Seeing through arthropod eyes: An AI-assisted, biomimetic approach for high-resolution, multi-task imaging

Area of Science:

• Biomimetics
• Computer Vision
• Robotics

Background:

• Arthropods possess advanced compound eyes with exceptional visual processing capabilities.
• Digital imaging can be enhanced by mimicking natural visual systems for wide-angle vision and depth perception.
• Limitations in current digital imaging include low resolving power, which biomimicry can address.

Purpose of the Study:

• To develop an artificial intelligence-assisted biomimetic system modeled after arthropod vision.
• To create a high-performance biomimetic compound eye camera.
• To achieve high-fidelity image reconstruction, 3D position prediction, and object recognition.

Main Methods:

• Development of a biomimetic compound eye camera with 4.3-megapixel resolution.
• Implementation of a multistage neural network for image processing and analysis.
• Utilizing rich visual information for computational imaging and target tracking.

Main Results:

• The camera captures full-color panoramic images with a 165° viewing angle and 40-micrometer resolving power.
• The system demonstrates high-fidelity image reconstruction, precise 3D position prediction, and accurate classification.
• Simultaneous 3D motion tracking of multiple miniature targets was achieved.

Conclusions:

• The AI-assisted biomimetic visual system provides a novel computational panoramic imaging solution.
• This technology advances applications in industry, medicine, and robotics by mimicking natural visual systems.
• The system offers a promising approach to combine wide-angle vision and depth perception with high accuracy.