Crowd density estimation based on conditional random field and convolutional neural networks

Abstract

Crowd density estimation has an important value in the management of public safety. This paper presents an algorithm of combining conditional random field (CRF) model and convolution neural networks (CNN) to estimate crowd density. Firstly, the CRF model with higher-order potentials is used to extract foreground information and get a binary graph. Then using the original image to recover the information of binary graph. Finally, the multi-stage CNN was designed to obtain quality feature of foreground for crowd density estimation. In the experiment, we validate the effectiveness of the proposed algorithm on Chun-xi road and Nanjing train station data sets. Experimental results indicated that the proposed method has a good performance for crowd density estimation in different scene.