Multi-Camera-Based Universal Measurement Method for 6-DOF of Rigid Bodies in World Coordinate System

Zuoxi Zhao, Yuchang Zhu, Yuanhong Li, Zhi Qiu, Yangfan Luo, Chaoshi Xie, Zhuangzhuang Zhang

TL;DR

This paper introduces a novel multi-camera system for accurately measuring the 6-DOF (six degrees of freedom) of rigid bodies in a world coordinate system. By leveraging Zhang Zhengyou's calibration method and a least-squares approach, the system offers a versatile, non-contact solution that is less reliant on expensive, specialized equipment and applicable to a wide range of object shapes.

What Problem Does This Paper Address?

Precisely measuring the 6-DOF (three translational and three rotational) of rigid bodies is crucial for numerous industrial applications, from precision machining to spacecraft docking. Existing methods often suffer from limitations: some require expensive, specialized instruments (like laser trackers), others are sensitive to environmental factors, and some are restricted to specific object shapes or require complex setups. Vision-based methods, while promising, often struggle with accurate depth measurement in monocular setups or provide pose relative to the camera, which is lost if the camera moves. This paper aims to bridge this gap by proposing a more universal, accurate, and accessible multi-camera vision-based system.

Methods

The core of the proposed method lies in a multi-camera calibration process and a robust 6-DOF calculation model.

  1. Multi-Camera Calibration: The system utilizes at least two ordinary cameras. It begins with Zhang Zhengyou's calibration method to determine the intrinsic (focal length, principal point) and extrinsic (rotation and translation) parameters of each camera. Crucially, it also calibrates the pose relationship between the world coordinate system and one of the camera coordinate systems. This is achieved using a checkerboard pattern, allowing for the determination of the transformation between the camera coordinate system and the world coordinate system.
  2. 6-DOF Calculation Model: The method employs a least-squares approach to calculate the 6-DOF of a rigid body. This involves placing at least four non-coplanar control points on the rigid body. The 3D coordinates of these control points in the rigid body's local coordinate system (B system) are measured manually. The system then uses the calibrated camera parameters and the pixel coordinates of these control points in the images to first determine their 3D coordinates in the reference camera's coordinate system (C1 system) and subsequently in the world coordinate system (W system). Finally, these world coordinates and the known B-system coordinates are used in a least-squares formulation to solve for the rotation matrix and translation vector that define the rigid body's pose (6-DOF) relative to the world coordinate system. The rotation matrix is then converted into Euler angles for the rotational DOF.

Datasets

Not discussed in the paper. The paper focuses on experimental validation using a specific setup and a measured object.

Key Insights and Ideas

  • Universality through Multi-Camera: By using multiple cameras and a robust calibration procedure, the system overcomes the depth ambiguity issues of monocular vision and provides a more stable and accurate measurement.
  • World Coordinate System Focus: A key innovation is calibrating the relationship between the camera system and the world coordinate system directly. This means the measured 6-DOF is independent of camera movement, making the measurement reproducible and not tied to a specific camera viewpoint.
  • Least-Squares for Robustness: Employing a least-squares method for calculating the 6-DOF from multiple control points enhances the system's robustness against noise and minor errors in pixel coordinate extraction.
  • Accessibility: The method emphasizes the use of "ordinary cameras," reducing the need for expensive, specialized equipment, which broadens its applicability.

Key Results

The proposed method was evaluated using the Phantom 3D high-speed photogrammetry system (P3HPS) as a ground truth.

  • Rotational DOF Accuracy: The average error for rotational degrees of freedom was less than 1.1 degrees.
  • Movement DOF Accuracy: The average error for translational degrees of freedom was less than 0.007 meters.
  • Calibration Reliability: Intrinsic camera parameters and pose relationships were estimated with reasonable accuracy, with standard deviations indicating good reliability.
  • Distance Sensitivity: The study found that accuracy generally improves as the measurement distance decreases, with the 2.50 m distance yielding the best results. Errors in the z-axis direction were consistently larger than in the x and y axes, a phenomenon attributed to factors like depth of field limitations and error amplification in calculations.

Limitations and Future Work

  • Moving Range Limitation: The paper acknowledges that for objects with very large moving ranges, the control points might exceed the cameras' field of view, making measurement impossible.
  • Dynamic Measurement: The current work primarily focuses on static measurements. The authors suggest that addressing dynamic measurement problems is a key area for future research.
  • Lens Distortion: The current calibration model does not explicitly account for lens distortion, which the authors note can contribute to relative errors. Future work could incorporate distortion correction.
  • Optimal Measurement Distance: While the study explored different distances, further experiments are suggested to find the optimal measurement distance for maximizing accuracy.

Summary updated

BibTeX
@inproceedings{zhao2020multi,
  author = {Zuoxi Zhao and Yuchang Zhu and Yuanhong Li and Zhi Qiu and Yangfan Luo and Chaoshi Xie and Zhuangzhuang Zhang},
  title = {{Multi-Camera-Based Universal Measurement Method for 6-DOF of Rigid Bodies in World Coordinate System}},
  booktitle = {Sensors},
  year = {2020},
  url = {https://openreview.net/forum?id=J6SI3p6S1z},
}
Figure