Anti-distortion bioinspired camera with an inhomogeneous photo-pixel array.

Area of Science:

• Optics and Photonics
• Machine Vision
• Flexible Electronics

Background:

• Bioinspired cameras with single lenses and curved image sensors offer economical solutions for space-constrained machine vision.
• Curved sensors, similar to the retina, aid in image focusing but introduce optical distortion.
• Image distortion can significantly impair machine vision performance, particularly in object recognition tasks.

Purpose of the Study:

• To develop an anti-distortion camera system for machine vision applications.
• To engineer a camera that corrects optical distortion inherent in curved image sensors.
• To demonstrate a novel method for distortion correction within the image sensor itself.

Main Methods:

• Designed a bioinspired camera utilizing a single lens and a curved photodiode array.
• Developed an anti-distortion engineering approach by arranging 4096 silicon photodiodes in a nonuniform, warped pattern on the curved surface.
• The photo-pixel distribution pattern was specifically warped to counteract the optical distortion introduced by the camera's optics.

Main Results:

• The developed camera effectively corrected image distortion across a 120° horizontal field of view.
• Acquired images exhibited no appreciable distortion.
• Neural network recognition accuracies confirmed the effectiveness of the distortion correction method.

Conclusions:

• A novel anti-distortion single-lens camera was successfully created using a specifically patterned curved photodiode array.
• Distortion correction achieved through photo-pixel array reconfiguration represents a form of in-sensor computing.
• This technology enhances the viability of curved sensor cameras for demanding machine vision applications.