水道港口2024，Vol.45Issue(6)：954-962,9.
基于改进YOLOv5s的船舶水尺检测模型研究
Study on improved YOLOv5s-based ship water gauge detection model
摘要
Abstract
Ship water gauge detection is a key in image-based detection technology of ship draft depth.During detection,it faces challenges such as gauge offset,rotation,distortion,and small target sizes,making accurate identification detection difficult.In this paper,an improved YOLOv5s ship water gauge detection model was proposed for enhancing feature extraction capabilities for irregular and distorted targets through the incorporation of Deformable Convolutional Networks(DCN).Additionally,the model incorporated the Convolutional Block Attention Module(CBAM)to improve feature representation in key areas of the target.A ship water gauge dataset was developed based on on-site monitoring images from the Gaogang Ship Lock in Taizhou for testing purposes.Results demonstrate that the improved model shows a significant improvement in terms of precision,recall,Fl score,mAP@0.5,and mAP@0.5:0.95 compared to the original model.Moreover,the overall performance of this model is superior to that of commonly used mainstream algorithms,effectively improving the accuracy of ship water gauge detection.It provides an important support for the automatic detection of ship draft depth.关键词
船舶水尺检测/YOLOv5s模型/可变形卷积/注意力机制/深度学习Key words
ship water gauge detection/YOLOv5s model/deformable convolution/attention mechanism/deep learning分类
交通工程引用本文复制引用
张桂荣,陈志宏,刘志荣,孙巍,侯利军..基于改进YOLOv5s的船舶水尺检测模型研究[J].水道港口,2024,45(6):954-962,9.基金项目
国家重点研发计划项目(2023YFC3206103) （2023YFC3206103）
江苏省水利科技项目(2022053) （2022053）
江苏省交通运输科技与成果转化项目(2024Y09) （2024Y09）
