基于改进YOLOv8的红外船舶检测  

Infrared Ship Detection Based on Improved YOLOv8

作　　者：王海群 魏培旭 解浩龙 左嘉炜 WANG Haiqun;WEI Peixu;XIE Haolong;ZUO Jiawei(School of Electrical Engineering,North China University of Science and Technology,Tangshan 063000,China)

机构地区：[1]华北理工大学电气工程学院,河北唐山063000

出　　处：《电光与控制》2025年第1期61-67,共7页Electronics Optics & Control

基　　金：河北省自然科学基金(F2021209006)。

摘　　要：针对现有红外船舶检测算法检测精度低和实时性不足问题,提出一种基于改进YOLOv8的红外船舶检测算法。首先,将设计的MCA机制引入到YOLOv8的主干网络,增强主干网络的多尺度特征提取能力;其次,对YOLOv8的检测头进行共享参数和重参数化设计,以此提升检测头的检测效率;然后,使用BiFPN结构改进YOLOv8的颈部网络,利用双向信息流和可学习权重加强网络的特征表达能力;最后,使用Faster Block对YOLOv8的C2f模块进行改进,保持精度的同时减少参数量,提升算法的检测速度。该算法在红外船舶数据集上测试,mAP值达到了93.1%,相比较原算法提高了2.5个百分点,参数量比原算法减少了32.6%。实验结果表明,改进后的算法比原算法有了较大提升,证明了改进算法的有效性。Aiming at the problems of low detection accuracy and lack of real-time performance of existing infrared ship detection algorithms,an infrared ship detection algorithm based on improved YOLOv8 is proposed.Firstly,the Multiscale Coordinate Attention(MCA)mechanism designed in this paper is introduced into the backbone network of YOLOv8 to enhance the capability of multi-scale feature extraction.Secondly,the YOLOv8 detection head is designed with shared parameters and re-parameterization,so as to improve the detection efficiency of the detection head.Then,the neck network of YOLOv8 is improved by using BiFPN structure,and the feature expression capability of the network is enhanced by bidirectional information flow and learnable weights.Finally,Faster Block is used to improve the C2f module of YOLOv8,which can maintain the accuracy while reducing the quantity of parameters and improve the detection speed of the model.The algorithm is tested on the infrared ship data set,and the mAP value reaches 93.1%,which is 2.5 percentage points higher than that of the original model,and the quantity of parameters is 32.6%lower than the original model.The experimental results show that the improved algorithm is much better than the original algorithm,which proves the effectiveness of the improved algorithm.

关 键 词：红外船舶检测 YOLOv8 注意力机制 BiFPN 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]