Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LD-SLRO: Latent Diffusion Structured Light for 3-D Reconstruction of Highly Reflective Objects

Sanghoon Jeon, Gihyun Jung, Suhyeon Ka, Jae-Sang Hyun

TL;DR

LD-SLRO tackles the persistent challenge of fringe projection profilometry on highly reflective, low-roughness surfaces where specular reflection distorts fringes. It introduces a latent-diffusion framework conditioned on two dedicated encoders—Diffuse Reflection Autoencoder and Specular Reflection Encoder—and a attention-guided denoiser to restore fringe patterns in a latent space, enabling flexible input/output fringe configurations. The method demonstrates superior fringe restoration and 3-D reconstruction accuracy versus state-of-the-art baselines, reducing RMSE from $1.8176\ \mathrm{mm}$ to $0.9619\ \mathrm{mm}$ on challenging objects. By separating diffuse and specular cues and leveraging a diffusion prior, LD-SLRO provides robust measurements under severe reflections, with practical impact for industrial inspection and automation pipelines.

Abstract

Fringe projection profilometry-based 3-D reconstruction of objects with high reflectivity and low surface roughness remains a significant challenge. When measuring such glossy surfaces, specular reflection and indirect illumination often lead to severe distortion or loss of the projected fringe patterns. To address these issues, we propose a latent diffusion-based structured light for reflective objects (LD-SLRO). Phase-shifted fringe images captured from highly reflective surfaces are first encoded to extract latent representations that capture surface reflectance characteristics. These latent features are then used as conditional inputs to a latent diffusion model, which probabilistically suppresses reflection-induced artifacts and recover lost fringe information, yielding high-quality fringe images. The proposed components, including the specular reflection encoder, time-variant channel affine layer, and attention modules, further improve fringe restoration quality. In addition, LD-SLRO provides high flexibility in configuring the input and output fringe sets. Experimental results demonstrate that the proposed method improves both fringe quality and 3-D reconstruction accuracy over state-of-the-art methods, reducing the average root-mean-squared error from 1.8176 mm to 0.9619 mm.

LD-SLRO: Latent Diffusion Structured Light for 3-D Reconstruction of Highly Reflective Objects

TL;DR

LD-SLRO tackles the persistent challenge of fringe projection profilometry on highly reflective, low-roughness surfaces where specular reflection distorts fringes. It introduces a latent-diffusion framework conditioned on two dedicated encoders—Diffuse Reflection Autoencoder and Specular Reflection Encoder—and a attention-guided denoiser to restore fringe patterns in a latent space, enabling flexible input/output fringe configurations. The method demonstrates superior fringe restoration and 3-D reconstruction accuracy versus state-of-the-art baselines, reducing RMSE from to on challenging objects. By separating diffuse and specular cues and leveraging a diffusion prior, LD-SLRO provides robust measurements under severe reflections, with practical impact for industrial inspection and automation pipelines.

Abstract

Fringe projection profilometry-based 3-D reconstruction of objects with high reflectivity and low surface roughness remains a significant challenge. When measuring such glossy surfaces, specular reflection and indirect illumination often lead to severe distortion or loss of the projected fringe patterns. To address these issues, we propose a latent diffusion-based structured light for reflective objects (LD-SLRO). Phase-shifted fringe images captured from highly reflective surfaces are first encoded to extract latent representations that capture surface reflectance characteristics. These latent features are then used as conditional inputs to a latent diffusion model, which probabilistically suppresses reflection-induced artifacts and recover lost fringe information, yielding high-quality fringe images. The proposed components, including the specular reflection encoder, time-variant channel affine layer, and attention modules, further improve fringe restoration quality. In addition, LD-SLRO provides high flexibility in configuring the input and output fringe sets. Experimental results demonstrate that the proposed method improves both fringe quality and 3-D reconstruction accuracy over state-of-the-art methods, reducing the average root-mean-squared error from 1.8176 mm to 0.9619 mm.
Paper Structure (18 sections, 15 equations, 7 figures, 3 tables, 2 algorithms)

This paper contains 18 sections, 15 equations, 7 figures, 3 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Principle
  3. Fringe Projection Profilometry
  4. Specular Reflection Induced Error
  5. Proposed Method
  6. Denoising Diffusion Probabilistic Model (DDPM)
  7. Network Architecture
  8. Diffuse Reflection Autoencoder
  9. Specular Reflection Encoder
  10. Denoiser Network
  11. Loss Function
  12. Fringe Pattern Design and Data Preparation
  13. Experiment
  14. Experimental Setup
  15. Ablation Experiment
  16. ...and 3 more sections

Figures (7)

  • Figure 1: Experimental setup and test objects. (a) Experimental system setup of FPP. (b) Photograph of highly reflective objects.
  • Figure 2: Typical fringe degradations on highly reflective surfaces. (a) Highly reflective fringe image (b) Non highly reflective fringe image (c) Fringe loss and distortion (d) Fringe contrast loss
  • Figure 3: The network architecture of the proposed LD-SLRO consists of three parts: a denoiser network, a pretrained diffuse reflection autoencoder, and a pretrained specular reflection encoder.
  • Figure 4: Effect of fringe pitch on specular blooming for highly reflective surfaces. (a) Captured binary fringe image with a pitch of 24 pixels. (b) Captured binary fringe image with a pitch of 36 pixels. (c) Intensity profiles indicating stronger blooming for the 24-pixel pattern and reduced blooming for the 36-pixel pattern.
  • Figure 5: Comparison of fringe image enhancement on highly reflective surfaces using 24-step PSP, HDRNet, DC-UNet, Y-FFC, and the proposed LD-SLRO. The proposed method more effectively suppresses specular highlights and restores high-quality fringe patterns, yielding clearer fringe patterns in overexposure regions.
  • ...and 2 more figures