| 67 | 0 | 37 |
| 下载次数 | 被引频次 | 阅读次数 |
针对水下图像识别中目标特征弱、小目标物体数量庞大且信息量少，以及水下环境光线吸收和散射导致目标识别困难等问题，提出C3k2＿RD-EMA-SOB-YOLOv8(CES-YOLOv8)改进算法。首先，引入C3k2模块替换原有的C2f模块，结合重参数化卷积和膨胀卷积设计轻量化的RD卷积块，降低了模型运算量。其次，引入高效多尺度注意力机制，提高网络对小目标物体关键信息的注意力，从而提升对小目标的检测性能。最后，添加小目标检测分支以获取较浅层的小目标特征信息，使之与深层特征充分融合，降低了小目标物体的漏检率。基于RUOD数据集上的实验结果表明，改进算法在mAP@0.5、mAP@0.5:0.95上领先原始模型1.0%和3.3%。与其他前沿的水下目标检测算法相比，CES-YOLOv8在轻量化的同时提高水下小目标的检测精度。
Abstract：To address the problems in underwater image recognition,including weak target features,a large number of small objects with limited information,and difficulties in target recognition caused by light absorption and scattering in underwater environments, an improved algorithm,C3k2＿RD-EMA-SOB-YOLOv8(CES-YOLOv8),is proposed.First,the C3k2 module is introduced to replace the original C2f module,and a lightweight RD convolution block is designed by combining re-parameterized refocusing convolution with dilated convolution,thereby reducing the computational cost of the model. Second,an efficient multi-scale attention mechanism is introduced to enhance the network's focus on key information of small objects,thereby improving small-object detection performance.Finally,a small-object detection branch is added to capture shallow feature information of small objects and fully fuse it with deep features,thus reducing the missed detection rate of small objects. Experimental results on the RUOD dataset show that,compared with the original model,the improved algorithm improves mAP@0. 5 and mAP@0. 5∶0. 95 by 1. 0%and 3. 3%,respectively. Compared with other cutting-edge underwater object detection algorithms,the present algorithm improves the detection accuracy while lightweighting the improvement,which effectively improves the ability to detect small underwater objects.
[1]Petillot Y R,Antonelli G,Casalino G,et al.Underwater robots:from remotely operated vehicles to intervention-autonomous underwater vehicles[J].IEEE Robotics&Automation Magazine,2019,26(2):94-101.
[2]Macreadie P I,Mclean D L,Thomson P G,et al.Eyes in the sea:unlocking the mysteries of the ocean using industrial,remotely operated vehicles(ROVs)[J].Science of the Total Environment,2018,634:1077-1091.
[3]Bingham B,Foley B,Singh H,et al.Robotic tools for deep water archaeology:surveying an ancient shipwreck with an autonomous underwater vehicle[J].Journal of Field Robotics,2010,27(6):702-717.
[4]Vasileiou M,Vlontzos G.A hybrid remotely operated underwater vehicle for maintenance operations in aquaculture:Practical insights from Greek fish farms[J].Computers and Electronics in Agriculture,2025,232:110045.
[5] Mosk A P,Lagendijk A,Lerosey G,et al.Controlling waves in space and time for imaging and focusing in complex media[J].Nature Photonics,2012,6(5),283-292.
[6]Ho M,El-Borgi S,Patil D,et al.Inspection and monitoring systems subsea pipelines:a review paper[J].Structural Health Monitoring,2020,19(2):606-645.
[7]He K,Gkioxari G,Dollár P,et al.Mask R-CNN[C]//Proceedings of the IEEE international conference on computer vision.2017:2961-2969.
[8]Girshick R.Fast R-CNN[C]//Proceedings of the IEEE international conference on computer vision.2015:1440-1448.
[9]Ren S,He K,Girshick R,et al.Faster R-CNN:Towards real-time object detection with region proposal networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2016,39(6):1137-1149.
[10]Cheng X,Yu J.RetinaNet with difference channel attention and adaptively spatial feature fusion for steel surface defect detection[J].IEEE Transactions on Instrumentation and Measurement,2020,70:1-11.
[11]Liu W,Anguelov D,et al.SSD:Single shot multibox detector[C]//Computer Vision-ECCV 2016:14th European Conference,Amsterdam,The Netherlands,Springer International Publishing,2016:21-37.
[12]Jiang P,Ergu D,Liu F,et al.A Review of Yolo algorithm developments[J].Procedia Computer Science,2022,199:1066-1073.
[13]苗力恒,田莹.改进YOLOv8n的多尺度轻量化水下目标检测[J].电子测量与仪器学报,2025,39(4):141-151.
[14]李玉,陈铭治.融合注意力机制的轻量化水下目标检测模型[J].计算机仿真,2025,42(4):172-176,429.
[15]张文凯,余敏,刘浩煜,等.低检测置信度下轻量化水下多目标跟踪算法[J].舰船科学技术,2025,47(6):128-133.
[16]赵晨,陈明.水下底栖生物轻量化目标检测算法YOLOv7-RFPCW[J].农业工程学报,2024,40(11):168-177.
[17]岳静飞,周孟然.基于改进YOLOv8的水下目标检测算法[J/OL].重庆工商大学学报(自然科学版),1-12[2024-10-29].http://link.cnk.net/urlid/500.1155.N.20241029.1248.002.
[18]Ultralytics,Track and count objects using YOLOv8,roboflow.[Online].Available:https://blog.roboflow.com/yolov8-tracking-and-counting/#object-detection-with-yolov8.[Accessed:May 24,2024].
[19]Ge Z,Liu S,Wang F,et al.YOLOX:Exceeding yolo series in 2021[EB/OL].arXiv:2107.08430,2021.
[20]CHEN X,JIANG N,YU Z,et al.Citrus leaf disease detection based on improved YOLO11 with C3K2[C]//International Conference on Computer Graphics,Artificial Intelligence,and Data Processing.SPIE,2025,13560:746-751.
[21]Yu F,Koltun V.Multi-scale context aggregation by dilated convolutions[EB/OL].arXiv:1511.07122,2015.
[22]Cai Z,Ding X,Shen Q,et al.Refconv:Reparameterized refocusing convolution for powerful convnets[J].IEEE Transactions on Neural Networks and Learning Systems,2025.DOI:10.1109/TNNLS.2025.3552654.
[23]Ouyang D,He S,Zhang G,et al.Efficient multi-scale attention module with cross-spatial learning[C]//ICASSP 2023 IEEE International Conference on Acoustics,Speech and Signal Processing(ICASSP).Rhodes Island,Greece:IEEE,2023:1-5.
[24]Hou Q,Zhou D,Feng J.Coordinate attention for efficient mobile network design[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Nashville,TN,USA:IEEE/CVF,2021:13708-13717.
[25]Fu C,Liu R,Fan X,et al.Rethinking general underwater object detection:datasets,challenges,and solutions[J].Neurocomputing,2023,517:243-256.
基本信息:
DOI：10.16389/j.cnki.cn42-1737/n.2026.04.008
中图分类号:TP391.41
引用信息:
[1]隗一凡,李浩,吴文俊,等.轻量化YOLO小目标检测算法在水下目标检测中的应用[J].江汉大学学报(自然科学版)().DOI:10.16389/j.cnki.cn42-1737/n.2026.04.008.
基金信息:
江汉大学2025年研究生科研创新基金项目(KYCXXJJ2025S08)
2026-07-03
2026-07-03
2026-07-03