Crowd density and quantity estimation method based on convolutional neural network

Abstract

The invention discloses a crowd density and quantity estimation method based on a convolutional neural network, and the method comprises the steps: firstly collecting a scene image, and marking the head position in the scene image as a training image set; secondly, generating a real crowd density distribution diagram for training according to the training images and the head marks thereof; then, building a convolutional neural network to return to the crowd density distribution diagram, calculating a loss function Loss, adjusting the network weight through the loss function by means of a stochastic gradient descent method, and training is ended when the model converges; and finally, inputting a to-be-predicted image into the trained convolutional neural network to obtain a predicted crowddensity distribution map, and performing summation operation on the whole crowd density distribution map to obtain a predicted total number of people. Compared with other existing methods, the methodhas the advantages that under the complex conditions of denseness, shielding, different visual angles and the like, the crowd counting accuracy can be improved, and public safety prevention and control are effectively enhanced.

Description

technical field [0001] The invention belongs to the field of machine vision applications, in particular to a method for estimating crowd density and quantity based on a convolutional neural network. Background technique [0002] Crowd counting in complex scenes is currently a research hotspot and difficulty in the industry and academia, and it has important application value in real life. Crowd counting is widely used in video surveillance, traffic monitoring, public safety, urban planning, and the construction of smart supermarkets, such as monitoring the number of people in an area where people tend to gather, and preventing crowds from getting out of control and stampede due to excessive crowd density. event. Due to complex situations such as denseness, occlusion, and different viewing angles, crowd counting in real scenes is still an unsolved problem. [0003] As a commonly used means of public safety prevention and control, machine vision is used to quantitatively mea...

AI Technical Summary

Problems solved by technology

The advantage is that it can accurately locate pedestrians or heads, but the disadvantage is that for high-density crowd images, the detection accuracy is poor

Images