Acta Photonica Sinica, Volume. 54, Issue 1, 0110002(2025)
Topology-Based Bidirectional Radar-Optical Fusion Algorithm for USV Detection at Sea
Figures & Tables(19)
Fig. 3. Radar camera topology secondary fusion algorithm framework
Fig. 6. Comparison of fitting effect between PROSAC algorithm and least squares algorithm
Table 1. Experimental configuration
View tableView in ArticleTable 1. Experimental configuration
Configuration Parameter System environment Windows 11 GPU RTX 406 0 Accelerated environment CUDA 12.1 Libraries PyTorch 1.11.0 Language Python 3.8
Table 2. Detailed parameters for different categories
View tableView in ArticleTable 2. Detailed parameters for different categories
Net Target category P/% R/% mAP@0.5/% Map comparison/% Speed（ms）/fps YOLOv7-tiny All 91 86.4 88.3 -Base 6.5/153.8 Experimental_Ship 95.4 1 99.5 Cargo 93.1 92.6 96.1 Floater 73.6 70.8 69.3 Crane_Ship 94.6 97.1 99.3 Speedboat 84 71.4 74.1 Fishing_Boat 1 86.4 91.5 RDC-YOLOv7-tiny All 96.1 90.1 93 ↑4.7 7.3/140 Experimental_Ship 93 1 99.6 ↑0.1 Cargo 96.7 95.7 98.2 ↑2.1 Floater 1 62.9 78.3 ↑9 Crane_Ship 99 1 99.6 ↑0.3 Speedboat 94.7 82.1 82.8 ↑8.7 Fishing_Boat 93.2 1 99.5 ↑8
Table 3. Comparison of YOLO lightweight models
View tableView in ArticleTable 3. Comparison of YOLO lightweight models
Model mAP@0.5/% Speed（ms）/FPS Map Comparison/% FPS Comparison RDC-YOLOv7-tiny 93 7.3/140 -Base -Base YOLOv8n 93.7 8.2/122.5 ↑0.7 ↓17.5 YOLOv9s 94.2 12.6/79.4 ↑1.2 ↓60.6
Table 4. Control-experiment
View tableView in ArticleTable 4. Control-experiment
Scale factor s=1 s=1.5 s=2 s=3 s=4 s=5 s=6 Time/s 0.07 0.13 0.22 0.46 0.54 0.74 0.97 Omission factor/% 26.3 10.5 0 0 0 0 0
Table 5. Algorithm execution time comparison
View tableView in ArticleTable 5. Algorithm execution time comparison
Time No coarse correlation gate/s Coarse correlation gate/s Reduction ratio/% Average reduction ratio/% Time 1 2.596 0.304 ↓88.29 ↓88.37 Time 2 0.873 0.093 ↓89.35 Time 3 1.388 0.174 ↓87.46
Table 6. Scene 1
View tableView in ArticleTable 6. Scene 1
IoU fusion Comprehensive similarity fusion PROSAC straight line fitting Coarse correlation gate Bidirectional fusion of unfused data Accuracy/% Comparison/% √ 85.92 -base √ 90.76 ↑4.84 √ √ 93.55 ↑7.63 √ √ √ 93.49 ↑7.57 √ √ √ √ 94.21 ↑8.29
Table 7. Scene 2
View tableView in ArticleTable 7. Scene 2
IoU fusion Comprehensive similarity fusion PROSAC straight line fitting Coarse correlation gate Bidirectional fusion of unfused data Accuracy/% Comparison/% √ 80.38 -base √ 87.32 ↑6.94 √ √ 90.03 ↑9.68 √ √ √ 89.97 ↑9.59 √ √ √ √ 90.52 ↑10.14
Table 8. Scene 3
View tableView in ArticleTable 8. Scene 3
IoU fusion Comprehensive similarity fusion PROSAC straight line fitting Coarse correlation gate Bidirectional fusion of unfused data Accuracy/% Comparison/% √ 90.42 base √ 94.26 ↑3.84 √ √ 95.89 ↑5.47 √ √ √ 95.89 ↑5.47 √ √ √ √ 96.23 ↑5.81
Tools
Get Citation
Copy Citation Text
Hao TANG, Xuan PENG, Wei XIONG, Yaqi CUI, Jianqiu HU, Huiyuan XING. Topology-Based Bidirectional Radar-Optical Fusion Algorithm for USV Detection at Sea[J]. Acta Photonica Sinica, 2025, 54(1): 0110002
Paper Information
Category:
Received: Jun. 25, 2024
Accepted: Sep. 12, 2024
Published Online: Mar. 5, 2025
The Author Email: Wei XIONG (xiongwei@csif.org.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20