Research On Crowd Flow Statistics System Based On Deep Learning

The crowd flow statistics algorithm is designed to automatically count the number of pedestrians in the image or video and to obtain a density map that reflects the population density distribution in the image.By statistics of the number and density distribution of the population in the picture,we can optimize the road design and the building distribution in some places where people are congested,so that the urban traffic is more intelligent and we can avoid trampling accidents.The important and practical value of automatic statistics of population flow makes it a key topic for many researchers.In this paper,a population flow statistics algorithm based on deep learning convolutional neural network is proposed based on the research of population flow statistics method.Based on this,a population automatic traffic statistics system is designed and implemented.The main work of this paper is as follows:(1)Two convolutional neural network structures are proposed to classify crowd images based on crowd density.Thus,the global features and local features are introduced into the network structure of crowd flow statistics algorithm,which improves the performance of the algorithm.(2)A crowd flow statistics method based on convolutional neural network is proposed,which combines the multi-scale features,global features and local features in the crowd images and generates a prediction density map that can reflect the population density distribution in the images.(3)Using the Django framework to implement a web-based crowd flow statistics system.The user can accurately predict the crowd density at the corresponding position in the image and the number of pedestrians in the image by uploading the corresponding image or video of the crowd in the web front end interface.The experimental results in this paper show that the proposed convolutional neural 'network structure combining global features,local features and multi-scale features in the image has higher levels of performance than the existing models in public datasets.The implementation of the crowd flow statistics system meets the demand of people to automatically analyze the number of pedestrians and crowd trends in image and video files.