[1]席阳丽,屈 丹,王芳芳,等.基于 FEW-YOLOv8 遥感图像目标检测算法[J].郑州大学学报(工学版),2025,46(04):62-69.[doi:10.13705/j.issn.1671-6833.2025.04.007]
　XI Yangli,QU Dan,WANG Fangfang,et al.Target Detection Algorithm Based on FEW-YOLOv8 Remote Sensing Images[J].Journal of Zhengzhou University (Engineering Science),2025,46(04):62-69.[doi:10.13705/j.issn.1671-6833.2025.04.007]
点击复制

基于 FEW-YOLOv8 遥感图像目标检测算法()
分享到：

《郑州大学学报(工学版)》[ISSN:1671-6833/CN:41-1339/T]

卷:
46
期数:
2025年04期
页码:
62-69
栏目:
出版日期:
2025-07-10

文章信息/Info

Title:
Target Detection Algorithm Based on FEW-YOLOv8 Remote Sensing Images
文章编号:
1671-6833(2025)04-0062-08
作者:
席阳丽1;  屈 丹2; 3;  王芳芳1;  都力铭1
1. 郑州大学 网络空间安全学院,河南 郑州 450001;2. 战略支援部队信息工程大学 信息系统工程学院,河南 郑州450001;3. 先进计算与智能工程(国家级)实验室,河南 郑州 450001
Author(s):
XI Yangli1;  QU Dan2; 3;  WANG Fangfang1;  DU Liming1
1. School of Cyber Science and Engineering, Zhengzhou University, Zhengzhou 450001, China; 2. School of Information System Engineering, Strategic Support Force Information Engineering University, Zhengzhou 450001, China; 3. Laboratory for Advanced Computing and Intelligent Engineering, Zhengzhou 450001, China
关键词:
遥感图像;  YOLOv8;  FasterNet 骨干网络;  EMA 注意力机制;  WIoU 损失函数
Keywords:
remote sensing images;  YOLOv8;  FasterNet backbone network;  EMA attention mechanism;  WIoU loss function
分类号:
TP389. 1
DOI:
10.13705/j.issn.1671-6833.2025.04.007
文献标志码:
A
摘要:
针对遥感图像目标检测任务中进行特征提取时缺少小目标信息,特征融合过程中部分信息丢失,小目标特征信息不明显,导致小目标检测精度不高的问题,提出了一种基于 FEW-YOLOv8 模型的遥感图像目标检测算法。首先,优化骨干网络架构,使用 FasterNet 骨干网络,更有效地提取了遥感图像中小目标的空间特征,使得网络模型更专注于微小目标,从而提升小目标检测精度。 其次,使用 EMA 注意力与 C2f 构建全新的 C2f_EMA 模块,替换Neck 结构中的 C2f 模块,在融合特征前进行特征注意力加强操作,使网络模型更突出特征信息中小目标部分,有效解决特征融合过程中小目标特征丢失问题。 最后,采用带有动态非单调 FM 的 WIoUv3 作为边界框的损失函数,提高了模型的边界框定位精度,并且提升了对小目标的检测性能。 实验结果显示:在 NWPU VHR-10 数据集上经过优化的 YOLOv8 算法的 mAP50 相较于原始 YOLOv8 算法提高了 7. 71 百分点,在 HRSC2016 和 DOTA v1. 0 上分别提高了 9. 70 百分点和 12. 32 百分点,证明所提算法能够有效提升遥感图像中小目标的检测精度。
Abstract:
Aiming at the problems of lack of small target information during feature extraction, partial loss of information during feature fusion, and inconspicuous small target feature information in remote sensing image target detection task, which lead to the low accuracy of small target detection, an algorithm for remote sensing image target detection based on FEW-YOLOv8 model was proposed. Firstly, the backbone network architecture was optimized to use the FasterNet backbone network, which extracted the spatial features of small targets in remote sensing images more efficiently, making the network model more focused on tiny targets, thus improving the small target detection accuracy. Secondly, the new C2f_EMA module was constructed using EMA attention and C2f to replace the C2f module in Neck network, and the feature attention enhancement operation was performed before fusing the features, so that the network model highlighted the small-target part of the feature information more, which effectively solved the problem of small-target feature loss in the process of feature fusion. Finally, WIoUv3, which had a dynamic non-monotonic FM, was used as the bounding box loss function to improve the accuracy of the model′s bounding box localization and strengthen the localization ability of small targets. The experimental results on NWPU VHR-10, HRSC2016 and DOTA v1. 0 datasets showed that the test mAP50 of the improved YOLOv8 algorithm was 7. 71, 9. 70 and 12. 32 percentage points higher than that of the original YOLOv8 algorithm, respectively, which proved that the proposed algorithm could effectively improve the detection accuracy of small targets in remote sensing images.

参考文献/References:

[1] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection [ C]∥2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE, 2016: 779-788. 

