Random Hybrid Chirp Sensing Matrix Implementation for Fast Compressive Sensing SAR Processing

Abstract

Compressive sensing uses the sparsity of signals and the incoherence of sensing matrices. The use of random sensing matrices ensures an easy configuration and a high probability of reconstruction, but there is no optimum algorithm that can avoid the lengthy computation time and high memory consumption burden. Deterministic sensing matrix equations are known to mitigate these problems, and among others, chirp sensing matrices can help to achieve fast data recovery. However, most deterministic sensing matrices suffer from increased internal interference compared with that of random sensing matrix groups, and consequently result in degraded performance. In this paper, we propose a novel compressive sensing reconstruction method that enables the acquisition of excellent sparse signal reconstruction performance of existing random sensing matrices and signal processing acceleration performance through deterministic sensing matrices. Accordingly, we propose a method that contributes to the increase in the vast amount of data that has been a chronic problem with SAR(Synthetic Aperture Radar) images and the acceleration of the processing speed.

Compressed Sensing Vs. Auto-Encoder: On the Perspective of Signal Compression and Restoration

“…The sensing matrices include random sensing matrices and structured sensing matrices. The random sensing matrices mean a matrix with a random distribution, offering the advantage of ease to construct [34]. However, it requires a large amount of calculation and memory, above all, there is no efficient algorithm to verify the RIP condition, so it is not suitable for real applications.…”