Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Irregularity Inspection using Neural Radiance Field

Tianqi Ding, Dawei Xiang

TL;DR

The paper tackles automated defect detection on large outdoor machinery by recasting it as a 3D irregularity inspection using Neural Radiance Fields (NeRF) to create digital twin models. It captures images of a standard reference rig and the field rig with a UAV, reconstructs two NeRF-based 3D models, and uses ICP to align them before a nearest-neighbor point cloud comparison to reveal 3D defects. The approach reduces memory demands, handles challenging lighting, and accelerates model generation, enabling safer, more objective inspections. The work demonstrates the method on synthetic and small-scale examples and outlines plans to scale to large-scale industrial scenarios with drone sampling and integrated neural-traditional modeling.

Abstract

With the increasing growth of industrialization, more and more industries are relying on machine automation for production. However, defect detection in large-scale production machinery is becoming increasingly important. Due to their large size and height, it is often challenging for professionals to conduct defect inspections on such large machinery. For example, the inspection of aging and misalignment of components on tall machinery like towers requires companies to assign dedicated personnel. Employees need to climb the towers and either visually inspect or take photos to detect safety hazards in these large machines. Direct visual inspection is limited by its low level of automation, lack of precision, and safety concerns associated with personnel climbing the towers. Therefore, in this paper, we propose a system based on neural network modeling (NeRF) of 3D twin models. By comparing two digital models, this system enables defect detection at the 3D interface of an object.

Irregularity Inspection using Neural Radiance Field

TL;DR

The paper tackles automated defect detection on large outdoor machinery by recasting it as a 3D irregularity inspection using Neural Radiance Fields (NeRF) to create digital twin models. It captures images of a standard reference rig and the field rig with a UAV, reconstructs two NeRF-based 3D models, and uses ICP to align them before a nearest-neighbor point cloud comparison to reveal 3D defects. The approach reduces memory demands, handles challenging lighting, and accelerates model generation, enabling safer, more objective inspections. The work demonstrates the method on synthetic and small-scale examples and outlines plans to scale to large-scale industrial scenarios with drone sampling and integrated neural-traditional modeling.

Abstract

With the increasing growth of industrialization, more and more industries are relying on machine automation for production. However, defect detection in large-scale production machinery is becoming increasingly important. Due to their large size and height, it is often challenging for professionals to conduct defect inspections on such large machinery. For example, the inspection of aging and misalignment of components on tall machinery like towers requires companies to assign dedicated personnel. Employees need to climb the towers and either visually inspect or take photos to detect safety hazards in these large machines. Direct visual inspection is limited by its low level of automation, lack of precision, and safety concerns associated with personnel climbing the towers. Therefore, in this paper, we propose a system based on neural network modeling (NeRF) of 3D twin models. By comparing two digital models, this system enables defect detection at the 3D interface of an object.
Paper Structure (12 sections, 11 figures, 1 algorithm)

This paper contains 12 sections, 11 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methods
  4. System Architecture
  5. Environment And Hardware Devices
  6. Software
  7. Hardware
  8. Iterative Closest Point and Point Cloud Comparison Algorithm
  9. Iterative Closest Point Algorithm
  10. Point Cloud Comparison Algorithm
  11. Experiments
  12. Conclusion

Figures (11)

  • Figure 1: Schematic representation of the NeRF procedure.[3]
  • Figure 2: AI Based Irregularity Inspection System Flowchart.
  • Figure 3: Canon EOS Rebel T4i Camera.
  • Figure 4: A illustration of ICP algorithm. Our goal is to find a transformation to make source and target model as close as possible.
  • Figure 5: Testing Tower Model.
  • ...and 6 more figures