Synthetic Aperture Sonar Systems Research Articles

Platform motion estimation in multi-band synthetic aperture sonar with coupled variational autoencoders.

Coherent processing in synthetic aperture sonar (SAS) requires platform motion estimation and compensation with sub-wavelength accuracy for high-resolution imaging. Micronavigation, i.e., through-the-sensor platform motion estimation, is essential when positioning information from navigational instruments is absent or inadequately accurate. A machine learning method based on variational Bayesian inference has been proposed for unsupervised data-driven micronavigation. Herein, the multiple-input multiple-output arrangement of a multi-band SAS system is exploited and combined with a hierarchical variational inference scheme, which self-supervises the learning of platform motion and results in improved micronavigation accuracy.

Platform motion estimation in multiple-input multiple-output synthetic aperture sonar with coupled variational autoencoders

Synthetic aperture sonar (SAS) utilizes the motion of the platform carrying the sonar system to synthesize an aperture that is much longer than thephysical antenna by coherently combining data from several pings. Coherent processing in SAS requires platform motion estimation and compensation with sub-wavelength accuracy for high-resolution imaging. Micronavigation, i.e., through-the-sensor platform motion estimation from spatio-temporal coherence measurements of diffuse backscatter on overlapping recordings between successive pings, is essential when positioning information from navigational instruments is absent or inadequately accurate. Representation learning with a variational autoencoder (VAE) offers an unsupervised data-driven micronavigation solution. In this study, we introduce a hierarchical variational model implemented with coupled VAEs to relate the common latent features between datasets of coherence measurements in broadband multiple-input multiple-output SAS systems. We show that self-supervising the training process of independently parameterized but coupled VAEs improves significantly the accuracy of the micronavigation estimates.

Three-dimensional observations of tidal plume fronts in estuaries using a synthetic aperture sonar array.

Synthetic aperture sonar (sas) systems are designed to observe stationary scatterers located near the sediment interface. Less commonly, a sas system may be used to observe scattering features located above the sonar in the water column. The Undersea Remote Sensing (USRS) project, sponsored by the Office of Naval Research, was a collaborative Directed Research Initiative (DRI) focused on studying dynamic estuarine water column features. During the USRS DRI, researchers from multiple institutions gathered to observe tidal features at various estuaries along the coast of the United States using both in situ and remote sensing techniques, including sas. The first studied estuary was the mouth of the Connecticut River (CTR). Data captured by a sas system deployed during a tidal event were post-processed to create three-dimensional observations of the structure of the leading edge of the CTR's ebb plume front. From these observations, lobed structures similar in scale to previously reported instabilities are revealed, with the present observations providing additional insight regarding the structure of the bubble distribution behind the front. Velocity estimates of plume features were also determined from sas data and shown to compare favorably with concurrent marine radar estimates.

A Novel Imaging Algorithm for Wide-Beam Multiple-Receiver Synthetic Aperture Sonar Systems

In existing imaging algorithms for wide-beam multiple-receiver synthetic aperture sonar (SAS) systems, the double-square-root (DSR) range history of each receiver is generally converted into the sum of a single-square-root (SSR) range history and an error term using displaced phase center aperture (DPCA) approximation. Therefore, before imaging, each receiver’s error term needs to be individually compensated in the azimuth frequency domain, which is computationally expensive. As a result, a novel wide-beam multiple-receiver SAS system algorithm with low complexity and high precision is suggested. First, the translation relationship between the range histories of the reference receiver and other receivers is used to derive an SSR approximation range history that takes into account the azimuth variance of the non-stop-hop-stop time while ignoring differential range curvature (DRC) between the range histories from different receivers. Then, using the principle of stationary phase (POSP), the two-dimensional (2-D) spectrum of the point target is obtained. Finally, the multiple-receiver data are transformed into monostatic SAS-equivalent data for imaging after phase correction, time delay correction, and azimuth reconstruction. The range-Doppler (RD) algorithm is used as an example to explain the specific steps of the proposed approach. Simulation data and ChinSAS data experiments verify that the proposed algorithm achieves an imaging performance that is comparable to that of the existing wide-beam algorithm, but with much higher computational efficiency, making it suitable for real-time imaging.

