A deep learning framework based on structured space model for detecting small objects in complex underwater...

Area of Science:

• Marine Biology
• Computer Vision
• Robotics

Background:

• Effective marine ecosystem monitoring relies on accurate underwater target detection.
• Current methods face challenges in balancing detection accuracy, model efficiency, and real-time performance.
• Small marine organisms are particularly difficult to detect in underwater environments.

Purpose of the Study:

• To propose an innovative approach for small target detection in underwater environments.
• To develop a high-accuracy, lightweight detection model for marine life monitoring.
• To improve the efficiency and applicability of underwater detection models for robotic deployment.

Main Methods:

• Combined the Structured Space Model (SSM) with feature enhancement techniques.
• Developed a lightweight detection model named UWNet.
• Evaluated UWNet's performance on detecting small marine organisms like starfish and scallops.

Main Results:

• UWNet demonstrated excellent detection accuracy, especially for challenging organisms.
• The model significantly reduced parameters (5% to 390%) compared to other models.
• Achieved substantial improvements in computational efficiency while maintaining high detection accuracy.

Conclusions:

• UWNet offers a superior solution for underwater small target detection.
• The model's lightweight design makes it suitable for deployment on underwater robots.
• This advancement aids in more effective marine life monitoring and ecosystem preservation.