一种基于改进YOLOv8的水下声纳图像识别方法及系统
By improving the YOLOv8 network model and combining a multidimensional parallel attention mechanism and a contrastive learning architecture, the problem of small target recognition in low-resolution and noisy environments in underwater sonar image recognition technology has been solved, achieving more efficient target recognition and lower resource consumption.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANJING UNIV OF INFORMATION SCI & TECH
- Filing Date
- 2025-01-26
- Publication Date
- 2026-07-17
AI Technical Summary
Existing underwater sonar image recognition technologies struggle to accurately identify small targets in low-resolution and high-noise environments, and existing algorithms suffer from problems such as large parameter counts, training degradation, and poor small target detection.
An improved YOLOv8 network model is adopted, which combines a multidimensional parallel attention mechanism and a contrastive learning architecture. The feature extraction capability is enhanced through the multidimensional parallel attention mechanism (MDPA), and the model weights are optimized by using a contrastive loss function to construct the YOLOv8-MDPA-CL network model.
It significantly improves the accuracy of target detail recognition and model generalization ability in complex underwater environments, enhances robustness, and reduces the number of parameters and computational complexity.
Smart Images
Figure CN119992305B_ABST