Synthetic Aperture Sonar (SAS) imaging technology is wildly used in the underwater applications. In the work process of SAS imaging, filtering technologies are important for SAS imaging, which can suppress different noises to improve signal quality. However, the existing filtering methods face many challenges, such as insufficient noise suppression, degradation of image detail, edge blurring and so on. Furthermore, the existing quality assessments for filtering methods are sometimes subjective, which limits the research development for filtering technologies. To solve these problems, we propose a hybrid filtering method with a no-reference quality assessment for SAS images in this paper. The proposed method includes two-stages, the first stage is to suppress local statistical interference, and the second stage is to preserve edge information by weighted smoothing. With the no-reference quality assessment, the hybrid filtering method and other filtering methods, including mid-value filtering and mean-value filtering methods, are investigated. The numerical results show that the no-reference quality assessment method can efficiently analyze different filtering methods, and the proposed methods can perform better than other filtering methods.

Underwater object detection with Synthetic Aperture Sonar (SAS) images faces many problems, including low contrast, blurred edges, high-frequency noise, and missed small objects. To improve these problems, this paper proposes a high-accuracy underwater object detection algorithm for SAS images, named the HAUOD algorithm. First, considering SAS image characteristics, a sonar preprocessing module is designed to enhance the signal-to-noise ratio of object features. This module incorporates three-stage processing for image quality optimization, and the three stages include collaborative adaptive Contrast Limited Adaptive Histogram Equalization (CLAHE) enhancement, non-local mean denoising, and frequency-domain band-pass filtering. Subsequently, a novel C2fD module is introduced to replace the original C2f module to strengthen perception capabilities for low-contrast objects and edge-blurred regions. The proposed C2fD module integrates spatial differential feature extraction, dynamic feature fusion, and Enhanced Efficient Channel Attention (Enhanced ECA). Furthermore, an underwater multi-scale contextual attention mechanism, named UWA, is introduced to enhance the model’s discriminative ability for multi-scale objects and complex backgrounds. The proposed UWA module combines noise suppression, hierarchical dilated convolution groups, and dual-dimensional attention collaboration. Experiments on the Sonar Common object Detection dataset (SCTD) demonstrate that the proposed HAUOD algorithm achieves superior performance in small object detection accuracy and multi-scenario robustness, attaining a detection accuracy of 95.1%, which is 8.3% higher than the baseline model (YOLOv8n). Compared with YOLOv8s, the proposed HAUOD algorithm can achieve 6.2% higher accuracy with only 50.4% model size, and reduce the computational complexity by half. Moreover, the HAUOD method exhibits significant advantages in balancing computational efficiency and accuracy compared to mainstream detection models.

There are many interesting machine learning (ML) applications in underwater acoustics. However, some ML algorithms require large amounts of training and testing data and there is a lack of open-source data for these purposes. The Point-based Sonar Signal Model (PoSSM) is a useful tool that generates synthetic time-series data appropriate for coherent signal processing applications. One underwater acoustics application is the identification of objects in synthetic aperture sonar (SAS) imagery. This paper introduces multiple datasets that provide a collection of SAS imagery from a generic SAS sonar above multiple seafloor textures and bathymetries. A variety of objects (e.g., cylinders, rocks, lobster traps) are contained in the imagery. This collection of synthetic data is suitable for training and testing ML algorithms that span a range of complexity, from constant false alarm rate (CFAR) automated detectors to convolutional neural networks (CNNs). These algorithms can perform a variety of tasks that include object detection and classification. In this paper, we test a CFAR detector on image data to estimate object locations. An example use-case of the synthetic data is shown via the training and evaluation of CNN-based classifiers for object recognition.

