Crowd counting method based on multi-level feature fusion

Abstract

The invention relates to a crowd counting method based on multi-level feature fusion. The method comprises the steps: preprocessing an acquired crowd image, generating a corresponding crowd density map by using the annotation information, constructing a multi-level feature fusion crowd counting network, initializing network weight parameters, and inputting a preprocessed crowd image and a crowd density map into the network; completing forward propagation, calculating loss of a forward propagation result and a real density map, updating model parameters, iterating forward propagation and updating the model parameters to a specified number of times, obtaining the crowd density map, and obtaining an estimated number of people. According to the method provided by the invention, the problem ofcrowd scale change in the crowd counting task can be solved, and crowd counting is more accurate.

Description

technical field [0001] The invention relates to the field of image crowd counting and the field of deep learning, in particular to a crowd counting method based on deep learning. Background technique [0002] Crowd counting is an important problem in the field of image processing and computer vision. Its goal is to automatically generate crowd density maps from crowd images and estimate the number of people in the scene. Crowd counting is widely used in traffic dispatching, security prevention and control, urban management and other fields. [0003] Traditional crowd counting methods require complex preprocessing of crowd images, and require manual design and extraction of human body features. In the case of cross-scenes, features need to be re-extracted, which has poor adaptability. In recent years, the successful application of convolutional neural networks has brought major breakthroughs to crowd counting tasks. Zhang[1] et al. proposed a convolutional neural network mo...

AI Technical Summary

Problems solved by technology

[0009] In order to solve the problem of crowd size differences in different scenarios in the prior art, the present invention proposes a crowd counting method based on multi-level feature fusion

Images