Journals Articles Conferences News About CLP
Submit a paper Email Alert
Search
Search
Articles
Journals
News
Advanced Search
Top Searches
2D MaterialsTransformation opticsQuantum PhotonicsSmall lasersSemiconductor UV PhotonicsSupersymmetric microring laser arrays

Laser & Optoelectronics Progress, Volume. 61, Issue 4, 0428009(2024)

Aircraft-Bunker Detection Method Based on Deep Learning in High-Resolution Remote-Sensing Images

Shushu Shi1,2, Yongqiang Chen1、*, Yingjie Wang1, and Chunle Wang1
Author Affiliations
  • 1Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100190, China
  • 2School of Electronic, Electrical and Communication Engineering, University of Chinese Academy of Sciences, Beijing 100049, China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (21)
    Equations (14)
    References (38)
    Cited By (2)
    Tools
    Paper Information
    Topics
    Figures & Tables(21)
    Google Earth images of different types of aircraft bunkers. (a) Tunnel hangar; (b) excavation hangar; (c) hangar; (d) revetment

    Fig. 1. Google Earth images of different types of aircraft bunkers. (a) Tunnel hangar; (b) excavation hangar; (c) hangar; (d) revetment

    Download full size
    Dataset sample images. (a) Objectives of different coverings; (b) multiple objectives under different backgrounds; (c) objectives under complex background

    Fig. 2. Dataset sample images. (a) Objectives of different coverings; (b) multiple objectives under different backgrounds; (c) objectives under complex background

    Download full size
    Objectives distribution in each image of dataset

    Fig. 3. Objectives distribution in each image of dataset

    Download full size
    Faster R-CNN framework

    Fig. 4. Faster R-CNN framework

    Download full size
    SSD framework

    Fig. 5. SSD framework

    Download full size
    RetinaNet framework

    Fig. 6. RetinaNet framework

    Download full size
    YOLOX framework

    Fig. 7. YOLOX framework

    Download full size
    Definition of rotated object representation

    Fig. 8. Definition of rotated object representation

    Download full size
    Ground truth. (a) Multi-objective distribution; (b) with confusing objectives; (c) multi-directional small objectives; (d) objectives under snow cover; (e) duplex aircraft bunkers

    Fig. 9. Ground truth. (a) Multi-objective distribution; (b) with confusing objectives; (c) multi-directional small objectives; (d) objectives under snow cover; (e) duplex aircraft bunkers

    Download full size
    Results of Faster R-CNN. (a) Multi-objective distribution; (b) with confusing objectives; (c) multi-directional small objectives; (d) objectives under snow cover; (e) duplex aircraft bunkers

    Fig. 10. Results of Faster R-CNN. (a) Multi-objective distribution; (b) with confusing objectives; (c) multi-directional small objectives; (d) objectives under snow cover; (e) duplex aircraft bunkers

    Download full size
    Results of SSD. (a) Multi-objective distribution; (b) with confusing objectives; (c) multi-directional small objectives; (d) objectives under snow cover; (e) duplex aircraft bunkers

    Fig. 11. Results of SSD. (a) Multi-objective distribution; (b) with confusing objectives; (c) multi-directional small objectives; (d) objectives under snow cover; (e) duplex aircraft bunkers

    Download full size
    Results of RetinaNet. (a) Multi-objective distribution; (b) with confusing objectives; (c) multi-directional small objectives; (d) objectives under snow cover; (e) duplex aircraft bunkers

    Fig. 12. Results of RetinaNet. (a) Multi-objective distribution; (b) with confusing objectives; (c) multi-directional small objectives; (d) objectives under snow cover; (e) duplex aircraft bunkers

    Download full size
    Results of YOLOv3. (a) Multi-objective distribution; (b) with confusing objectives; (c) multi-directional small objectives; (d) objectives under snow cover; (e) duplex aircraft bunkers

