A Side-Scan Sonar Seabed Target Detection Algorithm Based on YOLOv8-RDE
Side-scan sonar target detection is crucial in underwater exploration, but traditional algorithms suffer from inaccurate positioning, slow detection, and poor model generalization. To address these shortcomings, a side-scan sonar seabed target detection algorithm is proposed based on YOLOv8-RDE (Rep...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2025-01-01
|
| Series: | International Journal of Distributed Sensor Networks |
| Online Access: | http://dx.doi.org/10.1155/dsn/6543345 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850233936763944960 |
|---|---|
| author | Haoming Ji Daqi Zhu Mingzhi Chen |
| author_facet | Haoming Ji Daqi Zhu Mingzhi Chen |
| author_sort | Haoming Ji |
| collection | DOAJ |
| description | Side-scan sonar target detection is crucial in underwater exploration, but traditional algorithms suffer from inaccurate positioning, slow detection, and poor model generalization. To address these shortcomings, a side-scan sonar seabed target detection algorithm is proposed based on YOLOv8-RDE (RepSiLU-DySample-eSE) in this paper. This algorithm uses a rotating frame with a certain angle to improve the accuracy. Specifically, we introduce a RepSiLU module to replace certain Conv modules, making the model have stronger real-time performance. DySample is used instead of traditional upsampling modules. And an eSE attention mechanism is integrated into the detection head. These enable the model to focus more on key targets and enhances accuracy. Finally, we linearly blend the target image with the seabed background image to construct a new dataset. This significantly enhances the model’s detection capability under complex seabed interference. Experimental results show that the improved model achieves an mAP50 of 0.917 on the expanded dataset. This is a 4.6% improvement over the original model. The frame rate reaches 175 FPS, which is a 13.6% increase over the original YOLOv8n-OBB model. The improved model excels in both accuracy and speed. It is well-suited for real-time detection tasks in complex underwater environments. |
| format | Article |
| id | doaj-art-980341c7089343c9801d225bc90d2839 |
| institution | OA Journals |
| issn | 1550-1477 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Wiley |
| record_format | Article |
| series | International Journal of Distributed Sensor Networks |
| spelling | doaj-art-980341c7089343c9801d225bc90d28392025-08-20T02:02:47ZengWileyInternational Journal of Distributed Sensor Networks1550-14772025-01-01202510.1155/dsn/6543345A Side-Scan Sonar Seabed Target Detection Algorithm Based on YOLOv8-RDEHaoming Ji0Daqi Zhu1Mingzhi Chen2Laboratory of Underwater Vehicles and Intelligent SystemsLaboratory of Underwater Vehicles and Intelligent SystemsLaboratory of Underwater Vehicles and Intelligent SystemsSide-scan sonar target detection is crucial in underwater exploration, but traditional algorithms suffer from inaccurate positioning, slow detection, and poor model generalization. To address these shortcomings, a side-scan sonar seabed target detection algorithm is proposed based on YOLOv8-RDE (RepSiLU-DySample-eSE) in this paper. This algorithm uses a rotating frame with a certain angle to improve the accuracy. Specifically, we introduce a RepSiLU module to replace certain Conv modules, making the model have stronger real-time performance. DySample is used instead of traditional upsampling modules. And an eSE attention mechanism is integrated into the detection head. These enable the model to focus more on key targets and enhances accuracy. Finally, we linearly blend the target image with the seabed background image to construct a new dataset. This significantly enhances the model’s detection capability under complex seabed interference. Experimental results show that the improved model achieves an mAP50 of 0.917 on the expanded dataset. This is a 4.6% improvement over the original model. The frame rate reaches 175 FPS, which is a 13.6% increase over the original YOLOv8n-OBB model. The improved model excels in both accuracy and speed. It is well-suited for real-time detection tasks in complex underwater environments.http://dx.doi.org/10.1155/dsn/6543345 |
| spellingShingle | Haoming Ji Daqi Zhu Mingzhi Chen A Side-Scan Sonar Seabed Target Detection Algorithm Based on YOLOv8-RDE International Journal of Distributed Sensor Networks |
| title | A Side-Scan Sonar Seabed Target Detection Algorithm Based on YOLOv8-RDE |
| title_full | A Side-Scan Sonar Seabed Target Detection Algorithm Based on YOLOv8-RDE |
| title_fullStr | A Side-Scan Sonar Seabed Target Detection Algorithm Based on YOLOv8-RDE |
| title_full_unstemmed | A Side-Scan Sonar Seabed Target Detection Algorithm Based on YOLOv8-RDE |
| title_short | A Side-Scan Sonar Seabed Target Detection Algorithm Based on YOLOv8-RDE |
| title_sort | side scan sonar seabed target detection algorithm based on yolov8 rde |
| url | http://dx.doi.org/10.1155/dsn/6543345 |
| work_keys_str_mv | AT haomingji asidescansonarseabedtargetdetectionalgorithmbasedonyolov8rde AT daqizhu asidescansonarseabedtargetdetectionalgorithmbasedonyolov8rde AT mingzhichen asidescansonarseabedtargetdetectionalgorithmbasedonyolov8rde AT haomingji sidescansonarseabedtargetdetectionalgorithmbasedonyolov8rde AT daqizhu sidescansonarseabedtargetdetectionalgorithmbasedonyolov8rde AT mingzhichen sidescansonarseabedtargetdetectionalgorithmbasedonyolov8rde |