Abstract The synthetic aperture sonar (SAS) back projection (BP) algorithm is characterized by its ability to adapt to complex trajectories, facilitate motion error compensation, and achieve precise imaging. However, the computational complexity of the BP algorithm is high, making real-time imaging difficult to achieve in miniaturized and lightweight signal processing units. This paper proposes a focused back-projection imaging method for SAS sub-arrays under heterogeneous computing and leverages the high-performance parallel computing capabilities of the graphics processing unit (GPU) to accelerate the imaging process. The correctness and efficiency of the proposed rapid imaging algorithm are verified through simulation and actual measurement data. Experimental results show that compared to CPU parallel imaging algorithms, the acceleration ratio reaches 24.79, demonstrating excellent acceleration performance, which can meet the real-time signal processing requirements of SAS in miniaturized and lightweight signal processing units.

The performance of Convolutional Neural Net (CNN) based target recognition algorithms for Synthetic Aperture Sonar (SAS) imagery is strongly dependent on the seafloor background, with worse classification performance for samples with backgrounds outside the CNN training curriculum. Online performance estimation seeks to identify whether new samples had representation within the training curriculum, and if so whether a given model is expected to perform well on that sample. Performance estimation is critical in underwater object classification, given the diversity of seabed textures and sparsity of training data. Disentanglement learning is a data-driven approach to learning a latent representation that is conceptually interpretable.In the context of performance estimation, we use a disentanglement learning approach based on variational autoencoders to train a latent representation that encodes separately both target class and background type. Online performance estimation is performed by determining if a given sample has a background that is likely to be in the class of backgrounds included in the training data. Results are demonstrated using in-air linear SAS scans from the airSAS dataset.

Detecting bright point scatterers plays an important role in assessing the quality of many sonar, radar, and medical ultrasound imaging systems, especially for characterizing the resolution. Traditionally, prominent scatterers, also known as coherent scatterers, are usually detected by employing thresholding techniques alongside statistical measures in the detection processing chain. However, these methods can perform poorly in detecting point-like scatterers in relatively high levels of speckle background and can distort the structure of the scatterer when visualized. This paper introduces a fast image-processing method to visually identify and detect point scatterers in synthetic aperture imagery using the bright feature transform (BFT). The BFT is analytic, computationally inexpensive, and requires no thresholding or parameter tuning. We derive this method by analyzing an ideal point scatterer’s response with respect to pixel intensity and contrast around neighboring pixels and non-adjacent pixels. We show that this method preserves the general structure and the width of the bright scatterer while performing tone mapping, which can then be used for downstream image characterization and analysis. We then modify the BFT to present a difference of trigonometric functions to mitigate speckle scatterers and other random noise sources found in the imagery. We evaluate the performance of our methods on simulated and real synthetic aperture sonar and radar images, and show qualitative results on how the methods perform tone mapping on reconstructed input imagery in such a way to highlight the bright scatterer, which is insensitive to seafloor textures and high speckle noise levels.

Abstract. Hydrocarbon seepage at the deep seafloor fuels flourishing chemosynthetic communities. These seeps impact the functionality of the benthic ecosystem beyond hotspots of gas emission, altering the abundance, diversity, and activity of microbiota and fauna and affecting geochemical processes. However, these chemosynthetic ecotones (chemotones) are far less explored than the foci of seepage. To better understand the functionality of chemotones, we (i) mapped seabed morphology at the periphery of gas seeps in the deep eastern Mediterranean Sea, using video analyses and synthetic aperture sonar; (ii) sampled chemotone sediments and described burrowing using computerized tomography; (iii) explored nutrient concentrations; (iv) quantified microbial abundance, activity, and N2 fixation rates in selected samples; and (v) extracted DNA and explored microbial diversity and function using amplicon sequencing and metagenomics. Our results show that gas seepage creates burrowing intensity gradients at seep ecotones, with the ghost shrimp Calliax lobata primarily responsible for burrowing, which influences nitrogen and sulfur cycling through microbial activity. Burrow walls form a unique habitat, where macromolecules are degraded by Bacteroidota, and their fermentation products fuel sulfate reduction by Desulfobacterota and Nitrospirota. These, in turn, support chemosynthetic Campylobacterota and giant sulfur bacteria Thiomargarita, which can aid C. lobata nutrition. These interactions may support enhanced productivity at seep ecotones.

