Underwater High-Precision 3D Reconstruction System Based on Rotating Scanning

Qingsheng Xue, Qian Sun, Fupeng Wang, Haoxuan Bai, Bai Yang, Qian Li2021 · DOI

TL;DR

This paper introduces a novel underwater high-precision 3D laser scanning system that utilizes a rotating scanning mechanism and a green line laser. To overcome the challenges of underwater imaging, they developed a refraction compensation algorithm, achieving reconstruction errors of less than 0.6 mm, making it a promising low-cost alternative to expensive professional sensors.

What Problem Does This Paper Address?

Traditional underwater imaging methods, like acoustic detection, often suffer from poor visualization and susceptibility to noise. Optical 3D reconstruction offers a more intuitive and accurate approach, with applications ranging from underwater engineering and military operations to marine research and archaeology. However, existing optical methods face significant challenges in underwater environments, primarily due to light refraction at the air-water interface and through waterproof enclosures. This refraction distorts the perceived position of objects, leading to significant errors in 3D reconstruction. The paper aims to address this by developing a high-precision underwater 3D scanning system and, crucially, a method to compensate for these refraction-induced errors.

Methods

The core of the system is a rotary scanning setup comprising a low-light underwater camera, a green line laser projector, and a scanning turntable. The system is designed for a field of view of 50° (vertical) × 360° (horizontal).

Key methodological contributions include:

  • System Calibration: A comprehensive calibration process is detailed, covering camera parameters (internal and external) and system structure parameters. This establishes the relationship between 2D image coordinates and 3D world coordinates. The Zhang Zhengyou calibration method is employed for camera calibration.
  • Refraction Compensation Algorithm: This is the paper's main innovation. It addresses the double refraction occurring at the waterproof cover. The algorithm compensates for:
  • Light Plane Refraction: Correcting the perceived position of the projected laser line due to refraction.
  • CCD Image Refraction: Correcting the perceived position of the object's image on the camera sensor. This compensation is achieved through a series of geometric calculations based on Snell's law and the system's optical geometry.

  • Rotating Scanning: The system employs a turntable to achieve a 360° horizontal scan, enabling comprehensive 3D data acquisition of underwater objects.

Datasets

The paper does not explicitly mention or introduce any specific datasets for training. Instead, it focuses on experimental validation using:

  • Calibration Targets: A checkerboard pattern is used for camera calibration in both air and underwater conditions.
  • Standard Sphere: A precisely measured standard ball (radius 20.01 mm) is used for quantitative error analysis and to evaluate the accuracy of the reconstruction system, both with and without the refraction compensation algorithm.
  • Real Underwater Object: A conch shell is used as a test object for demonstrating the 3D reconstruction capabilities in a natural underwater scenario.

Key Insights and Ideas

  • Refraction is a Major Hurdle: The paper strongly emphasizes that refraction is a critical factor affecting underwater optical 3D reconstruction accuracy, often overlooked or inadequately addressed in previous work.
  • Double Refraction Compensation is Crucial: The system's design accounts for the two instances of refraction (air-to-glass and glass-to-water) encountered by light, leading to a more robust compensation mechanism.
  • Rotary Scanning Enhances Coverage: The integration of a rotating stage significantly expands the scanning range, allowing for more complete 3D models of underwater objects.
  • Low-Cost Potential: The system's design, using readily available components like a green line laser and a standard underwater camera, positions it as a cost-effective alternative to expensive professional systems.
  • Harris Algorithm Suitability: The paper finds the Harris corner detection algorithm to be robust and suitable for feature extraction in their underwater scanning context, even with noise.

Key Results

  • High Accuracy in Air: The system achieves an average measurement error of less than 0.16 mm for a standard sphere in air.
  • Significant Underwater Improvement: After applying the refraction correction algorithm, the average measurement error for the standard sphere in water is reduced to less than 0.6 mm.
  • Error Range: The reconstruction error is reported to be less than 0.6 mm within a working distance of 140 mm to 2500 mm.
  • Error vs. Distance: The study notes that reconstruction error generally increases with measurement distance.
  • Successful Underwater Reconstruction: The system successfully reconstructed a conch shell in an underwater environment, demonstrating its practical applicability.

Limitations and Future Work

  • Working Distance and Error: The paper acknowledges that reconstruction error increases with distance. While the system performs well within its specified range, further improvements for very long-range scanning might be needed.
  • Environmental Factors: The current refraction compensation is based on specific refractive indices for water and air. Variations in water turbidity or salinity could potentially affect accuracy, though this is not explicitly detailed as a limitation.
  • Future Work: The authors suggest that their low-cost, simple-structured line laser scanning system can serve as a valuable reference for future underwater 3D laser scanning system development, potentially replacing more expensive professional sensors. Further research could focus on optimizing the system for even higher precision, faster scanning speeds, and robustness in more challenging underwater conditions.

Summary updated

BibTeX
@misc{xue2021underwater,
  author = {Qingsheng Xue and Qian Sun and Fupeng Wang and Haoxuan Bai and Bai Yang and Qian Li},
  title = {{Underwater High-Precision 3D Reconstruction System Based on Rotating Scanning}},
  year = {2021},
  doi = {10.3390/s21041402},
}
Figure