Abstract: Based on back propagation neural network (BPNN), this paper establishes a channel model. The color information of red, green and blue channels (RGB) in satellite images is used to characterize the environmental characteristics of radio wave propagation path in wireless communication. We build the data set combining the color information and the distance between the measuring points and the base station and iteratively train the parameters of the net to predict the propagation path loss. The results given by the channel model show that the correlation coefficient between the predictions and the measured data reaches 0.83, absolute mean error is controlled at 0.66 dB, and standard deviation is controlled at 6.65 dB, which indicate that this model can reliably predict the propagation path loss of radio waves in wireless communication in the absence of a detailed model and material parameters of a certain scene. In the end, the model is compared with the traditional channel modeling method in many aspects. The results show that the model can provide prediction results that are slightly different from traditional channel modeling methods under the same computing resources, while greatly reducing the required time. The model can quickly predict propagation path loss in the optimization of wireless communication network system.