Active Autofocusing Method With Plane Tilt Correction for AR Microdisplay Inspection

Abstract

An active autofocusing technology with plane tilt correction is proposed for the first time for use in display application scenarios. This technology is based on the variance of the Laplacian operator and machine vision principles. A deep learning-based multilayer perceptual classification neural network is proposed to quickly determine the tilt state of the plane and facilitate device-camera alignment. In testing, the neural network achieved 99.3% accuracy. The technology has demonstrated feasibility for leveling and focusing in commercial microdisplays for augmented reality (AR) applications within the micrometer-scale focus range through experimental validation. Our proposed autofocusing method is suitable for industrial inspection and offers the advantages of a lower risk of sample damage caused by auxiliary components and easy integration.