SAMBA: A Scatter-Guided Masked Bidirectional Mamba Foundation Model for SAR Target Recognition

2026-06-30

AI Summary

This work addresses the challenges of synthetic aperture radar (SAR) automatic target recognition, which suffers from scarce labeled data, computationally intensive existing self-supervised methods, and neglect of SAR-specific electromagnetic scattering characteristics. To overcome these limitations, we propose SAMBA, an efficient self-supervised foundation model that integrates a linear-complexity bidirectional Mamba encoder, a physics-informed three-level scattering-guided masking strategy (SG-MAE) derived from SAR scattering priors, and a lightweight SpatialMix feature interaction module. The model is trained via a two-stage cross-domain pretraining scheme. Extensive experiments demonstrate that SAMBA achieves state-of-the-art performance across seven downstream classification and detection tasks while using significantly fewer parameters than CNN- and Transformer-based baselines. Notably, the SG-MAE strategy substantially enhances few-shot transferability.

Abstract

Synthetic aperture radar automatic target recognition (SAR ATR) is critical for Earth observation and defense, but its practical deployment is constrained by scarce annotated training data. Self-supervised pre-training alleviates this label bottleneck, yet prevailing Transformer architectures incur prohibitive quadratic computational complexity, and conventional universal masking neglects the unique electromagnetic scattering properties intrinsic to SAR imagery. To address these limitations, we propose SAMBA (Scattering-Guided Bidirectional Mamba), an efficient self-supervised pre-training foundation model for SAR target interpretation. Our framework features three core innovations: (i) a linear-complexity Mamba encoder with a mid-sequence class token to mitigate computational bottlenecks; (ii) a three-level hierarchical Scattering-Guided Masked Autoencoder (SG-MAE) masking strategy guided by SAR physical priors, aligning the pretext task with SAR's intrinsic imaging mechanism; (iii) a lightweight SpatialMix feature interaction module to enhance cross-region feature fusion. We also design a two-stage cross-domain pre-training pipeline to optimize the overall pre-training process. Extensive evaluations demonstrate that SAMBA consistently delivers superior performance across all pre-training configurations, with substantially fewer parameters than both CNN and Transformer baselines. Compared with the default masking strategy in standard MAE, the proposed SG-MAE strategy further boosts the model's few-shot transfer capability. Benchmarking on seven downstream datasets covering classification and detection tasks shows SAMBA achieves state-of-the-art (SOTA) performance on most metrics, fully validating its robust generalizability across diverse SAR interpretation tasks. Source code and pre-trained weights are publicly available at https://github.com/mynswkk/SAMBA.

Problem

• SAR ATR
• self-supervised pre-training
• electromagnetic scattering
• data scarcity
• computational complexity

Innovation

• Mamba
• Scattering-Guided Masking
• Self-Supervised Pre-training
• SAR ATR
• Linear Complexity