Fast and high-accuracy detection of underwater targets based on side scan sonar images has great potential for marine fisheries, underwater security, marine mapping, underwater engineering and other applications. The following problems, however, must be addressed when using low-resolution side scan sonar images for underwater target detection: (1) the detection performance is limited due to the restriction on the input of multi-scale images; (2) the widely used deep learning algorithms have a low detection effect due to their complex convolution layer structures; (3) the detection performance is limited due to insufficient model complexity in training process; and (4) the number of samples is not enough because of the bad dataset preprocessing methods. To solve these problems, an improved neural network for underwater target detection—which is based on side scan sonar images and fully utilizes spatial pyramid pooling and online dataset preprocessing based on the You Look Only Once version three (YOLO V3) algorithm—is proposed. The methodology of the proposed approach is as follows: (1) the AlexNet, GoogleNet, VGGNet and the ResNet networks and an adopted YOLO V3 algorithm were the backbone networks. The structure of the YOLO V3 model is more mature and compact and has higher target detection accuracy and better detection efficiency than the other models; (2) spatial pyramid pooling was added at the end of the convolution layer to improve detection performance. Spatial pyramid pooling breaks the scale restrictions when inputting images to improve feature extraction because spatial pyramid pooling enables the backbone network to learn faster at high accuracy; and (3) online dataset preprocessing based on YOLO V3 with spatial pyramid pooling increases the number of samples and improves the complexity of the model to further improve detection process performance. Three-side scan imagery datasets were used for training and were tested in experiments. The quantitative evaluation using Accuracy, Recall, Precision, mAP and F1-Score metrics indicates that: for the AlexNet, GoogleNet, VGGNet and ResNet algorithms, when spatial pyramid pooling is added to their backbone networks, the average detection accuracy of the three sets of data was improved by 2%, 4%, 2% and 2%, respectively, as compared to their original formulations. Compared with the original YOLO V3 model, the proposed ODP+YOLO V3+SPP underwater target detection algorithm model has improved detection performance through the mAP qualitative evaluation index has increased by 6%, the Precision qualitative evaluation index has increased by 13%, and the detection efficiency has increased by 9.34%. These demonstrate that adding spatial pyramid pooling and online dataset preprocessing can improve the target detection accuracy of these commonly used algorithms. The proposed, improved neural network with spatial pyramid pooling and online dataset preprocessing based on the YOLO V3 method achieves the highest scores for underwater target detection results for sunken ships, fish flocks and seafloor topography, with mAP scores of 98%, 91% and 96% for the above three kinds of datasets, respectively.

A multibeam water column image (WCI) can provide detailed seabed information and is an important means of underwater target detection. However, gas plume targets in an image have no obvious contour information and are susceptible to the influence of underwater environments, equipment noises, and other factors, resulting in varied shapes and sizes. Compared with traditional detection methods, this paper proposes an improved YOLOv7 (You Only Look Once vision 7) network structure for detecting gas plume targets in a WCI. Firstly, Fused-MBConv is used to replace all convolutional blocks in the ELAN (Efficient Layer Aggregation Networks) module to form the ELAN-F (ELAN based on the Fused-MBConv block) module, which accelerates model convergence. Additionally, based on the ELAN-F module, MBConv is used to replace the 3 × 3 convolutional blocks to form the ELAN-M (ELAN based on the MBConv block) module, which reduces the number of model parameters. Both ELAN-F and ELAN-M modules are deep residual aggregation structures used to fuse multilevel features and enhance information expression. Furthermore, the ELAN-F1M3 (ELAN based on one Fused-MBConv block and three MBConv blocks) backbone network structure is designed to fully leverage the efficiency of the ELAN-F and ELAN-M modules. Finally, the SimAM attention block is added into the neck network to guide the network to pay more attention to the feature information related to the gas plume target at different scales and to improve model robustness. Experimental results show that this method can accurately detect gas plume targets in a complex WCI and has greatly improved performance compared to the baseline.