Abstract The United Nations' sustainable development goal to designate 30% of the oceans as marine protected areas by 2030 requires practical management tools, and in turn ecologically meaningful mapping of the seafloor. Particularly challenging is the mesophotic zone, a critical component of the marine system, a biodiversity hotspot, and a potential refuge. Here, we introduce a novel seafloor habitat management workflow, integrating cm‐scale synthetic aperture sonar (SAS) and multibeam bathymetry surveying with efficient ecotope characterization. In merely 6 h, we mapped ~5 km2 of a complex mesophotic reef at sub‐metric resolution. Applying a deep learning classifier on the SAS imagery, we classified four habitats with an accuracy of 84% and defined relevant fine‐scale ecotones. Visual census with precise in situ sampling guided by SAS images for navigation were utilized for ecological characterization of mapped units. Our preliminary fish surveys indicate the ecological importance of highly complex areas and rock/sand ecotones. These less abundant habitats would be largely underrepresented if surveying the area without prior consideration. Thus, our approach is demonstrated to generate scalable habitat maps at resolutions pertinent to relevant biotas, previously inaccessible in the mesophotic, advancing ecological modeling and management of large seascapes.

The sonar image has limitations on the physical spatial resolution due to system configuration and underwater environment, which often leads to challenges for underwater targets detection. Here, the deep learning method is applied to enhance the physical spatial resolution of underwater sonar images. Specifically, the U-shaped end-to-end neural network which contains down-sampling and up-sampling parts is proposed to improve the physical spatial resolution limited by the array aperture. The single target and multiple cases are considered separately. In both cases, the normalized loss on the testing sets declines rapidly, and the predicted high-resolution images own great agreement with the ground truth eventually. Further improvements in resolution are focused on, that is, compressing the predicted high-resolution image to its physical spatial resolution limitation. The results show that the trained end-to-end neural network could map high resolution targets to the impulse responses at the same location and amplitude with an uncertain target number. The proposed convolutional neural network approach could give a practical alternative to improve the physical spatial resolution of underwater sonar images.

Improved hardware and processing techniques such as synthetic aperture sonar have led to imaging sonar with centimeter resolution. However, practical limitations and old systems limit the resolution in modern and legacy datasets. This study proposes using single image super resolution based on a conditioned diffusion model to map between images at different resolutions. This approach focuses on upscaling legacy, low-resolution sonar datasets to enable backward compatibility with newer, high-resolution datasets, thus creating a unified dataset for machine learning applications. The study demonstrates improved performance for classifying upscaled images without increasing the probability of false detection. The increased probability of detection was 7% compared to bicubic interpolation, 6% compared to convolutional neural networks, and 2% compared to generative adversarial networks. The study also proposes two sonar specific evaluation metrics based on acoustic physics and utility to automatic target recognition.

Classification of Imaging Artifacts in Synthetic Aperture Sonar With Bayesian Deep Learning

Small Sample Image Classification for Synthetic Aperture Sonar Based on Super-Resolution Reconstruction and Improved Self-supervised Contrastive Learning

Interferometric synthetic aperture sonar (InSAS) is an emerging sonar technology for high-resolution mapping and imaging of the seafloor. This technology is increasingly utilized for defense- and commercial-related applications. However, its application for scientific and environmental purposes remains limited. In this article, we describe the development of InSAS as a tool for seafloor characterization. We discuss the potential applications for InSAS that extend its use beyond traditional defense and offshore infrastructure related surveys to applications for habitat classification, environmental monitoring, and seafloor geological characterization.