Guest editorial: recent advances in synthetic aperture sonar technology

Abstract[Correction added on 28‐July‐2023, after first online publication: The first affiliation has been changed and bio for Gary Heald has been added in this version.]Compared to traditional sidescan sonar, synthetic aperture sonar (SAS) or Interferometric synthetic aperture sonar (InSAS) systems can provide much higher resolution images across the whole swath. Due to the high resolution and fast mapping rate at a given speed, SAS/InSAS technology is gaining increasing interest within the field of underwater engineering. Navigation, autonomous underwater vehicles (AUVs), sonar technolog and electronics have achieved great progress in recent years. Thanks to these developments, the SAS/InSAS has also been pushed into a new stage. Firstly, the lightweight and compact SAS can be mounted on ever smaller, highly flexible platforms. Secondly, current algorithms are computationally more efficient. Thirdly, the SAS/InSAS images are further improved based on various methods. Last but not least, SAS/InSAS technology is preferred by underwater mapping users due to its high resolution. The current Special Issue is focused on research ideas, articles and experimental studies related to “Recent Advances in Synthetic Aperture Sonar Technology” for development, operating mode, algorithms and applications the various aspects of SAS/InSAS in applications.

Interferometric synthetic aperture sonar large scale simulated data benchmark of a ship wreck

AbstractInterferometric synthetic aperture sonar (InSAS) systems produce high resolution seafloor imagery and 3‐D bathymetry by exploiting the phase differences between image pairs formed from data collected by vertically separated arrays. This process is non‐trivial due to the inherent ambiguity of interferometric phase estimates, and the effect of layover, where the response of multiple scatterers with different positions are mapped onto the same pixel during imaging. These effects are particularly problematic for complex scenes with strong bathymetry, such as ship wrecks. This is an unsolved problem and improved algorithms for InSAS bathymetry estimation are needed. However, developing and assessing such algorithms using data from fielded InSAS systems is difficult because of a lack of accurate ground truth for the platform motion and the true 3‐D shape of the scene. This paper presents an openly available synthetic data set as an alternative to field‐collected data for use by InSAS algorithm developers. The dataset is derived from a high‐resolution 3‐D photogrammetry scan of the SS Thistlegorm wreck. It was produced by a point‐based diffraction model using parameters similar to the Minehunting UUV for Shallow water Covert Littoral Expeditions (MUSCLE) InSAS.

Approximate extraction of late-time returns via morphological component analysis

A fundamental challenge in acoustic data processing is to separate a measured time series into relevant phenomenological components. A given measurement is typically assumed to be an additive mixture of myriad signals plus noise whose separation forms an ill-posed inverse problem. In the setting of sensing elastic objects using active sonar, we wish to separate the early-time return from the object's geometry from late-time returns caused by elastic or compressional wave coupling. Under the framework of morphological component analysis (MCA), we compare two separation models using the short-duration and long-duration responses as a proxy for early-time and late-time returns. Results are computed for a broadside response using Stanton's elastic cylinder model as well as on experimental data taken from an in-air circular synthetic aperture sonar system, whose separated time series are formed into imagery. We find that MCA can be used to separate early and late-time responses in both the analytic and experimental cases without the use of time-gating. The separation process is demonstrated to be compatible with image reconstruction. The best separation results are obtained with a flexible, but computationally intensive, frame based signal model, while a faster Fourier transform based method is shown to have competitive performance.

Approximate extraction of late-time returns via morphological component analysis

A fundamental challenge in acoustic data processing is to separate a measured time series into relevant phenomenological components. In the setting of sensing elastic objects using active sonar, we wish to separate the early-time returns (e.g., returns from the object's exterior geometry) from late-time returns caused by elastic or compressional wave coupling. Under the framework of morphological component analysis (MCA), we compare two separation models using the short-duration and long-duration responses as a proxy for early-time and late-time returns. Results are computed for broadside geometries using Stanton's elastic cylinder model as well as experimental data taken from an in-air circular synthetic aperture sonar (AirSAS) system, whose separated time series are formed into imagery. We find that MCA can be used to separate early and late-time responses in both the analytic and experimental cases without the use of time-gating. The separation process is demonstrated to be robust to noise and compatible with AirSAS image reconstruction. The best separation results are obtained with a flexible, but computationally intensive, frame based signal model, while a faster Fourier Transform based method is shown to have competitive performance.

