Crowd density estimation method and system based on content awareness module

Abstract

The invention provides a crowd density estimation method and system based on a content perception module, and relates to the technical field of crowd density estimation, and the method comprises the steps: carrying out the preprocessing of to-be-predicted image data; extracting deep semantic information and multi-layer fusion context information according to the processed image data; dynamically generating content perception parameters according to the deep semantic information based on a content perception module; initializing the weight and offset of a crowd density estimation module convolution layer according to the content perception parameters; and inputting the multilayer fusion context information into the initialized crowd density estimation module to obtain a crowd density image. When the crowd density is predicted each time, the deep semantic information of the picture data is obtained, and then the parameters of the convolution layer of the crowd density estimation module are dynamically generated, so that crowd density estimation is carried out more flexibly and specifically, and a better prediction result can be obtained in a complex and changeable scene.

Description

technical field [0001] The invention relates to the technical field of crowd density estimation, in particular to a crowd density estimation method and system based on a content-aware module. Background technique [0002] In recent years, the human population has increased and the crowd density in public places has increased significantly. Stampede accidents caused by excessive crowd density have occurred all over the world, causing huge personal and property losses. With the rapid development of technology in the field of computer vision, crowd counting tasks based on surveillance images and videos play a vital role in security and traffic control. At the same time, the changing environment, shooting angles, and occlusion of dense crowds have brought huge challenges to this task. [0003] With the development of deep learning and convolutional neural network (CNN), researchers began to use CNN to accurately estimate the number of crowds from images and videos. Most of the...

AI Technical Summary

Problems solved by technology

However, the networks currently used in the field of crowds are not flexible enough to load fixed parameters during prediction, and it is difficult to deal with the characteristics of changing crowd image shooting angles, irregular crowd distribution, perspective and occlusion

Images