Multiclass Radio Frequency Interference Detection and Suppression for SAR Based on the Single Shot MultiBox Detector

Area of Science:

• Remote Sensing
• Signal Processing
• Artificial Intelligence

Background:

• Radio frequency interference (RFI) degrades synthetic aperture radar (SAR) imaging quality.
• RFI suppression in SAR echoes is a challenging blind source separation problem.

Purpose of the Study:

• To propose a novel method for multiclass RFI detection and suppression in SAR data.
• To improve SAR imaging quality by effectively mitigating RFI.

Main Methods:

• Utilized the Single Shot Multibox Detector (SSD) for RFI detection, identification, and estimation.
• Created a simulated echo-interference dataset using short-time Fourier transform (STFT).
• Reconstructed and mitigated RFI signals using adaptive filtering based on SSD predictions.

Main Results:

• The proposed method effectively detects, identifies, and estimates multiclass RFI.
• Successfully increased the signal-to-interference-noise ratio (SINR) of RFI-contaminated SAR echoes.
• Improved the peak sidelobe ratio (PSLR) after pulse compression.

Conclusions:

• The SSD-based approach offers an effective solution for RFI suppression in SAR.
• Validated through simulated experiments, demonstrating significant improvements in SAR data quality.