基于改进YOLOv8的航海雷达图像目标检测算法  

Target detection algorithm for marine radar images based on improved YOLOv8

作　　者：康睿 徐海祥[1,2] 冯辉[1,2] KANG Rui;XU Haixiang;FENG Hui(School of Naval Architecture,Ocean and Energy Power Engineering,Wuhan University of Technology,Wuhan 430063,China;Key Laboratory of High Performance Ship Technology,Wuhan University of Technology,Wuhan 430063,China)

机构地区：[1]武汉理工大学船海与能源动力工程学院,武汉430063 [2]武汉理工大学高性能船舶技术教育部重点实验室,武汉430063

出　　处：《大连海事大学学报》2024年第3期23-30,共8页Journal of Dalian Maritime University

基　　金：国家自然科学基金资助项目(51979210,52371374)。

摘　　要：针对内河船舶航行场景复杂、航海雷达图像形状与颜色特征较少且难以标注的问题,提出一种改进的YOLOv8航海雷达图像目标检测方法。首先,为缓解标注错误与模型过拟合问题,在模型训练阶段引入标签平滑策略;然后,结合雷达图像特有的位置先验信息,设计一种基于坐标的卷积结构用于同时提取目标的形状、颜色和位置特征。为验证该方法的有效性和优越性,对采集的长江航道雷达图像在不同天气环境下进行对比试验。结果表明,本文方法在保证目标检测实时性的同时,精确率达到91.52%,平均精度较经典YOLOv8提高了5.17%,可为提升内河航运现代化与智能化管理水平提供技术支持。In order to solve the problems of complex navigation scenes of inland waterway vessels,few shape and color characteristics of marine radar images,and the difficulty of annotation,an improved YOLOv8 marine radar image target detection method was proposed.Firstly,to alleviate the issues of annotation errors and model overfitting,a label smoothing strategy was introduced during the model training phase.Then,combining the unique positional prior information of radar images,a coordinate based convolution structure was designed to simultaneously extract the shape,color,and positional features of the target.To verify the effectiveness and superiority of the proposed method,comparative experiments were conducted on the collected radar images of the Yangtze River channel under different weather conditions.Results show that the proposed method achieves an accuracy rate of 91.52%while ensuring realtime object detection,with an average accuracy improvement of 5.17%compared to the classic YOLOv8,which can provide technical support for improving the modernization and intelligent management level of inland waterway transportation.

关 键 词：内河航运 航海雷达图像 目标检测 YOLOv8 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TN957.52[自动化与计算机技术—控制科学与工程]