Multi-scale Feature Fusion in Wireless Propagation Model Optimization Algorithms
Li Zheng, Jie Min, Chenxi Guo, Tianfeng Yan · 2021
In order to address the effects of various environmental variables on radio propagation and improve the quality of radio propagation. Inspired by the traditional empirical classification method and based on an artificial environment, this paper improves the experiments on the path well prediction results of the wireless propagation model according to the optimization algorithm with satellite remote sensing images as the input data of the network, and selects RESNet50 with the highest recognition efficiency by comparing the experimental data of eight deep learning networks such as VGG16 and RESNet50 under the self-built wireless transmission environment. By adding a feature pyramid to the network, the recognition accuracy of environmental variables reaches 96.4%, and the recognition time is 1.29 seconds per 1000 samples.