Research on underwater robot ranging technology based on semantic segmentation and binocular vision

Area of Science:

• Robotics
• Computer Vision
• Optical Engineering

Background:

• Underwater vehicles require accurate target identification and distance measurement.
• Existing methods may lack efficiency or accuracy in complex underwater environments.
• Binocular vision and light refraction principles form the basis for underwater imaging models.

Purpose of the Study:

• To propose a new distance measurement system for underwater vehicles.
• To enhance underwater target identification and distance estimation.
• To provide a theoretical basis for underwater camera calibration.

Main Methods:

• Developed an underwater imaging model based on light refraction and binocular ranging.
• Implemented a system using semantic segmentation (Deeplabv3+) for underwater target identification.
• Integrated target maps generated by semantic segmentation with binocular ranging for distance measurement.

Main Results:

• The proposed method achieved comparable measurement accuracy to traditional binocular ranging.
• Measurement speed was increased by 30% compared to methods using original drawings.
• The error rate of the distance measurement was controlled within 5%.

Conclusions:

• The developed system effectively meets the operational needs of underwater robots.
• Semantic segmentation significantly improves the efficiency of binocular ranging for underwater targets.
• The system provides a robust solution for underwater vehicle navigation and target interaction.