Crowd Density Estimation Based on Multi-scale Feature Fusion and Information Enhancement

Abstract

Aiming at the problems such as diverse target scales and large-scale changes in crowds in dense crowd scenarios, a crowd density estimation method based on multi-scale feature fusion and information enhancement is proposed. Firstly, considering that small-scale targets account for a relatively large proportion in the image, based on the VGG-16 network, the dilated convolution module is introduced to mine the detailed information of the image. Secondly, in order to make full use of the multi-scale information of the target, a new context-aware module is constructed to extract the contrast features between different scales. Finally, considering the characteristic of continuous changes in the target scale, a multi-scale feature aggregation module is designed to enhance the sampling range of dense scales and multi-scale information interaction, thereby improving the network performance. Experiments on public datasets show that the proposed method in this paper can effectively estimate the population density compared with other advanced methods.