Qingfa Zhang

The unique hollowed-out catamaran hulls and complex environmental conditions in lake areas hinder traditional ranging algorithms (combining target detection and stereo matching) from accurately obtaining depth information near the center of ships. This not only impairs the navigation of electric tourist boats but also leads to high computing resource consumption. To address this issue, this study proposes a ranging method integrating improved ORB (Oriented FAST and Rotated BRIEF) with stereo vision technology. Combined with traditional optimization techniques, the proposed method calculates target distance and angle based on the triangulation principle, providing a rough alternative solution for the “gap period” of stereo matching-based ranging. The method proceeds as follows: first, it acquires ORB feature points with relatively uniform global distribution from preprocessed binocular images via a local feature weighting approach; second, it further refines feature points within the ROI (Region of Interest) using a quadtree structure; third, it enhances matching accuracy by integrating the FLANN (Fast Library for Approximate Nearest Neighbors) and PROSAC (Progressive Sample Consensus) algorithms; finally, it applies the screened matching point pairs to the triangulation method to obtain the position and distance of the target ship. Experimental results show that the proposed algorithm improves processing speed by 6.5% compared with the ORB-PROSAC algorithm. Under ideal conditions, the ranging errors at 10m and 20m are 2.25% and 5.56%, respectively. This method can partially compensate for the shortcomings of stereo matching in ranging under the specified lake area scenario.

Accurate and efficient detection of small-scale targets on dynamic water surfaces remains a critical challenge in the deployment of unmanned surface vehicles (USVs) for maritime applications. Complex background interference-such as wave motion, sunlight reflections, and low contrast-often leads to missed or false detections, particularly when using conventional convolutional neural networks. To address these issues, this study introduces LMS-YOLO, a lightweight detection framework built upon the YOLOv8n architecture and optimized for realtime marine object recognition. The proposed network integrates three key components: (1) a C2f-SBS module incorporating StarNet-based Star Blocks, which streamlines multi-scale feature extraction while reducing parameter overhead; (2) a Shared Convolutional Lightweight Detection Head (SCLD), designed to enhance detection precision across scales using a unified convolutional strategy; and (3) a Mixed Local Channel Attention (MLCA) module, which reinforces context-aware representation under complex maritime conditions. Evaluated on the WSODD and FloW-Img datasets, LMS-YOLO achieves a 5.5% improvement in precision and a 2.3% gain in mAP@0.5 compared to YOLOv8n, while reducing parameter count and computational cost by 37.18% and 34.57%, respectively. The model operates at 128 FPS on standard hardware, demonstrating its practical viability for embedded deployment in marine perception systems. These results highlight the potential of LMS-YOLO as a deployable solution for high-speed, high-accuracy marine object detection in real-world environments.

Aiming at the problems of insufficient navigation area recognition accuracy, fuzzy boundary of obstacle segmentation, and high consumption of computational resources in the autonomous navigation of water navigation sensors, such as USVs, this paper proposes a DMTN-Net network architecture based on DeeplabV3+ to improve the accuracy and efficiency of environment sensing. Firstly, DMTN-Net adopts the lightweight MobileNetV2 as the backbone, which reduces the amount of computation. Secondly, the innovative N-Decoder structure integrates cSE and Triplet Attention, which enhances the feature representation and improves the segmentation performance. Finally, various experiments were conducted on the MassMind dataset, Pascal VOC2007 dataset, and related sea areas. The experimental results show that DMTN-Net performs well on MassMind and Pascal VOC2007 datasets, and compared with other mainstream networks, the indexes of mIoU, mPA, and mPrecision are significantly improved, and the computational cost is greatly reduced. In addition, the offshore navigation experiments further validate its performance advantages and provide solid support for the practicalization of USV waterborne sensors.

Advanced ship detection technologies play a critical role in improving maritime safety by enabling the rapid identification of vessels and other maritime targets, thereby mitigating the risk of collisions and optimizing traffic efficiency. Traditional detection methods often demonstrate high sensitivity to minor variations in target appearance but face significant limitations in generalization, making them ill-suited to the complex and dynamic nature of maritime environments. To address these challenges, an enhanced ship detection method, referred to as YOLOv5-SE, has been proposed, which builds upon the YOLOv5 framework. This approach incorporates attention mechanisms within the backbone network to improve the model's focus on key features of small targets, dynamically adjusting the importance of each channel to boost representational capacity and detection accuracy. In addition, a refined version of the Complete Intersection over Union (CIoU) loss function has been introduced to optimize the loss associated with target bounding box prediction, thereby improving localization accuracy and ensuring more precise alignment between predicted and ground-truth boxes. Furthermore, the conventional coupled detection head in YOLOv5 is replaced by a Decoupled Head, facilitating better adaptability to various target shapes and accelerating model convergence. Experimental results demonstrate that these modifications significantly enhance ship detection performance, with mean Average Precision (mAP) at IoU 0.5 reaching 94.9% and 95.1%, representing improvements of 3.1% and 1.2% over the baseline YOLOv5 model, respectively. These advancements underscore the efficacy of the proposed methodology in improving detection accuracy and robustness in challenging maritime settings.