data with only a handful of frames.

论文关键词：People counting,Sparse representation,Fast l1-minimization,Random projection,Convolutional neural network,Semi-supervised learning

论文评审过程：Received 29 September 2014, Revised 16 January 2015, Accepted 13 February 2015, Available online 5 March 2015, Version of Record 17 June 2015.

论文官网地址：https://doi.org/10.1016/j.patcog.2015.02.009