Related Experiment Video
Updated: Jun 6, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
470
A Dynamic Interference Detection Method of Underwater Scenes Based on Deep Learning and Attention Mechanism.
Shuo Shang1, Jianrong Cao1, Yuanchang Wang1
1School of Information and Electrical Engineering, Shandong Jianzhu University, Jinan 250101, China.
Biomimetics (Basel, Switzerland)
|November 26, 2024
Summary
This study introduces an improved YOLOv8 algorithm for underwater dynamic target detection, enhancing feature extraction and loss functions. The new method achieves 95.1% mAP, improving accuracy in complex underwater environments.
Area of Science:
- Robotics
- Computer Vision
- Artificial Intelligence
Background:
- Underwater robot vision systems face challenges in 3D reconstruction due to high dynamic interference and poor lighting.
- Existing target detection algorithms are insufficient for complex underwater environments.
Purpose of the Study:
- To propose an improved YOLOv8 algorithm for accurate underwater dynamic target detection.
- To enhance the performance of underwater robot vision systems.
Main Methods:
- Modified the feature extraction layer of the YOLOv8 network by improving the Bottleneck convolutional structure.
- Integrated an improved SE attention mechanism for enhanced feature extraction.
- Replaced the CIoU loss function with MPDIoU loss to accelerate model convergence.
Main Results:
- The improved YOLOv8 algorithm achieved a mean Average Precision (mAP) of 95.1%.
- Demonstrated superior accuracy in detecting underwater dynamic targets, particularly small ones in complex scenes.
Conclusions:
- The proposed algorithm significantly improves underwater dynamic target detection accuracy.
- The enhancements provide a more robust solution for underwater robot vision systems.