Seafloor Slope Estimation and Its Theoretical Accuracy in Interferometric Synthetic Aperture Sonar

This paper describes a synthetic aperture sonar (SAS) dataset collected in-air consisting of four types of targets in four environments of different complexity. The in-air laboratory based experiments produced data with a level of fidelity and ground truth accuracy that is not easily attainable in data collected underwater. The range of complexity, high level of data fidelity, and accurate ground truth provides a rich dataset with acoustic features on multiple scales. It can be used to develop new signal-processing and image reconstruction algorithms, as well as machine learning models for object detection and classification. It may also find application in model verification and validation for acoustic simulators. The dataset consists of raw acoustic time series returns, associated environmental conditions, hardware configuration, array motion, as well as the reconstructed imagery.

Synthetic aperture image enhancement with near-coinciding Nonuniform sampling case

Morphological component analysis (MCA) involves the separation of time series into components based on morphological factors, such as duration or envelope type. Acoustic returns from sonar systems can be separated into short-duration and long-duration components using MCA, and the resulting time series formed into traditional synthetic aperture sonar (SAS) images. We present an application of this approach to classification of data from hollow and solid spherical targets collected using a linear in-air SAS system (AirSAS). AirSAS time series were separated into morphological components and, utilizing online image formation, the resulting imagery processed through a multi-branch convolutional neural network (CNN) architecture to classify targets by type. Classification results are included for a variety of scene backgrounds and representation modalities. This is a challenging classification problem since the targets have the same exterior geometry and are distinguished by non-specular acoustic phenomenology.

Synthetic aperture sonar interferometry relies on the interferogram of two single look complex images to estimate bathymetry. The phase difference measurements have variance, which is typically reduced by spatial smoothing at the cost of horizontal resolution. The high resolution intensity image is related to the bathymetry because of the observation geometry. We therefore suggest an approach that constrains the filtering around edges found by intensity image segmentation. We demonstrate our suggested method on simulated data and show quantitative and qualitative improvements in both the horizontal resolution and the shape resolvability of small objects. We demonstrate a 30% improvement in RMSE of the bathymetric estimate, and observe that the estimated bathymetry more closely renders the real object shape for a small, but elevated object. We demonstrate our suggested method on real data and show similar results.

When a synthetic aperture sonar (SAS) system operates under low-frequency broadband conditions, the azimuth range coupling of the point target reference spectrum (PTRS) is severe, and the high-resolution imaging range is limited. To solve the above issue, we first convert multi-receivers’ signal into the equivalent monostatic signal and then divide the equivalent monostatic signal into range subblocks and the range frequency subbands within each range subblock in order. The azimuth range coupling terms are converted into linear terms based on piece-wise linear approximation (PLA), and the phase error of the PTRS within each subband is less than π/4. Then, we use the chirp-z transform (CZT) to correct range cell migration (RCM) to obtain low-resolution results for different subbands. After RCM correction, the subbands’ signals are coherently summed in the range frequency domain to obtain a high-resolution image. Finally, different subblocks are concatenated in the range time domain to obtain the final result of the whole swath. The processing of different subblocks and different subbands can be implemented in parallel. Computer simulation experiments and field data have verified the superiority of the proposed method over existing methods.

Abstract The validation of seafloor scattering models used for seabed characterisation requires quantifying the contributions from the sediment interface and volume to the total acoustic returns. At low‐frequencies, direct measurements of sediment volume scattering have rarely been made, due to the bias in interface roughness scattering caused by large beamwidths of low‐frequency sonars. Endfire Synthetic Aperture Sonar (EF‐SAS) can achieve narrower beamwidths by forming a vertically oriented synthetic array as a transmitter and/or receiver and moving it through the water column. The narrower beamwidths achieved by EF‐SAS allow for more accurate measurements of volume scattering by reducing interface scattering bias in acoustic returns. The application of EF‐SAS for sediment characterisation is explored for the first time. The authors demonstrate that EF‐SAS can be used to construct the angular response curve for both interface and volume scattering as well as to estimate the attenuation and reflection coefficients, which can be inverted for grain size.