Ultra-High Compression and Rapid Reconstruction of DAS Data Based on Adaptive Compressed Reconstruction Algorithm

Distributed acoustic sensing (DAS) systems face critical challenges in data transmission, storage, and real-time processing due to the extremely high density of signal acquisition in large-scale deployments. Conventional compressed sensing methods often rely on manually predefined parameters, exhibit poor robustness under complex noise conditions, and suffer from low reconstruction accuracy and high computational cost. To overcome these limitations, this study proposes an adaptive compressive sampling pursuit (ACoSaMP) algorithm that dynamically estimates signal sparsity from the frequency-band energy distribution, thereby eliminating manual tuning. Integrated into the CoSaMP framework, the proposed approach reduces computational complexity while enhancing compression efficiency. Experimental validation in both laboratory experiments and real-world scenarios involving three representative external disturbances (rammer operation, manual excavation, and vehicle/subway vibration) demonstrates that ACoSaMP achieves ultra-low compression ratios of 1.1%, reconstruction times of 25 ms, 85 ms, and 81 ms, and Pearson correlation coefficients (PCC) consistently above 0.9. Compared with orthogonal matching pursuit (OMP), ACoSaMP improves compression efficiency by a factor of 19.2 and accelerates reconstruction speed by 18.5 times. In addition, support vector machine (SVM) classification of the reconstructed signals achieves 95.3% accuracy, exceeding that obtained with uncompressed signals by 5.5%.