Algorithm Research And System Design Of Crowd Density Estimation Based On Deep Learning

In the process of building intelligent video surveillance systems,the analysis of the crowd is a very meaningful project.With the unprecedented breakthrough in the field of computer vision,a lot of research work shows that the image features of neural network learning have more generalization ability and representativeness than traditional image features.Based on the actual work situation and social needs,this paper uses the deep learning method to study the population density estimation problem.The specific work content is as follows:?1?A multi-scale feature fusion population density estimation algorithm based on deep learning is proposed.The algorithm uses the method based on population density-map regression to learn the mapping relationship between crowd image and population density distribution map through convolutional neural network.Due to the existence of perspective distortion,the scale corresponding to pedestrians in different positions in the crowd image also changes.This paper designs a deep single-column convolutional neural network structure,uses different levels of feature extraction modules to extract pedestrian features,and then fuses different levels of feature maps to solve the scale change problem in crowd images.Finally,after fusion The feature map is regressed to generate a population density map,and the population density map is integrated to obtain the total number of people in the crowd image.?2?The convolutional neural network model proposed in this paper is trained and tested in the crowd dataset ShanghaiTech and UCF_CC₅0,and the dataset is also augmented.Different algorithms are used on the basis of the original dataset.The image is transformed to different degrees,so that the neural network can adapt to different application scenarios and enhance the robustness of the neural network.Finally,the experimental results are compared with some current crowd density estimation algorithms.The experimental results show that the proposed algorithm has better performance in the accuracy and robustness of the estimation results.?3?A simple and feasible intelligent crowd density estimation system is designed.The system consists of real-time image acquisition module,crowd density estimation module and alarm module.It can monitor the scene density of the scene captured by the camera in real time,and can start the alarm in time when the population density reaches the peak.