[2] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector [ C ] ∥ European Conference on Computer Vision. Cham: Springer, 2016: 21-37. 
[3] GAO F, CAI C X, JIA R H, et al. Improved YOLOX for pedestrian detection in crowded scenes [ J ] . Journal of Real-Time Image Processing, 2023, 20(2) : 1-13. 
[4] XU D Q, WU Y Q. FE-YOLO: a feature enhancement network for remote sensing target detection [ J] . Remote Sensing, 2021, 13(7) : 1311. 
[5] CHEN L Q, SHI W X, DENG D X. Improved YOLOv3 based on attention mechanism for fast and accurate ship detection in optical remote sensing images [ J] . Remote Sensing, 2021, 13(4) : 660. 
[6] XU D Q, WU Y Q. Improved YOLO-V3 with DenseNet for multi-scale remote sensing target detection [ J] . Sensors, 2020, 20(15) : 4276. 
[7] POUDEL R P K, BONDE U, LIWICKI S, et al. ContextNet: exploring context and detail for semantic segmentation in real-time[EB / OL]. (2018-11-05) [ 2024-10- 12] . http:∥arxiv. org / abs/ 1805. 04554. 
[8] CHEN J T, LEI B W, SONG Q Y, et al. A hierarchical graph network for 3D object detection on point clouds[C] ∥2020 IEEE / CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 389-398. 
[9] CHEN J R, KAO S H, HE H, et al. Run, don′t walk: chasing higher FLOPS for faster neural networks[C]∥2023 IEEE / CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 12021-12031. 
[10] ZHOU T, HUANG B, LI R R, et al. An attention-based deep learning model for citywide traffic flow forecasting [ J ] . International Journal of Digital Earth, 2022, 15 (1) : 323-344. 
[11] GAO R X, WANG T F. Motion deblurring algorithm for wind power inspection images based on Ghostnet and SE attention mechanism [ J] . IET Image Processing, 2023, 17(1) : 291-300 
[12] LI G B, SHI G L, JIAO J. YOLOv5-KCB: a new method for individual pig detection using optimized K-means, CA attention mechanism and a bi-directional feature pyramid network[ J] . Sensors, 2023, 23(11) : 5242. 
[13] OUYANG D L, HE S, ZHANG G Z, et al. Efficient multiscale attention module with cross-spatial learning[C]∥2023 IEEE International Conference on Acoustics, Speech and Signal Processing. Piscataway: IEEE, 2023: 1-5. 
[14] YUAN Z, FANG W, ZHAO Y M, et al. Research of insect recognition based on improved YOLOv5[ J] . Journal on Artificial Intelligence, 2021, 3(4) : 145-152. 
[15] WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-theart for real-time object detectors [ C] ∥2023 IEEE / CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 7464-7475. 
[16] YIN L L. Analysis recognition of ghost pepper and cilipadi using mask RCNN and YOLO[ J] . Przegl d Elektrotechniczny, 2023, 1(8) : 94-99. 
[17] LIU S, QI L, QIN H F, et al. Path aggregation network for instance segmentation [ C]∥2018 IEEE / CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 8759-8768. 
[18] ZHENG Z H, WANG P, LIU W, et al. Distance-IoU loss: faster and better learning for bounding box regression[ J] . Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(7) : 12993-13000. 
[19] WANG P J, BAYRAM B, SERTEL E. A comprehensive review on deep learning based remote sensing image super-resolution methods [ J ] . Earth Science Reviews, 2022, 232: 104110. 
[20] PENG X C, CHEN Y Z, CAI X W, et al. An improved YOLOv7-based model for real-time meter reading with PConv and attention mechanisms[ J] . Sensors, 2024, 24(11) : 3549. 
[21] TONG Z J, CHEN Y H, XU Z W, et al. Wise-IoU: bounding box regression loss with dynamic focusing mechanism[EB / OL] . (2023-04-08) [2024-10-12] . https: ∥arxiv. org / abs/ 2301. 10051. 
[22] 贾云飞, 郑红木, 刘闪亮. 基于 YOLOv5s 的金属制品 表面缺陷的轻量化算法研究[ J] . 郑州大学学报( 工 学版) , 2022, 43(5) : 31-38. 
JIA Y F, ZHENG H M, LIU S L. Lightweight surface defect detection method of metal products based on YOLOv5s [ J] . Journal of Zhengzhou University ( Engineering Science) , 2022, 43(5) : 31-38. 
[23] 刘庆华, 杨欣仪, 接浩, 等. 基于融合 GhostNetV2 的 YOLO v7 水稻籽粒检测[ J] . 农业机械学报, 2023, 54 (12) : 253-260, 299. 
LIU Q H, YANG X Y, JIE H, et al. Rice grain detection based on YOLO v7 fusing of GhostNetV2[ J] . Transactions of the Chinese Society for Agricultural Machinery, 2023, 54(12) : 253-260, 299. 
[24] 胡瑛,刘狄昆,刘拯,等. 基于改进 YOLOv5 的复杂场景 下交通标志识别方法[ J] . 湖南工程学院学报( 自然科 学版) , 2024,34(2) :31-38. 
HU Y, LIU D K, LIU Z, et al. Traffic sign recognition method in complex scenes based on improved YOLOv5 [ J] . Journal of Hunan Institute of Engineering ( Natural Science Edition) , 2024,34(2) :31-38. 
[25] 刘磊. YOLOv4 交通信号灯检测[ J] . 电子世界, 2021 (15) : 92-94. 
LIU L. YOLOv4 traffic light detection [ J] . Electronics World, 2021(15) : 92-94. 
[26] 邓翔宇, 裴浩媛, 盛迎. 基于网络融合的改进 MobileViT 人脸表情识别[ J] . 计算机工程与科学, 2024, 46(6) : 1072-1080. 
DENG X Y, PEI H Y, SHENG Y. Facial expression recognition based on network fusion to improve MobileViT [ J] . Computer Engineering & Science, 2024, 46 ( 6) : 1072-1080. 
[27] 胡施 威, 邓 建 新, 王 浩 宇, 等. 基 于 改 进 EfficientNetB0 模型的葡萄叶部病害识别方法[ J] . 现代电子 技术, 2024, 47(15) : 73-80. 
HU S W, DENG J X, WANG H Y, et al. Grape leaf disease identification method based on improved EfficientNetB0 model[ J] . Modern Electronics Technique, 2024, 47(15) : 73-80.

更新日期/Last Update: 2025-07-13