Modeling the effect of random roughness on synthetic aperture sonar image statistics.

A model has been developed to predict the effect of random seafloor roughness on synthetic aperture sonar (SAS) image statistics, based on the composite roughness approximation-a physical scattering model. The continuous variation in scattering strength produced by a random slope field is treated as an intensity scaling on the image speckle produced by the coherent SAS imaging process. Changes in image statistics caused by roughness are quantified in terms of the scintillation index (SI). Factors influencing the SI include the seafloor slope variance, geo-acoustic properties of the seafloor, the probability density function describing the speckle, and the signal-to-noise ratio. Example model-data comparisons are shown for SAS images taken at three different sites using three different high-frequency SAS systems. Agreement between the modeled and measured SI show that it is possible to link range-dependent image statistics to measurable geo-acoustic properties, providing the foundation necessary for solving problems related to the detection of targets using high-frequency imaging sonars, including performance prediction or adaptation of automated detection algorithms. Additionally, this work illustrates the possible use of SAS systems for remote sensing of roughness parameters such as root mean square slope or height.

The Study of Selecting a Test Area for Validating the Proposal Specification of InSAS(Interferometric Synthetic Aperture Sonar)

This paper provides a case study of development testing and evaluation of design goal of Interferometric SAS (Synthetic Aperture Sonar) system that is developing supported by Civil-Military Technology Cooperation Center in offshore fields. For Deep water operating capabilities evaluation, We have surveyed candidate field, bathymetric mapping and target identification over 200 m depth, East Sea. In testing phase, We have provided environmental information of testing field include water column, seabed and weather condition in real time. And to compare excellency of developing InSAS, we have gather commercial imaging sonar system data with same target. This case study will support the Test Readiness Review of future underwater surveillance system developing via investigate marine testing field environment, testing facilities and planning.

Three-Dimensional Imaging of Circular Array Synthetic Aperture Sonar for Unmanned Surface Vehicle

Synthetic aperture sonar (SAS) and interferometric synthetic aperture sonar (InSAS) have a range layover phenomenon during underwater observation, the AUV-mounted circular synthetic aperture sonar (CSAS) system, that insonifies targets using multiple circular scans that vary in height and can eliminate the layover phenomenon. However, this observation method is time-consuming and difficult to compensate. To solve this problem, the circular array synthetic aperture sonar (CASAS) based on the equivalent phase center was established for unmanned surface vehicles. Corresponding to the echo signal model of circular array synthetic aperture sonar, a novel three-dimensional imaging algorithm was derived. Firstly, the echo datacube was processed by signal calibration with near-field to far-field transformation and grid interpolation. Then, the sparse recover method was adopted to achieve the scattering coefficient in the height direction by sparse Bayesian learning. Thirdly, the Fourier slice theorem was adopted to obtain the 2D image of the ground plane. After the reconstruction of all height slice cells was accomplished, the final 3D image was obtained. Numerical simulations and experiments using the USV-mounted CASAS system were performed. The imaging results verify the effectiveness of the 3D imaging algorithm for the proposed model and validate the feasibility of CASAS applied in underwater target imaging and detection.

Efficient imaging method for multireceiver SAS

With the Loffeld's bistatic formula, the point target reference spectrum (PTRS) of multireceiver synthetic aperture sonar (SAS) can be decomposed into the quasi monostatic term and bistatic deformation (BD) term. The former term is analogous to the monostatic SAS PTRS while the latter one describes the bistatic phase introduced by the bistatic displacement between the transmitter and receiver. The multireceiver SAS echo signal can be transformed into monostatic SAS equivalent data after correcting the BD term. Based on the range sub-block processing method, this data transformation is conducted by using the preprocessing step. Then, fast imaging processors based on monostatic SAS system can be directly exploited to process the monostatic SAS equivalent data. In this paper, the authors mainly focus on the application of non-linear chirp scaling algorithm. At last, the simulations and real data are used to validate the presented method. The processing results show that the imaging performance of presented method is nearly close to that of traditional method. Besides, the imaging efficiency is highly improved compared to the phase centre approximation (PCA) method.

