A Fish-like Binocular Vision System for Underwater Perception of Robotic Fish.

Area of Science:

• Robotics
• Biomimetics
• Computer Vision

Background:

• Biological fish possess advanced visual systems enabling broad-spectrum perception.
• Current underwater robots often have limited fields of view and require protective casings.
• Fish-inspired designs offer potential for enhanced underwater robotic vision.

Purpose of the Study:

• To develop a fish-like binocular vision system for underwater robots.
• To improve the visual perception capacity and operational range of underwater robots.
• To address challenges of distortion and parallax in wide-field underwater vision.

Main Methods:

• Designed a fish-like binocular vision system based on biological principles for streamlined shapes and visual field overlap.
• Developed a visual field stitching algorithm to merge binocular images and correct distortion/parallax.
• Implemented an orientation alignment method with yaw and pitch angle scales for object reference.

Main Results:

• The fish-like vision system achieved a horizontal field of view of 306.56° in underwater experiments.
• Validated the effectiveness of the visual field stitching algorithm for seamless image merging.
• Confirmed the utility of the orientation alignment method for object orientation reference.

Conclusions:

• The developed vision system significantly advances underwater robotic perception capabilities.
• The fish-inspired approach provides a novel and effective method for wide-field underwater vision.
• This work offers valuable insights for designing future biomimetic visual systems for marine applications.