    Fig. 13. Results of YOLOv3. (a) Multi-objective distribution; (b) with confusing objectives; (c) multi-directional small objectives; (d) objectives under snow cover; (e) duplex aircraft bunkers

    Download full size
    Results of YOLOX. (a) Multi-objective distribution; (b) with confusing objectives; (c) multi-directional small objectives; (d) objectives under snow cover; (e) duplex aircraft bunkers

    Fig. 14. Results of YOLOX. (a) Multi-objective distribution; (b) with confusing objectives; (c) multi-directional small objectives; (d) objectives under snow cover; (e) duplex aircraft bunkers

    Download full size
    Results of YOLOX. (a) Side-by-side distribution of objectives; (b) centrally distributed multiple objectives; (c) multi-directional objectives

    Fig. 15. Results of YOLOX. (a) Side-by-side distribution of objectives; (b) centrally distributed multiple objectives; (c) multi-directional objectives

    Download full size
    Results of R-YOLOX with KLD loss. (a) Side-by-side distribution of objectives; (b) centrally distributed multiple objectives; (c) multi-directional objectives

    Fig. 16. Results of R-YOLOX with KLD loss. (a) Side-by-side distribution of objectives; (b) centrally distributed multiple objectives; (c) multi-directional objectives

    Download full size
    Results of YOLOX with Smooth L1 loss. (a) Side-by-side distribution of objectives; (b) centrally distributed multiple objectives; (c) multi-directional objectives

    Fig. 17. Results of YOLOX with Smooth L1 loss. (a) Side-by-side distribution of objectives; (b) centrally distributed multiple objectives; (c) multi-directional objectives

    Download full size
    Detection results of aircraft bunkers at an airport

    Fig. 18. Detection results of aircraft bunkers at an airport

    Download full size

    • Table 1. Parameter settings in model training

      View table

      Table 1. Parameter settings in model training

      ModelBackboneBatch sizeInput image sizeInitial learning rateIterations
      Faster R-CNNResNet504600×6000.0001150
      SSDVGG168300×3000.0020150
      RetinaNetResNet508600×6000.0001150
      YOLOv3DarkNet538640×6400.0010150
      YOLOX-sSPP-DarkNet538640×6400.0001150

    • Table 2. Comparison of detection accuracy of different model

      View table

      Table 2. Comparison of detection accuracy of different model

      ModelPrecision /%Recall /%AP50 /%F1FPSSize /MBTime /s
      Faster R-CNN78.2896.0095.580.8612.14108.170.61
      SSD90.4992.9095.460.9287.8090.610.19
      RetinaNet92.9495.0197.310.9428.60138.910.62
      YOLOv394.6496.0096.950.9529.54235.040.68
      YOLOX96.8595.4597.700.9642.9734.300.32

    • Table 3. The comparison of rotation detection accuracy

      View table

      Table 3. The comparison of rotation detection accuracy

      ModelLRegPrecision /%Recall /%AP50 /%F1
      YOLOXSmooth L1 loss86.6091.6992.060.89
      R-YOLOXKLD loss93.8494.5797.250.94
    Tools

    Get Citation

    Copy Citation Text

    Shushu Shi, Yongqiang Chen, Yingjie Wang, Chunle Wang. Aircraft-Bunker Detection Method Based on Deep Learning in High-Resolution Remote-Sensing Images[J]. Laser & Optoelectronics Progress, 2024, 61(4): 0428009

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Set citation alerts for article
    Save article for my favorites
    Paper Information

    Category: Remote Sensing and Sensors

    Received: Oct. 19, 2022

    Accepted: Dec. 15, 2022

    Published Online: Feb. 20, 2024

    The Author Email: Yongqiang Chen (clyq@163.com)

    DOI:10.3788/LOP222827

    CSTR:32186.14.LOP222827

    Topics

    laser devices and laser physics
    Lasers and Laser Optics
    Laser physics
    laser manufacturing
    Instrumentation, Measurement and Metrology