High-precision semantic segmentation of remote sensing imagery is crucial in geospatial analysis. It plays an immeasurable role in fields such as urban governance, environmental monitoring, and natural resource management. However, when confronted with complex objects (such as winding roads and dispersed buildings), existing semantic segmentation methods still suffer from inadequate target recognition capabilities and multi-scale representation issues. This paper proposes a neural network model, LMVMamba (LoRA Multi-scale Vision Mamba), for semantic segmentation of remote sensing images. This model integrates the advantages of convolutional neural networks (CNNs), Transformers, and state-space models (Mamba) with a multi-scale feature fusion strategy. It simultaneously captures global contextual information and fine-grained local features. Specifically, in the encoder stage, the ResT Transformer serves as the backbone network, employing a LoRA fine-tuning strategy to effectively enhance model accuracy by training only the introduced low-rank matrix pairs. The extracted features are then passed to the decoder, where a U-shaped Mamba decoder is designed. In this stage, a Multi-Scale Post-processing Block (MPB) is introduced, consisting of depthwise separable convolutions and residual concatenation. This block effectively extracts multi-scale features and enhances local detail extraction after the VSS block. Additionally, a Local Enhancement and Fusion Attention Module (LAS) is added at the end of each decoder block. LAS integrates the SimAM attention mechanism, further enhancing the model’s multi-scale feature fusion capability and local detail segmentation capability. Through extensive comparative experiments, it was found that LMVMamba achieves superior performance on the OpenEarthMap dataset (mIoU 52.3%, OA 69.8%, mF1: 68.0%) and LoveDA (mIoU 67.9%, OA 80.3%, mF1: 80.5%) datasets. Ablation experiments validated the effectiveness of each module. The final results indicate that this model is highly suitable for high-precision land-cover classification tasks in remote sensing imagery. LMVMamba provides an effective solution for precise semantic segmentation of high-resolution remote sensing imagery.

Reconstructing high-resolution underwater topography using side scan sonar (SSS) images with self-constraint may fail at the marginal part of the surveying area. To solve this problem, the bathymetric data are integrated as the reference to obtain high-resolution and high-accuracy topography. First, based on the seafloor reflection model, the SSS-based reconstructed terrain is obtained. Then, the integration model of SSS-based reconstructed results and bathymetric data in the same surveying area is proposed. Experiments have been used to test the effectiveness of the proposed method and good results have been achieved.

Multibeam bathymetry has become an effective underwater target detection method by using echo signals to generate a high-resolution water column image (WCI). However, the gas plume in the image is often affected by the seafloor environment and exhibits sparse texture and changing motion, making traditional detection and segmentation methods more time-consuming and labor-intensive. The emergence of convolutional neural networks (CNNs) alleviates this problem, but the local feature extraction of the convolutional operations, while capturing detailed information well, cannot adapt to the elongated morphology of the gas plume target, limiting the improvement of the detection and segmentation accuracy. Inspired by the transformer’s ability to achieve global modeling through self-attention, we combine CNN with the transformer to improve the existing YOLOv7 (You Only Look Once version 7) model. First, we sequentially reduce the ELAN (Efficient Layer Aggregation Networks) structure in the backbone network and verify that using the enhanced feature extraction module only in the deep network is more effective in recognising the gas plume targets. Then, the C-BiFormer module is proposed, which can achieve effective collaboration between local feature extraction and global semantic modeling while reducing computing resources, and enhance the multi-scale feature extraction capability of the model. Finally, two different depths of networks are designed by stacking C-BiFormer modules with different numbers of layers. This improves the receptive field so that the model’s detection and segmentation accuracy achieve different levels of improvement. Experimental results show that the improved model is smaller in size and more accurate compared to the baseline.

Remote sensing object detection remains challenging because objects often appear with large scale variation, dense spatial layouts, and strong interference from complex geographical backgrounds. To address these coupled difficulties, we propose EDM-Net, an end-to-end multi-scale detector that organizes feature processing into three coordinated stages: adaptive extraction, intra-scale interaction, and cross-scale fusion. First, an efficient sparse mixture-of-experts (ES-MoE) module is embedded in the backbone to allocate scale-specific convolutional experts according to scene-level feature responses, providing a more adaptive feature basis than a single static extraction path. Second, a dynamic mixing intra-scale feature interaction (DMIFI) module is introduced into the Transformer encoder. This module combines global self-attention with dynamic spatial mixing, thereby preserving long-range context while reintroducing local two-dimensional inductive bias for dense and small objects. Third, a multi-scale synergistic attention fusion (MSAF) module aligns adjacent feature levels through parallel local and global attention branches and structural re-parameterization, reducing semantic dilution during feature aggregation. Comprehensive experiments on three large-scale remote sensing benchmark datasets, DIOR, NWPU VHR-10, and RSOD, demonstrate that EDM-Net consistently improves over the re-trained RT-DETR-R18 baseline under the same experimental protocol, attaining mAP50 scores of 83.7%, 95.6%, and 95.8% respectively. Additional ablation and scale-specific analyses indicate that the three modules contribute complementary gains, especially for small and densely distributed objects. These results suggest that coordinated extraction, interaction, and fusion can improve remote sensing object detection under complex scale and background conditions.