Guest Editorial: Synthetic aperture in sonar and radar

Abstract

Traditional sonar and radar both emerged during the early parts of the 20th century revolutionising the way we perceive and interpret the world around us, both above and under water. Particularly, synthetic aperture technologies have provided the tool for obtaining high-resolution images, namely synthetic aperture sonar (SAS) and synthetic aperture radar (SAR), which have opened the gates for a wide array of applications, ranging from military surveillance and environmental monitoring to archaeological exploration and disaster management.

Since the beginning of radar and sonar, it has been recognised that there are many areas of common interest. These include detection, classification, localisation and tracking of targets against a background of reverberation, noise or clutter, using either acoustic or electromagnetic energy. Over the past few decades there have been significant advances in both domains in the use of synthetic aperture imaging techniques—in radar for high resolution imaging from aircraft and satellites, defence surveillance purposes, geophysical and oceanographic remote sensing and environmental monitoring. In sonar it has been applied in high resolution imaging of objects on the seabed (including clutter) for the offshore industry and maritime mine countermeasures. Despite these common goals there has been very little cross-fertilization between the two scientific communities. This special issue is aimed at collecting scientific papers from both communities with the objective of contributing to increasing the exchange of knowledge between the two fields.

For this special issue, we received 10 papers, which underwent peer review. Papers were accepted or rejected based on the quality and fit with the special issue theme. Three of the five accepted papers focus on SAS, whereas the remaining two concern SAR.

The paper by Hansen and Sæbø presents a novel method for optimising the collection geometry for long-range synthetic aperture sonar interferometry (InSAS). As InSAS performance strongly depends on the collection geometry, the authors focus on determining the performance metrics and their dependence on geometrical parameters and then define a model and a procedure for optimising the overall performance. The theoretical work produced in this paper is well supported with evidence provided by real data.

The paper by Lane et al. shows how to implement target recognition and classification in SAR images with low-SWAP processing hardware. The authors utilise three different machine learning (ML)-based approaches to implement target detection and classification applied to SAR images, namely the RetinaNet, EfficientNet and Yolov5. The ML-based algorithms are trained by using a powerful cloud-based server, but they run on very low-SWAP devices, emulating their use in small and low-powered platforms. The authors make use of diverse types of SAR images to explore the algorithm effectiveness across various scenarios and SAR systems. The authors carried out a deep assessment of the proposed techniques to provide insights about which algorithm to prefer and in which case.

The paper by Olson and Geilhufe focuses on model selection techniques for seafloor scattering statistics in SAS images of complex seafloors. The problem of estimating seafloor scattering statistics has been studied for decades as it heavily affects the ability of systems to detect targets. In particular, it is important to determine the statistical model that produces the best data fit. The authors provide a thorough analysis of various model selection techniques and compare them against real data collected with the HISAS-1032. Results show how different model selections may produce significant changes in the system performances.

The paper by Hagelberg et al. explores a variety of bistatic and multistatic SAR configurations to highlight their impact on combined incoherent and coherent change detection (ICD/CCD). The authors define a set of metrics and produce a large amount of real data in a controlled environment (laboratory) to assess performances for various radar system configurations, including bistatic, multistatic and polarimetric. The authors show how performances can be improved when using multi-dimensional (multistatic and polarimetric) data over simple bistatic data. The authors also elaborate on the cost and limitations of using multidimensional SAR systems against simple monostatic and bistatic systems.

Finally, the paper by Steele and Lyons presents a completely new method for characterising seafloor sediments by using endfire SAS. The authors propose this new method to improve the performance of sediment volume characterisation by strongly reducing the biasing produced by interface roughness scattering caused by the large beamwidth of low-frequency sonars. By using endfire SAS (EF–SAS) a narrower beam can be produced by the sonar, therefore reducing such bias. The authors provide evidence of the effectiveness of their method by using real data.

All papers selected for this special issue presented concrete and innovative results and showed advancements in both SAS and SAR fields with the introduction and reuse of concepts and tools such as multidimensionality, machine/deep learning and endfire SAS to solve new and traditional problems in both fields.