Real-Time Deformation Monitoring and Imaging for Long-Track SAR Based on GPU-Accelerated Parallel Multistream Processing

Track Synthetic Aperture Radar (SAR) provides an efficient, accurate, and stable approach for geological hazard early warning and structural health monitoring. However, high-resolution track SAR systems must process massive and complex data during large-scale, high-density deformation monitoring, which often leads to phase-wrapping errors between adjacent frames and inaccurate deformation estimation. Existing optimization methods typically rely on hardware performance or isolated improvements to individual algorithmic stages, without introducing parallel optimization at the architectural level. As a result, they fail to reduce the overall computational complexity of the algorithm. To address these limitations, this paper proposes a parallel multi-stream real-time processing and imaging framework. The proposed method adopts a two-stream strategy to optimize preprocessing tasks—such as range compensation and error correction—effectively mitigating system blocking caused by parameter dependencies. For the computationally intensive Stolt interpolation, a nested interpolation scheme combined with an outer kernel partitioning strategy is introduced to enhance interpolation efficiency and accuracy. Address mapping ensures synchronization and alignment among sub-domain data blocks. Experimental results demonstrate an acceleration factor of approximately 16, while preserving negligible amplitude and phase errors in the reconstructed images. These results confirm the effectiveness and robustness of the proposed method, providing a high-efficiency real-time imaging solution suitable for rapid disaster warning and on-site emergency response scenarios.