《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (1): 292-300.DOI: 10.11772/j.issn.1001-9081.2024010125

• 多媒体计算与计算机仿真 • 上一篇    下一篇

融合注意力和上下文信息的遥感图像小目标检测算法

刘赏1(), 周煜炜1, 代娆1, 董林芳1, 刘猛2   

  1. 1.天津财经大学 理工学院,天津 300222
    2.河北省水文工程地质勘查院(河北省遥感中心),石家庄 050021
  • 收稿日期:2024-02-05 修回日期:2024-04-01 接受日期:2024-04-07 发布日期:2024-05-09 出版日期:2025-01-10
  • 通讯作者: 刘赏
  • 作者简介:周煜炜(1998—),女,天津人,硕士研究生,主要研究方向:图像处理、目标检测;
    代娆(1998—),女,天津人,硕士研究生,主要研究方向:图像处理、姿态估计;
    董林芳(1972—),女,河北张家口人,副教授,博士,主要研究方向:人工智能、自然语言处理;
    刘猛(1986—),男,河北石家庄人,硕士,主要研究方向:图像处理、遥感地质。
  • 基金资助:
    河北省财政项目(13000023P00F2D410374D);天津市科技计划项目(22ZLZKZF00480)

Small target detection algorithm in remote sensing images integrating attention and contextual information

Shang LIU1(), Yuwei ZHOU1, Rao DAI1, Linfang DONG1, Meng LIU2   

  1. 1.School of Science and Engineering,Tianjin University of Finance and Economics,Tianjin 300222,China
    2.Hydrogeological Engineering Geological Exploration Institute (Hebei Remote Sensing Center),Shijiazhuang Hebei 050021,China
  • Received:2024-02-05 Revised:2024-04-01 Accepted:2024-04-07 Online:2024-05-09 Published:2025-01-10
  • Contact: Shang LIU
  • About author:ZHOU Yuwei, born in 1998, M. S. candidate. Her research interests include image processing, object detection.
    DAI Rao, born in 1998, M. S. candidate. Her research interests include image processing, pose estimation.
    DONG Linfang, born in 1972, Ph. D., associate professor. Her research interests include artificial intelligence, natural language processing.
    LIU Meng, born in 1986, M. S. His research interests include image processing, remote sensing geology.
  • Supported by:
    Financial Project of Hebei Province(13000023P00F2D410374D);Science and Technology Program of Tianjin(22ZLZKZF00480)

摘要:

对多尺度的遥感图像进行小目标检测时,基于深度学习的目标检测算法容易出现误检和漏检的情况。这是因为此类算法的特征提取模块进行了多次的下采样操作;而且未能根据不同类别、不同尺度的目标关注所需的上下文信息。为了解决该问题,提出一种融合注意力和上下文信息的遥感图像小目标检测算法ACM-YOLO(Attention-Context-Multiscale YOLO)。首先,应用细粒度的查询感知稀疏注意力以减少小目标特征信息的丢失,从而避免漏检;其次,设计局部上下文增强(LCE)函数以更好地关注不同类别的遥感目标所需的上下文信息,从而避免误检;最后,使用加权双向特征金字塔网络(BiFPN)强化特征融合模块对遥感图像小目标的多尺度特征融合能力,从而改善算法检测效果。在DOTA数据集和NWPU VHR-10数据集上进行对比实验和消融实验,以验证所提算法的有效性和泛化性。实验结果表明,在2个数据集上所提算法的平均精确率均值(mAP) 分别达到了77.33%和96.12%,而相较于YOLOv5算法,召回率分别提升了10.00和7.50个百分点。可见,所提算法能有效提升mAP和召回率,减少误检和漏检。

关键词: 遥感图像, 小目标检测, 稀疏采样, 局部上下文信息增强, 多尺度特征融合

Abstract:

When detecting small targets in multi-scale remote sensing images, target detection algorithms based on deep learning are prone to false detection and missed detection. One of the reasons is that the feature extraction module carries out multiple down-sampling operations. The second reason is the failure to pay attention to the contextual information required by different categories and different scales of targets. To solve this problem, a small object detection algorithm in remote sensing images integrating attention and contextual information ACM-YOLO (Attention-Context-Multiscale YOLO) was proposed. Firstly, to reduce the loss of small target feature information, fine-grained query aware sparse attention was applied, thereby avoiding missed detection. Secondly, to pay more attention to the contextual information required by different categories of remote sensing targets, the Local Contextual Enhancement (LCE) function was designed, thereby avoiding false detection. Finally, to strengthen multi-scale feature fusion capability of the feature fusion module on small targets in remote sensing images, the weighted Bi-directional Feature Pyramid Network (BiFPN) was adopted, thereby improving detection effect of the algorithm. Comparison experiments and ablation experiments were performed on DOTA dataset and NWPU VHR-10 dataset to verify effectiveness and generalization of the proposed algorithm. Experimental results show that on the two datasets, the proposed algorithm has the mean Average Precision (mAP) reached 77.33% and 96.12% respectively, and the Recall increases by 10.00 and 7.50 percentage points, respectively, compared with YOLOv5 algorithm. It can be seen that the proposed algorithm improves mAP and recall effectively, which reduces false detection and missed detection.

Key words: remote sensing image, small target detection, sparse sampling, local contextual information enhancement, multi-scale feature fusion

中图分类号: