Three-array feature fusion crowd density estimation algorithm based on residual network

In order to solve the technical problem of multi-scale caused by the view angle of a camera in a crowd density estimation task, the invention provides a three-array feature fusion crowd density estimation algorithm based on a residual network, an improved ResNet34 network is used for extracting picture feature information, output is processed by using three columns of cavity convolution with different voidage rates, and the crowd density estimation efficiency is improved. The principle of the hole convolution is to perform interval zero supplementation operation on a convolution kernel on the basis of common convolution, so that a receptive field can be expanded under the condition of not increasing network parameters, and the capability of capturing image multi-scale information by the network is improved. And finally, the outputs of the three arrays are cascaded to obtain a high-quality density prediction map. According to the method, the influence of the density, the shielding property and the multi-scale