[20p-C43-1~11] 13.1 Fundamental properties, surface and interface, and simulations of Si related materials

3:00 PM - 3:15 PM JST(6:00 AM - 6:15 AM UTC)English Presentation

[20p-C43-8] Synthetic Image Generation of Microstructure Surfaces Using Physically Based Rendering Techniques

Zhen-Wei Tsai¹, 〇(M1)Chao-Ching Ho¹ (1.Nat'l Taipei Uni. of Tech.)

Keywords:

Physics-Based Rendering,Microstructure Surface Imaging,Defect Detection

This study presents a novel approach for synthesizing microstructure surface images through physics-based rendering techniques. Traditional deep learning models for defect detection in microstructures require extensive defect data, which is often scarce due to high-yield production processes. To address this challenge, a physically based rendering method was adopted, utilizing rendering software to create a virtual environment and generate numerous synthetic images. This approach enables the modeling of microstructured surfaces under various viewing directions and lighting conditions.A 3D rendering engine was employed to set microstructure parameters, including groove size, depth, and spacing. The virtual environment allowed for arbitrary adjustments of light source types, intensities, camera sensor sizes, positions, and material properties. The optical path difference parameter was used to represent various structural colors resulting from interference between multiple waves. An interference shader based on established derivations was implemented to calculate the intensity of interference waves.To visualize color, the spectral distribution was converted to the RGB color system via the XYZ color system, integrating wavelengths in the visible region. The method demonstrated its capability to render diffraction and interference phenomena on microstructure surfaces accurately. Future work will focus on using quantitative metrics to evaluate the realism of rendered images compared to actual captured images.This research provides a robust framework for generating synthetic microstructure surface images, enhancing neural network training for defect detection and offering significant potential for applications in semiconductor manufacturing and optical characterization.