Imaging sonar is a crucial tool for underwater visual perception. Compared to 2D sonar images, 3D sonar images offer superior spatial positioning capabilities, although the data acquisition cost is higher and lacks open source references for data annotation, target detection, and semantic segmentation. This paper utilizes 3D imaging sonar to collect underwater data from three types of targets with 1534 effective frames, including a tire, mannequin, and table, in Liquan Lake, Shanxi Province, China. Based on these data, this study focuses on three innovative aspects as follows: rapid underwater data annotation, loss function optimization, and unsupervised moving target extraction in water. For rapid data annotation, a batch annotation method combining human expertise and multi-frame superposition is proposed. This method automatically generates single-frame target detection boxes based on multi-frame joint segmentation, offering advantages in speed, cost, and accuracy. For loss function optimization, a density-based loss function is introduced to address the issue of overfitting in dense regions due to the uneven distribution of point cloud data. By assigning different weights to data points in different density regions, the model pays more attention to accurate predictions in a sparse area, resulting in a 6.939 improvement in mIOU for semantic segmentation tasks, while lakebed mIOU achieved a high score of 99.28. For unsupervised moving target extraction, a multi-frame joint unsupervised moving target association extraction method called the Double DBSCAN, D-DBSCAN, is proposed. This method simulates human visual sensitivity to moving targets in water and uses a joint D-DBSCAN spatial clustering approach with single-frame and inter-frame superposition, achieving an improvement of 21.3 points in mAP. Finally, the paper summarizes the three proposed innovations and provides directions for further research.

In the realm of underwater environment detection, achieving information matching stands as a pivotal step, forming an indispensable component for collaborative detection and research in areas such as distributed mapping. Nevertheless, the progress in studying the matching of underwater side-scan sonar images has been hindered by challenges including low image quality, intricate features, and susceptibility to distortion in commonly used side-scan sonar images. This article presents a comprehensive overview of the advancements in underwater sonar image processing. Building upon the novel SchemaNet image topological structure extraction model, we introduce a feature matching model grounded in side-scan sonar images. The proposed approach employs a semantic segmentation network as a teacher model to distill the DeiT model during training, extracting the attention matrix of intermediate layer outputs. This emulates SchemaNet’s transformation method, enabling the acquisition of high-dimensional topological structure features from the image. Subsequently, utilizing a real side-scan sonar dataset and augmenting data, we formulate a matching dataset and train the model using a graph neural network. The resulting model demonstrates effective performance in side-scan sonar image matching tasks. These research findings bear significance for underwater detection and target recognition and can offer valuable insights and references for image processing in diverse domains.

The Underwater Sonar Moving Target (USMT) dataset is a forward-looking sonar (FLS) image dataset for low-pixel underwater moving-target perception and segmentation. It contains 21 temporally ordered sequences and 7,136 annotated rectangular B-scan FLS images collected during AUV experiments in a controlled lake environment. Each sonar image is accompanied by a pixel-level annotation in LabelMe JSON format and a TXT file containing the corresponding sonar acquisition metadata. The annotations were generated using X-AnyLabeling with the SAM 2 video model and subsequently reviewed and refined manually. The dataset can support research on underwater intelligent perception, sonar image segmentation, and temporal modeling of moving targets.