AI for Performance-Optimized Quantization in Future SAR Systems

Inhalt:
Next-generation SAR systems will be capable of high-resolution wide-swath acquisitions, which will inevitably result in a significant increase of the onboard data volume to be acquired by the system. This, in turn, will lead to severe constraints in terms of onboard memory requirements and downlink capacity. In this context, the onboard quantization of SAR raw data represents an aspect of crucial importance, since it acts as a trade-off between achievable product quality and resulting on-board data volume. In this paper, we investigate the use of artificial intelligence (AI), and in particular of deep learning (DL), for flexible on-board SAR raw data quantization, with the aim of deriving an optimized and adaptive data rate allocation given a desired performance metric and requirements in the resulting focused SAR/InSAR products without relying on a priori information on the acquired scene. The derived bitrate maps (BRMs) can then be used for adapting a BAQ quantizer to the local characteristics of the input data and to the desired output performance. Different performance parameters can be used, such as the Signal-to-Quantization Noise Ratio (SQNR), the InSAR coherence loss or the resulting interferometric phase error, extending the capabilities of the architecture and, ideally, providing multiple bitrate estimations for a single input scene, depending on the specific application requirement. In view of a potential on-board implementation, a possible hardware architecture for the proposed compression scheme is presented as well.