Model validation for simulated synthetic aperture sonar time series data

Synthetic aperture sonar (SAS) systems have grown in popularity due to their ability to create high-resolution sonar imagery with a high area coverage rate. Researchers continue to find new uses for SAS data products. To support the development of new uses and algorithms, it is important to be able to accurately simulate SAS systems with synthetic time series data. This simulated data can be used for many purposes, including estimating system signal-to-noise ratio and performance in a variety of environments, testing signal processing algorithms, and training algorithms to recognize objects in SAS imagery. For these processing tasks to be successful, the simulator must accurately represent physical phenomena in the oceanic underwater environment, which can be extremely complex. This presentation will outline a process for validation and verification of a high-frequency SAS time series simulator, and discuss several useful tests for validating model fidelity. Two fundamental concepts will be explored in depth: the sonar equation and spatial coherence theory. The sonar equation is important because it accounts for correct implementation of multiple lower level models such as geometry, beam patterns, propagation, and seafloor scattering strength. Spatial coherence is important because it is relied on to form coherent imagery.

Unsupervised learning of platform motion in synthetic aperture sonar.

Synthetic aperture sonar (SAS) provides high-resolution acoustic imaging by processing coherently the backscattered signal recorded over consecutive pings as the bearing platform moves along a predefined path. Coherent processing requires accurate estimation and compensation of the platform's motion for high quality imaging. The motion of the platform carrying the SAS system can be estimated by cross-correlating redundant recordings at successive pings due to the spatiotemporal coherence of statistically homogeneous backscatter. This data-driven approach for estimating the motion of the SAS platform is essential when positioning information from navigational instruments is absent or inadequately accurate. Herein, the problem of platform motion estimation from coherence measurements of diffuse backscatter is formulated in a probabilistic framework. A variational autoencoder is designed to disentangle the ping-to-ping platform displacement from three-dimensional (3D) spatiotemporal coherence measurements. Unsupervised representation learning from unlabeled data offers robust 3D platform motion estimation. Including a small amount of labeled data during training improves further the platform motion estimation accuracy.

Iterative, Deep Synthetic Aperture Sonar Image Segmentation

Synthetic aperture sonar (SAS) systems produce high-resolution images of the seabed environment. Moreover, deep learning has demonstrated superior ability in finding robust features for automating imagery analysis. However, the success of deep learning is conditioned on having lots of labeled training data, but obtaining generous pixel-level annotations of SAS imagery is often practically infeasible. This challenge has thus far limited the adoption of deep learning methods for SAS segmentation. Algorithms exist to segment SAS imagery in an unsupervised manner, but they lack the benefit of state-of-the-art learning methods and the results present significant room for improvement. In view of the above, we propose a new iterative algorithm for unsupervised SAS image segmentation combining superpixel formation, deep learning, and traditional clustering methods. We call our method Iterative Deep Unsupervised Segmentation (IDUS). IDUS is an unsupervised learning framework that can be divided into four main steps: 1) A deep network estimates class assignments. 2) Low-level image features from the deep network are clustered into superpixels. 3) Superpixels are clustered into class assignments (which we call pseudo-labels) using $k$-means. 4) Resulting pseudo-labels are used for loss backpropagation of the deep network prediction. These four steps are performed iteratively until convergence. A comparison of IDUS to current state-of-the-art methods on a realistic benchmark dataset for SAS image segmentation demonstrates the benefits of our proposal even as the IDUS incurs a much lower computational burden during inference (actual labeling of a test image). Finally, we also develop a semi-supervised (SS) extension of IDUS called IDSS and demonstrate experimentally that it can further enhance performance while outperforming supervised alternatives that exploit the same labeled training imagery.

An Improved Imaging Algorithm for Multi-Receiver SAS System with Wide-Bandwidth Signal

