An Efficient Echo Reconstruction Method by Physics-Driven Deep Learning

This study presents an innovative deep learning method, the physics-driven dual-channel complex convolutional neural network (PDCConvNet), to address the challenges of radar target echo reconstruction. Traditional scattering center (SC) extraction techniques are often hindered by high computational complexity and slow processing speeds, particularly in multi-parameter inversion tasks. The proposed method incorporates three key designs. First, a signal denoising and separation module is employed to isolate SCs while maintaining a fixed structure for the network's output nodes, ensuring that each sub-echo is associated with a unique set of SC parameters. Second, a dual-channel complex convolutional structure is introduced to independently extract the amplitude and phase features of the input echo, ensuring the preservation and propagation of complete scattering information throughout the network. Finally, the mathematical formulation of the SC model is embedded into the network's loss architecture, guiding the optimization of network parameters and improving SC extraction accuracy. Numerical results demonstrate that the proposed approach significantly outperforms traditional optimization techniques in terms of echo reconstruction efficiency, offering a more effective solution for radar echo reconstruction.