Full-coverage and high-efficiency seabed sediment detection and identification are critical elements of digital marine construction that support the three-dimensional and thematic development of maritime spatial geographic information systems. With the development of multibeam echo sounder (MBES), the use of MBES backscatter intensity and bathymetry data to extract backscatter angular response (AR) features has increased. Using backscatter intensity features and seabed terrain features for classification is an effective way to achieve large-scale seabed sediment classification. However, it was still limited by small sample size problems and the poor stability of the classification model due to the complexity of performing seabed sediment sampling. In response to the above issues, this paper proposes a sample enhancement method based on simple linear iterative clustering (SLIC) superpixel segmentation to address the problems. First, a superpixel-based sample homogeneity expansion method is combined with multibeam backscatter intensity images to achieve adaptive sample range selection. Then, a Random Forest (RF)-based model is constructed using MBES backscatter intensity and seabed terrain features. To assess the model’s validity, the experiment uses data from an extensive MBES survey and field sampling information from the Celtic Sea, UK. It achieves accurate predictions for the area’s eight sediment types. The experimental results show that the proposed method expands 60 groups of original sample points to 9293 groups of valid sample points, and achieves an overall classification accuracy and kappa coefficient of 84.05% and 0.81 for the seabed sediment, respectively. In addition, the classification accuracy is significantly better compared to the traditional methods.

Triangulated irregular networks (TINs) are widely used in terrain visualization due to their accuracy and efficiency. However, the conventional algorithm for multi-scale terrain rendering, based on TIN, has many problems, such as data redundancy and discontinuities in scale transition. To solve these issues, a method based on a detail-increment model for the construction of a continuous-scale hierarchical terrain model is proposed. First, using the algorithm of edge collapse, based on a quadric error metric (QEM), a complex terrain base model is processed to a most simplified model version. Edge collapse records at different scales are stored as compressed incremental information in order to make the rendering as simple as possible. Then, the detail-increment hierarchical terrain model is built using the incremental information and the most simplified model version. Finally, the square root of the mean minimum quadric error (MMQE), calculated by the points at each scale, is considered the smallest visible object (SVO) threshold that allows for the scale transition with the required scale or the visual range. A point cloud from Yanzhi island is converted into a hierarchical TIN model to verify the effectiveness of the proposed method. The results show that the method has low data redundancy, and no error existed in the topology. It can therefore meet the basic requirements of hierarchical visualization.

Riverbed sediments are crucial in river ecosystems, significantly impacting water quality, environmental protection, and water resource management. With the increasing ease of obtaining high-resolution river data, integrating multibeam technology and field sampling has become one of the popular methods for identifying riverbed sediments. However, limitations in underwater field sampling techniques result in insufficient sampling data, leaving many unlabeled high-resolution acoustic data underutilized, which hinders practical training and deployment of classifiers. To address this issue, this paper proposes a sediment sample enhancement method based on superpixel segmentation, confident learning, and active learning to improve model performance. First, superpixel segmentation is employed to expand field sampling data and increase the number of samples. Next, confident learning removes noisy labels from the expanded samples, improving data quality. Finally, the centers of superpixel blocks are introduced as unlabeled data and active learning techniques are used to select high-quality unlabeled data for labeling. Experiments combining multibeam data from the FuchunRiver region with actual sediment sampling data classify five riverbed sediments. The proposed method expands the original 18 sampling points to 8,333 effective sampling points. Comparison results show that the proposed method achieves a classification accuracy of 85.3%, an improvement of 11.1% to 11.3% compared to traditional classification methods, offering a new approach for large-scale benthic environment detection in river basins.

Refraction errors induced by inaccurate or insufficient sound-speed profiles (SSPs) can result in significant distortion of seafloor topography during multibeam bathymetric data processing. The current refraction error correction methods are focused on SSP inversion in shallow and medium water depth areas with relatively flat terrain. To address the problems encountered in deep-sea environments, an improved SSP inversion method is introduced that uses an empirical orthogonal function (EOF) decomposition and multiobjective evolutionary algorithm based on decomposition (MOEA/D). Initially, SSPs obtained from the Argo historical data set are decomposed by EOF decomposition, which extracts the orthogonal eigenvectors and the range of reconstruction coefficients. Then, the absolute mean value and standard deviation of depth differences within overlapping swathes are used as constraint conditions, forming dual objective functions. The MOEA/D algorithm is employed to determine the optimal EOF coefficients. Finally, ray tracing is conducted using the inverted SSP to correct the seafloor distortion. The method was verified using experimental data in complex deep-sea terrain. The maximum depth difference within the overlapping area of adjacent swaths notably decreases from 14.9 to 5.6 m, and the root mean square error decreases from 6.4 to 1.2 m.

Bottom tracking is an essential step in sidescan sonar image processing, which plays a crucial role in geometric distortion correction, geocoding, and image stitching. However, it is difficult to achieve accurate and automatic bottom tracking due to the influence of reflected surface echoes or suspensions in water. Therefore, this article proposes a robust and automatic bottom-tracking method considering multiple influencing factors. First, the bottom-tracking range is delineated by determining whether there are surface echoes. Then, the intensity difference between adjacent sampling intervals of the sonar image is calculated and bottom tracking is performed by an adaptive threshold. Finally, the bottom-tracking results are smoothed by combining them with the robust linear regression algorithm. Experimental results show that the detection accuracy of the proposed method is above 95%, which is higher than the results of the conventional threshold method (79.3%), Laplacian of Gaussain (LOG) operator (77.0%), and Canny operator (87.0%). The proposed method can adaptively adjust the threshold parameters and has a better bottom-tracking result with less computational complexity.