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. 60, Issue 12, 1228007(2023)

Improved Aircraft Detection of Optical Remote Sensing Image Based on Faster R-CNN

Xin Yang, Qiong Wang, Yazhou Yao, and Zhenmin Tang*
Author Affiliations
  • School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing 210094, Jiangsu, China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (18)
    Equations (5)
    References (29)
    Cited By (2)
    Tools
    Paper Information
    Topics
    Figures & Tables(18)
    Faster R-CNN algorithm framework after basic module introduced

    Fig. 1. Faster R-CNN algorithm framework after basic module introduced

    Download full size
    Structure of ResNet50

    Fig. 2. Structure of ResNet50

    Download full size
    Structure of ResNet50 and FPN

    Fig. 3. Structure of ResNet50 and FPN

    Download full size
    Structure of CBAM

    Fig. 4. Structure of CBAM

    Download full size
    Cutting process of sliding window

    Fig. 5. Cutting process of sliding window

    Download full size
    Post-processing of prediction results

    Fig. 6. Post-processing of prediction results

    Download full size
    Examples of aircraft categories

    Fig. 7. Examples of aircraft categories

    Download full size
    Distribution of aircraft number by class

    Fig. 8. Distribution of aircraft number by class

    Download full size
    Distribution of aircraft size

    Fig. 9. Distribution of aircraft size

    Download full size
    Feature map before adding channel attention

    Fig. 10. Feature map before adding channel attention

    Download full size
    Feature map after adding channel attention

    Fig. 11. Feature map after adding channel attention

    Download full size
    Feature map comparison before and after adding spatial attention.(a) Input; (b) feature map of channel 14 that is not added to CBAM; (c) feature map of channel 14 after CBAM added

    Fig. 12. Feature map comparison before and after adding spatial attention.(a) Input; (b) feature map of channel 14 that is not added to CBAM; (c) feature map of channel 14 after CBAM added

    Download full size
    Comparison before and after use of post-processing

    Fig. 13. Comparison before and after use of post-processing

    Download full size

    • Table 1. Performance comparison of feature extraction network before and after lightweight

      View table

      Table 1. Performance comparison of feature extraction network before and after lightweight

      Feature extraction networkmF1Size /106FPS
      ResNet5085.982241.916.3
      ResNet41-387.480225.818.2
      Res2Net5078.855941.614.1
      Res2Net41-383.005625.515.7
      ResNext10159.866260.010.6
      ResNext92-377.356844.411.4
      ResNet50+FPN-486.583041.417.1

    • Table 2. Performance comparison of different target detection algorithms before and after lightweight

      View table

      Table 2. Performance comparison of different target detection algorithms before and after lightweight

      MethodmF1Size /106FPS
      Faster R-CNN+ResNet5085.982241.916.3
      Faster R-CNN+ResNet41-387.480225.818.2
      Cascade R-CNN+ResNet5087.174069.210.7
      Cascade R-CNN+ResNet41-388.534653.112.9
      FCOS+ResNet5074.134832.116.9
      FCOS+ResNet41-378.877116.118.5

    • Table 3. Performance comparison of different target detection tasks before and after lightweight

      View table

      Table 3. Performance comparison of different target detection tasks before and after lightweight

      DatasetMethodmF1Size /106FPS
      RSAICP-planeFaster R-CNN+ResNet5085.982241.916.3
      Faster R-CNN+ResNet41-387.480225.818.2
      SAR-shipFaster R-CNN+ResNet5090.274841.933.1
      Faster R-CNN+ResNet41-391.161725.837.9
      CASIA-aircraftFaster R-CNN+ResNet5096.933042.125.5
      Faster R-CNN+ResNet41-397.324026.027.8

    • Table 4. Performance comparison of CBAM attention mechanism

      View table

      Table 4. Performance comparison of CBAM attention mechanism

      Feature extraction networkmF1Size /106FPS
      ResNet4187.480225.818.2
      ResNet41-3_CBAM-1687.767725.800515.3
      ResNet41-3_CBAM-888.780325.80115.3
      ResNet41-3_CBAM-487.974525.80215.3

    • Table 5. Experiment results of midline single frame prediction and multi-scale training

      View table

      Table 5. Experiment results of midline single frame prediction and multi-scale training

      Feature extraction networkmF1
      ResNet41-387.4802
      ResNet41-3_ms88.9047
      ResNet41-3_ms*88.9465
      ResNet41-3_CBAM-888.7803
      ResNet41-3_CBAM-8_ms88.7926
      ResNet41-3_CBAM-8_ms*88.9682
    Tools

    Get Citation

    Copy Citation Text

    Xin Yang, Qiong Wang, Yazhou Yao, Zhenmin Tang. Improved Aircraft Detection of Optical Remote Sensing Image Based on Faster R-CNN[J]. Laser & Optoelectronics Progress, 2023, 60(12): 1228007

    Download Citation

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

    Category: Remote Sensing and Sensors

    Received: May. 24, 2022

    Accepted: Jul. 4, 2022

    Published Online: Jun. 5, 2023

    The Author Email: Zhenmin Tang (tzm.cs@njust.edu.cn)

    DOI:10.3788/LOP221679

    Topics

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