Research on Defect Detection Method of Fusion Reactor Vacuum Chamber Based on Photometric Stereo Vision

Guodong Qin, Haoran Zhang, Yong Cheng, Youzhi Xu, Feng Wang, Shijie Liu, Xiaoyan Qin, Ruijuan Zhao, Congju Zuo, Aihong Ji2024 · DOI

TL;DR

This paper introduces a novel approach for detecting and reconstructing defects within the dim, challenging environment of a fusion reactor vacuum chamber. By combining an improved low-light image enhancement technique with a photometric stereo vision-based 3D reconstruction method, the system achieves accurate defect localization and dimensional analysis, crucial for maintaining the integrity of these complex machines.

What Problem Does This Paper Address?

Fusion reactors, essential for future energy production, rely on vacuum chambers that are inherently difficult to inspect. The internal surfaces, particularly the first wall and target plates, are subjected to extreme thermal and electromagnetic loads, leading to surface degradation, cracks, and other failures. These defects can compromise the reactor's performance and safety. However, the vacuum chamber's design and lack of adequate internal lighting create extremely dim conditions, making visual inspection and defect detection a significant challenge. Existing non-destructive testing methods are often insufficient for precise localization and 3D reconstruction of these defects, especially fine cracks, within such an environment.

Methods

The paper proposes a two-pronged approach:

  1. Improved Multi-Scale Retinex Low-Light Image Enhancement: To overcome the dimness, the authors developed an enhanced Retinex-based algorithm. This method adaptively adjusts weights across different scales to better recover details in low-light images while preserving image clarity and contrast. This is crucial for making otherwise invisible defects visible.
  2. Photometric Stereo Vision for Defect Reconstruction: For 3D reconstruction, the paper leverages photometric stereo vision. This technique involves capturing multiple images of the defect under different, controlled light source directions. By analyzing how the surface reflects light from these varied angles, the system can calculate the surface normal vectors and, subsequently, the 3D shape of the defect. To optimize this process, they also designed a simulation system to determine the ideal placement and configuration of light sources for uniform and efficient illumination within the chamber.

Datasets

Not discussed in the paper. The experiments were conducted on a simulated matte plate with artificially introduced defects.

Key Insights and Ideas

  • Adaptive Retinex for Extreme Low-Light: The standard Retinex algorithm struggles with highly uneven illumination. The adaptive weighting across multiple scales in their improved method is key to recovering details in the fusion reactor's challenging low-light conditions without introducing excessive noise.
  • Optimized Lighting is Crucial for Photometric Stereo: The paper highlights that the effectiveness of photometric stereo vision is highly dependent on the illumination strategy. Their simulation system for light source positioning is a practical innovation, reducing the need for extensive physical experimentation.
  • Combining Enhancement and Reconstruction: The true power of the system lies in the synergy between enhancing the dim images first and then applying photometric stereo. This allows for the detection of defects that would be missed by either method alone.

Key Results

The proposed system demonstrated significant improvements:

  • Image Enhancement: The enhanced low-light images showed broader gray levels, improved brightness, and better contrast, making defects more discernible.
  • Defect Reconstruction Accuracy: The system achieved a maximum depth error of less than 24.0% and a maximum width error of less than 15.3% for defect reconstruction.
  • Practical Implementation: A robotic platform with a binocular stereo-vision camera was successfully built and used for experimental validation.

Limitations and Future Work

  • Limitations: The authors acknowledge that the shape of reconstructed defects can exhibit some aberration. This is attributed to:
  • Restrictions in the integrable edges required for photometric stereo vision.
  • Camera viewing angle limitations, which can lead to shadows and highlights that affect gradient map accuracy.
  • Future Work: The researchers plan to continue investigating defect detection algorithms for fusion reactor vacuum chambers. Future experiments will focus on defect detection under varying illumination conditions, utilizing data from the China Fusion Engineering Test Reactor.

Summary updated

BibTeX
@misc{qin2024research,
  author = {Guodong Qin and Haoran Zhang and Yong Cheng and Youzhi Xu and Feng Wang and Shijie Liu and Xiaoyan Qin and Ruijuan Zhao and Congju Zuo and Aihong Ji},
  title = {{Research on Defect Detection Method of Fusion Reactor Vacuum Chamber Based on Photometric Stereo Vision}},
  year = {2024},
  doi = {10.3390/s24196227},
}

Loading related content...

Figure