Real-time crowd stable state recognition method and device based on convolutional neural network

Abstract

The invention relates to a real-time crowd stable state recognition method and device based on a convolutional neural network. The real-time crowd stable state recognition method comprises the following steps: obtaining an input image, taking the input image as the input of a multi-column convolutional neural network model, and obtaining the number of crowds in a given grid region; performing image correction on the input image to obtain an actual area of the given grid region; obtaining a crowd density value of the given grid region based on the crowd number and the actual area; and identifying a crowd stable state of each given grid region based on the crowd density value, wherein the multi-column convolutional neural network model comprises a plurality of parallel convolutional neural networks with the same structure, and the sizes of convolution kernels of the convolutional neural networks are different, and the output of the convolutional neural networks generates a two-dimensional density map matrix through 1 * 1 filter mapping, and the number of people in a given grid region is obtained. Compared with the prior art, the real-time crowd stable state recognition method has theadvantages of high precision and the like.

Description

technical field [0001] The present invention relates to a crowd state information identification method and device, in particular to a convolutional neural network-based real-time crowd steady state identification method and device. Background technique [0002] Population stability analysis is a challenging research hotspot with important safety implications. Among them, crowd density is a direct and effective basis for analyzing crowd stability. With the improvement of the computing power of graphics processing units and machine deep learning capabilities, the convolutional neural network (CNN) in the deep learning system is more widely used in high-precision image processing. At present, the H.265 high-definition high-compression video technology of the increasingly popular video surveillance system (VSS) in public places effectively supports the real-time acquisition of high-definition images of crowd distribution. The convolutional neural network provides technical su...

AI Technical Summary

Problems solved by technology

[0003] So far, there are still some deficiencies in the analysis of crowd stability based on image processing: 1) The original image of the real-time video surveillance system has perspective distortion, which cannot be corrected in time, resulting in a large deviation in the estimation of crowd density

2) Lack of effective crowd stability analysis dynamic models and devices to timely determine the stability of floating crowds to assist crowd flow control

Images