Optics and Precision Engineering, Volume. 33, Issue 8, 1289(2025)
Remote sensing object detection algorithm based on ultra fusion residual marching geometric perception
Figures & Tables(9)
Fig. 6. NWPU-VHR-10 and RSOD data set ablation experiment target detection effect diagram
Table 1. Experimental results contrasting
View tableView in ArticleTable 1. Experimental results contrasting
Models\Metrics NWPU-VHR-10 data set RSOD data set mPrecision mRecall mAP mF1 Score mPrecision mRecall mAP mF1 Score DETR 0.569 4 0.656 6 0.572 5 0.607 0 0.731 2 0.884 5 0.863 5 0.792 5 SSD 0.851 4 0.757 7 0.843 9 0.781 0 0.947 4 0.768 3 0.912 2 0.800 0 YOLOv5 0.912 1 0.896 2 0.922 9 0.901 0 0.932 3 0.783 2 0.906 4 0.792 5 HIC-YOLOv5［ 46 ］- - 0.876 0 - - - 0.896 0 - Drone-YOLO［ 46 ］- - 0.927 0 - - - 0.906 0 - Xu et al［ 46 ］- - 0.944 0 - - - 0.930 0 - YOLOv7 0.919 7 0.811 8 0.926 9 0.849 0 0.945 4 0.819 2 0.908 0 0.855 0 YOLOv7-tiny - - 0.921 0 - - - 0.945 0 - Tang et al［ 47 ］- - 0.971 6 - - - 0.962 7 - Wang et al［ 48 ］- - 0.916 5 - - - 0.901 6 - Cheng et al［ 49 ］- - 0.938 7 - - - 0.862 9 - YOLOv8 0.953 9 0.914 1 0.968 3 0.930 0 0.850 5 0.925 6 0.949 5 0.877 5 Zhang et al［ 50 ］0.934 0 0.902 0 0.940 0 0.917 0 0.963 0 0.927 0 0.944 0 0.971 0 YOLOv9［ 51 ］- - - - 0.934 5 0.896 5 0.920 5 0.915 3 YOLOv10［ 51 ］- - - - 0.944 7 0.910 2 0.933 5 0.927 3 SPPF-Mamba［ 52 ］- - 0.901 0 - - - 0.945 3 - Li et al［ 53 ］- - 0.926 0 - - - 0.951 0 - MLA-YOLO［ 54 ］- - 0.946 8 - - - 0.939 3 - YGNet［ 55 ］- - 0.889 0 - - - 0.962 0 - PR-Deformable DETR［ 56 ］- - 0.883 0 - - - 0.951 0 - LMATDet［ 57 ］- - 0.954 3 - - - 0.949 8 - SEB-YOLO［ 58 ］- - 0.935 0 0.939 0 - YOLO-UG 0.961 6 0.915 8 0.972 4 0.934 0 0.967 5 0.922 6 0.969 9 0.942 5
Table 2. Results of ablation experiment on NWPU-VHR-10 data set
View tableView in ArticleTable 2. Results of ablation experiment on NWPU-VHR-10 data set
UFRPM GFI MFF mPrecision mRecall mAP mF1 Score Params/M Complexity/GMac × × × 0.919 7 0.811 8 0.926 9 0.849 0 50.60 94.35 √ × × 0.931 6 0.865 0 0.950 6 0.895 0 50.60 105.06 √ √ × 0.941 8 0.871 9 0.956 9 0.901 0 50.60 105.06 √ √ √ 0.961 6 0.915 8 0.972 4 0.934 0 50.88 106.88
Table 3. Results of ablation experiment on RSOD data set
View tableView in ArticleTable 3. Results of ablation experiment on RSOD data set
UFRPM GFI MFF mPrecision mRecall mAP mF1 Score Params/M Complexity/GMac × × × 0.945 4 0.819 2 0.908 0 0.855 0 50.56 94.28 √ × × 0.901 1 0.908 8 0.954 3 0.902 5 50.56 105.00 √ √ × 0.912 0 0.878 9 0.956 3 0.887 5 50.56 105.00 √ √ √ 0.967 5 0.922 6 0.969 9 0.942 5 50.84 106.81
Tools
Get Citation
Copy Citation Text
Chenshuai BAI, Xiaofeng BAI, Kaijun WU, Haowen WANG. Remote sensing object detection algorithm based on ultra fusion residual marching geometric perception[J]. Optics and Precision Engineering, 2025, 33(8): 1289
Paper Information
Category:
Received: Aug. 30, 2024
Accepted: --
Published Online: Jul. 1, 2025
The Author Email: Kaijun WU (wkj@mail.lzjtu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20