Research of Crowd Density Estimation Algorithm Based on CAN

Population estimation refers to the use of various methods and tools to estimate the number of people in a specific area or location, which has broad significance in multiple fields. It helps to better understand population behavior, make wise decisions, and prevent emergency situations. This article introduces a crowd density estimation algorithm based on an improved Context AwareNet (CAN). Using the classic vgg-16 network structure and the ShanghaiTech dataset, the function of estimating the number of people in the target image is achieved. After the training and testing of the network model, it is found that the effect of void convolution is not as good as that of ordinary convolution, and the model has serious errors for low-density test images. By optimizing the parameters, improving the network structure, and modifying the data set, two crowd density estimation models for different types of people are generated. By combining the two models, the identification accuracy of low density people is improved and the generalization ability of the algorithm is enhanced without affecting the accuracy rate of the algorithm.