Answers are grounded in this paper and cite their sources.
Micrometer-scale defects in compact near-eye display (NED) optical modules induce light ray deflections that accumulate through multiple refractions and reflections, causing viewpoint-dependent localized distortions. Traditional detection methods require disassembly, compromising localization accuracy and risking secondary damage. We propose an inverse modeling approach that employs a convolutional neural network (CNN) trained on multi-view distortion vector fields (DVFs) to reconstruct defect geometry (including peak-to-valley (P-V) amplitude and 2D position) and identify the specific lens surface containing the defect. Experiments demonstrate that our method achieves micrometer-scale reconstruction accuracy and precise spatial localization without module disassembly. The reconstructed defect model provides a foundation for subsequent compensation of localized distortions, including the dynamic distortions induced by eye movement, offering a non-destructive solution for quality assurance in multi-lens NED systems.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: