Journal of Applied Optics, Volume. 46, Issue 3, 505(2025)

Small object detection algorithm of UAV for visible light images based on YOLO-SCAT

Haiyong CHEN1, Boyang LIU1, and Xingwei YAN2,3、*
Author Affiliations
  • 1School of Artificial Intelligence, Hebei University of Technology, Tianjin 300401, China
  • 2College of Electronic Science, National University of Defense Technology, Changsha 410073, China
  • 3Tianjin Institute of Advanced Technology, Tianjin 300459, China
  • show less
    Figures & Tables(14)
    Diagram of YOLO-SCAT algorithm model
    Diagram of SRU unit model
    Diagram of CRU unit model
    Diagram of CAM module model
    Diagram of SAM module model
    Diagram of CBAM module model
    Schematic diagram of ELAN module improvement
    Diagrams of algorithm detection results
    • Table 1. Comparison of overall performance between YOLO-SC, YOLO-AT, YOLO-SCAT and YOLOv7 on unmanned aerial vehicle datasets

      View table
      View in Article

      Table 1. Comparison of overall performance between YOLO-SC, YOLO-AT, YOLO-SCAT and YOLOv7 on unmanned aerial vehicle datasets

      ModelP/%R/%mAP@0.5/%mAP@0.5:0.95/%F1/%FPSGflops
      YOLOv793.986.190.450.289.880.6103.2
      YOLO-SC95.486.192.251.090.577.5103.1
      YOLO-AT(CBAM)93.888.993.151.691.380.0105.3
      YOLO-SCAT94.494.494.752.994.476.9103.8
    • Table 2. Comparative experimental results of ELAN module improvement

      View table
      View in Article

      Table 2. Comparative experimental results of ELAN module improvement

      BackboneP/%R/%mAP@0.5/%mAP@0.5:0.95/%F1/%FPS
      YOLOv793.986.190.450.289.880.6
      a92.687.590.746.490.079.4
      b92.687.491.347.889.978.7
      c96.987.490.949.894.178.1
      d95.486.192.251.090.577.5
    • Table 3. Introduction of attention mechanism experiment in original YOLOv7 %

      View table
      View in Article

      Table 3. Introduction of attention mechanism experiment in original YOLOv7 %

      AttentionPRmAP@0.5
      Original93.986.190.4
      SE92.990.393.0
      ECA85.791.792.4
      EMA91.893.093.0
      CBAM93.888.993.1
    • Table 4. Comparative experimental results of attention mechanism %

      View table
      View in Article

      Table 4. Comparative experimental results of attention mechanism %

      AttentionPRmAP@0.5mAP@0.5:0.95F1
      SE90.794.493.651.392.5
      ECA94.494.494.550.294.4
      EMA91.281.691.247.886.1
      CBAM94.494.494.752.994.4
    • Table 5. YOLO-SCAT generalization validation experiment

      View table
      View in Article

      Table 5. YOLO-SCAT generalization validation experiment

      ModelP/%R/%mAP@0.5/%F1/%FPS
      YOLOv792.372.790.981.380.6
      YOLO-SCAT93.993.995.093.976.9
    • Table 6. Comparative experimental results of YOLO-SCAT and mainstream target detection algorithms

      View table
      View in Article

      Table 6. Comparative experimental results of YOLO-SCAT and mainstream target detection algorithms

      ModelP/%R/%mAP@0.5/%F1/%FPS
      Faster-RCNN28.873.954.441.018.2
      SSD86.967.870.876.143.8
      YOLOX86.766.170.975.040.7
      YOLOv584.086.486.185.140.0
      YOLOv793.986.190.489.880.6
      Ours94.494.494.794.476.9
    Tools

    Get Citation

    Copy Citation Text

    Haiyong CHEN, Boyang LIU, Xingwei YAN. Small object detection algorithm of UAV for visible light images based on YOLO-SCAT[J]. Journal of Applied Optics, 2025, 46(3): 505

    Download Citation

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

    Category: SPECIAL COLUMN ON UNMANNED INTELLIGENT SENSING TECHNOLOGY

    Received: Jan. 29, 2024

    Accepted: --

    Published Online: May. 28, 2025

    The Author Email: Xingwei YAN (晏行伟)

    DOI:10.5768/JAO202546.0311004

    Topics