为确保运输类船舶安全航行，提高海上交通的监管效果，本文研究了运输类船舶海上多点礁石视频监测方法。通过在运输类船舶上搭载高清摄像机采集航行过程中的海面图像，利用YOLOv7模型进行单帧图像的多点礁石初步检测，并在YOLOv7的特征提取部分引入了归一化的NAM轻量级注意力机制模块，以增强模型对礁石特征的提取能力；采用LSTM网络进行跨帧间的礁石位置预测，以提高监测的准确性和稳定性，实现运输类船舶海上多点礁石的持续监测。实验结果表明，该方法在不同海况下均能准确监测到多点礁石，且消融实验表明加入NAM轻量级注意力机制和LSTM，IoU峰值提高至0.91，FLOPs维持12.3×109次，验证了对检测准确性和计算效率的提升。
In the field of maritime transportation, reefs are a serious threat to the safe navigation of transportation class ships. In order to ensure the safe navigation of transportation class ships and improve the regulatory effect of sea transportation, this paper proposes a research on the video monitoring method of multi-point reefs at sea for transportation class ships. By carrying a high-definition camera on the transport class ships to collect sea surface images during navigation, the YOLOv7 model is used to carry out the initial detection of multi-point reefs in a single frame image, and the normalized NAM lightweight attention mechanism module is introduced into the feature extraction part of YOLOv7 to enhance the model's ability of extracting features of the reefs; the LSTM network is used to carry out the prediction of the reef position across frames to improve the accuracy and stability of monitoring, and realize the continuous monitoring of multi-point reefs at sea for transportation class ships. The experimental results show that the method can accurately monitor multi-point reefs under different sea conditions, and the ablation experiments show that by adding the NAM lightweight attention mechanism and LSTM, the peak IoU is increased to 0.91 and the FLOPs are maintained at 12.3 × 109 times, which verifies the enhancement of detection accuracy and computational efficiency.
2025,47(3): 163-166 收稿日期：2024-9-28
DOI：10.3404/j.issn.1672-7649.2025.03.027
分类号：U675.79
作者简介：刘鑫(1989-),女,硕士,讲师,研究方向为交通运输工程
参考文献：
[1] 王志鹏, 郝寨柳, 田于逵, 等. 基于FEM/SPH耦合方法的极地运输船舶冰阻力预报研究[J]. 船舶力学, 2021, 25(1): 9?15.
WANG Z P, HAO Z L, TIAN Y K, et al. Research on ice resistance prediction of a polar vessel based on FEM/SPH coupling method[J]. Journal of Ship Mechanics, 2021, 25(1): 9?15.
[2] 孙琳. 基于群智能优化算法的船舶最优运输路线规划[J]. 船舶科学技术, 2024, 46(12): 166?169.
SUN L. Optimal transportation route planning for ships based on swarm intelligence optimization algorithm[J]. Ship Science and Technology, 2024, 46(12): 166?169.
[3] 许述文, 焦银萍, 白晓惠, 等. 基于频域多通道图特征感知的海面小目标检测[J]. 电子与信息学报, 2023, 45(5): 1567?1574.
XU S W, JIAO Y P, BAI X H, et al. Small target detection based on frequency domain multichannel graph feature perception on sea surface[J]. Journal of Electronics & Information Technology, 2023, 45(5): 1567?1574.
[4] 王海峰, 行鸿彦, 陈梦, 等. 基于精细复合多尺度散布熵与XGBoost的海面小目标检测方法[J]. 电子测量与仪器学报, 2023, 37(1): 12?20.
WANG H F, XING H Y, CHEN M, et al. Small target detection method based on refined composite multiscale dispersion entropy and XGBoost[J]. Journal of Electronic Measurement and Instrumentation, 2023, 37(1): 12?20.
[5] 曹禹, 刘光宇, 穆琳琳, 等. 基于多区域最优选择策略的声呐图像目标检测[J]. 西北工业大学学报, 2023, 41(1): 153?159.
CAO Y, LIU G Y, MU L L, et al. Sonar image target detection based on multi-region optimal selection strategy[J]. Journal of Northwestern Polytechnical University, 2023, 41(1): 153?159.
[6] 商林. 基于智能物联网平台的船舶运输路线规划系统[J]. 船舶科学技术, 2020, 42(20): 175?177.
SHANG L. Design of shipping route planning system based on intelligent Internet of things platform[J]. Ship Science and Technology, 2020, 42(20): 175?177.
[7] 王均成, 贺超, 赵志源, 等. 基于YOLOv3-tiny的视频监控目标检测算法[J]. 电子技术应用, 2022, 48(7): 30?33+39.
WANG J C, HE C, ZHAO Z Y, et al. Video surveillance object detection method based on YOLOv3-tiny[J]. Application of Electronic Technique, 2022, 48(7): 30?33+39.
