Radar signal sorting based on image semantic segmentation

This paper explores the application of image semantic segmentation networks in radar signal sorting. Unlike traditional methods, it does not depend on predefined parameters or prior information, so it can show enhanced adaptability in dealing with a more complex electromagnetic environment. Firstly, the pulse descriptor PDW of collected signals is encoded into corresponding sequence images, and then U-net is used to train pulse sequence images. The trained model can be applied with high accuracy in radar signal sorting. At the same time, an attention mechanism is introduced into the U-net, which improves the recognition rate of signal features.