When the multi-receiver synthetic aperture sonar (SAS) works with a wide-bandwidth signal, the performance of the range-Doppler (R-D) algorithm is seriously affected by two approximation errors, i.e., point target reference spectrum (PTRS) error and residual quadratic coupling error. The former is generated by approximating the PTRS with the second-order term in terms of the instantaneous frequency. The latter is caused by neglecting the cross-track variance of secondary range compression (SRC). In order to improve the imaging performance in the case of wide-bandwidth signals, an improved R-D algorithm is proposed in this paper. With our method, the multi-receiver SAS data is first preprocessed based on the phase center approximation (PCA) method, and the monostatic equivalent data are obtained. Then several sub-blocks are generated in the cross-track dimension. Within each sub-block, the PTRS error and residual quadratic coupling error based on the center range of each sub-block are compensated. After this operation, all sub-blocks are coerced into a new signal, which is free of both approximation errors. Consequently, this new data is used as the input of the traditional R-D algorithm. The processing results of simulated data and real data show that the traditional R-D algorithm is just suitable for an SAS system with a narrow-bandwidth signal. The imaging performance would be seriously distorted when it is applied to an SAS system with a wide-bandwidth signal. Based on the presented method, the SAS data in both cases can be well processed. The imaging performance of the presented method is nearly identical to that of the back-projection (BP) algorithm.

Spatial Coherence of Speckle for Repeat-Pass Synthetic Aperture Sonar Micronavigation

Accurate positioning of autonomous underwater vehicles is a major challenge. The long-term drift is problematic if global position updates are not available, and for applications such as repeat-pass interferometry and coherent change detection, millimeter accuracy is needed. Repeat-pass synthetic aperture sonar (SAS) micronavigation is one potential technique for countering both challenges. While single-pass SAS micronavigation enabled successful coherent processing within one track, the potential is that repeat-pass SAS micronavigation can support coherent processing between passes. Both techniques are based on recognizing the speckle pattern in the seafloor return, but repeat-pass SAS micronavigation has additional challenges with the larger temporal and spatial separations between the observations. In this study, we investigate the spatial correlation of speckle as observed from SAS systems. We divide the different contributions to spatial decorrelation into three groups: 1) speckle decorrelation; 2) footprint mismatch; and 3) stretching. We examine each contribution separately and develop simplified formulas for their decorrelation baselines. When correlating synthetic aperture images, decorrelation from stretching dominates. When correlating single-element data recorded at low grazing angles common to SAS, speckle decorrelation dominates. We validate our findings on experimental data, and by combining elements into larger effective elements, we demonstrate increasing the across-track baseline for repeat-pass SAS micronavigation updates from less than 1 to 10 m.

Spatially variant autofocus for circular synthetic aperture sonar.

Circular synthetic aperture sonar (CSAS) is a method for improving the resolution and target detection capabilities of a synthetic aperture sonar system. CSAS data are difficult to focus because of their large aperture sizes and elevation sensitivity. This difficulty has sometimes been addressed by using transponders or distributing isotropic scatterers in the field of view of the system; however, this comes at the cost of reduced practicality. As an alternative, map-drift based multipoint autofocus ("multilateration") was proposed by Cantalloube and Nahum [IEEE Trans. Geosci. Remote Sens. 49, 3730-37 (2011)] for autofocusing analogous circular synthetic aperture radar data. Multilateration also solves the problem of aberration spatial variance by providing a three-dimensional navigation correction. In circular synthetic aperture focusing problems, though, correcting aberrations is a joint navigation and elevation estimation problem, and the present work extends the multilateration approach to simultaneously solve both a navigation solution and coordinate corrections for the multilateration control patches. Additionally, methods for addressing the stability and behavior of the inverse problem are addressed, and an adaptive weighting scheme for reducing the influence of outliers is presented. The field results demonstrate near optimal point-spread functions on distributions of natural isotropic scatterers and robustness in regions with bathymetric variability.

예인형 간섭계측합성개구소나 시스템 개발 및 실시간 높이 정보 추출 방법

예인형 간섭계측합성개구소나 시스템 개발 및 실시간 높이 정보 추출 방법

Persistent scatterers detection from phase statistics of multi-view synthetic aperture sonar imaging

Persistent scatterers (PS) refer to stationary point reflectors that exhibit strong phase stability, irrespective of their amplitude. Synthetic aperture sonar (SAS) systems provide high-resolution seafloor imaging by adding coherently the backscattered waves from multiple views along the synthetic antenna. Traditionally, PS detection relies on the magnitude of the resulting image, thus it is susceptible to high levels of speckle noise. This study demonstrates that PS can be detected more reliably from pixel-wise phase statistics of multi-view complex SAS images. Phase-based feature extraction aims to facilitate automatic target recognition tasks without training data.