Deep SAR Imaging and Motion Compensation.

Area of Science:

• Remote Sensing
• Signal Processing
• Artificial Intelligence

Background:

• Traditional Synthetic Aperture Radar (SAR) imaging uses compressive sensing (CS) and matrix sensing (MS) to reduce sampling by leveraging sparse or low-rank properties.
• Existing methods may not fully exploit the inherent redundancy in SAR backscattering coefficients, limiting sampling efficiency.

Purpose of the Study:

• To propose a novel deep SAR imaging algorithm that exploits backscattering coefficient redundancy for improved sampling efficiency.
• To develop a deep SAR motion compensation algorithm to address practical application challenges.

Main Methods:

• A deep learning approach utilizing an auto-encoder structure to exploit redundancy in the backscattering coefficient.
• Simultaneous estimation of auto-encoder parameters and backscattering coefficient by optimizing reconstruction loss on down-sampled SAR echo.
• Development of a deep SAR motion compensation algorithm to mitigate motion error effects.

Main Results:

• The proposed deep SAR imaging algorithm effectively utilizes backscattering coefficient redundancy.
• The auto-encoder structure with a lower-dimensional latent layer reduces parameters and improves efficiency.
• The deep motion compensation algorithm successfully eliminates motion error impacts on imaging results.
• Validation of both algorithms on simulated and real SAR data.

Conclusions:

• The novel deep learning approach significantly enhances SAR imaging sampling efficiency by exploiting data redundancy.
• The integrated deep motion compensation algorithm addresses practical requirements for real-world SAR applications.
• The proposed methods offer a promising advancement for SAR imaging technology.