We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Underwater Target Detection Based on Improved YOLOv7.
- Authors
Junshang Fu; Ying Tian
- Abstract
Underwater target detection is an important part of marine exploration. However, in complex underwater environments due to factors like light absorption and scattering, as well as variations in water quality and clarity. These challenges result in inaccurate target feature extraction, sluggish detection speeds, and insufficient robustness in the detection methods. In order to address these issues, an enhanced YOLOv7 network (YOLOv7-SPNW-D) is proposed for underwater target detection in this study. The SPD-MP module structure replaces the MP module in the neck network to capture small targets and enhance detection accuracy. A novel NWD loss function is employed to facilitate smoother extraction of small target features. This enhances feature extraction and improves network inference speed. Additionally, incorporating a small target detection module enables the providing of more comprehensive small target information within a deep feature map. This, in turn, improves the capture of small target features in complex backgrounds, and avoids feature loss and enhancing model exactness. Through ablation experiments on the URPC dataset, it is shown that the improved YOLOv7-SPNW-D algorithm performs better than the original YOLOv7 algorithm, with the mAP50 value increased to 87.0%, proving the effectiveness of this method. In conclusion, the improved YOLOv7-SPNW-D model is more suitable for underwater marine organism target detection.
- Subjects
WATER quality; LIGHT absorption; LIGHT scattering; MARINE organisms; MARINE resources; SUBMERGED structures; FEATURE extraction
- Publication
IAENG International Journal of Computer Science, 2024, Vol 51, Issue 4, p422
- ISSN
1819-656X
- Publication type
Article