TL;DR
This paper introduces a novel self-calibration method for light-pen-based 3D vision coordinate measurement systems. By combining spherical fitting, generalized least squares, and the principle of position invariance, the method accurately determines the light pen's stylus tip center, crucial for precise measurements, and establishes a defined "pose domain" for optimal system performance.
What Problem Does This Paper Address?
Modern manufacturing, particularly in industries like automotive and aerospace, demands increasingly high-precision, real-time measurement capabilities. Traditional Coordinate Measuring Machines (CMMs) are often too large and have limited measurement ranges for on-site applications. While other technologies like laser trackers and optical scanners exist, they can be expensive, complex, and lack portability. Vision measurement systems, especially light-pen-type systems, offer a promising alternative due to their small size, flexibility, and convenience. However, the accuracy of these systems is significantly influenced by the light pen's stylus tip center and its orientation (pose). Accurately determining the stylus tip's true position and defining a suitable "pose domain" (the range of orientations for reliable measurement) is critical for achieving the required precision. This paper addresses the challenge of accurately calibrating the stylus tip center and defining this pose domain without relying on external, auxiliary equipment.
Methods
The core of this paper is a self-calibration method that leverages several key techniques:
- Spherical Fitting: The principle of position invariance is used. When a feature point on the measured object is viewed from different poses of the light pen, its projection onto the image plane changes. However, the actual 3D position of the stylus tip center remains constant. By capturing multiple images of the same measurement point with varying light pen poses, the system can fit these observed points to a sphere. The center of this fitted sphere represents the actual 3D position of the stylus tip center.
- Generalized Inverse Method of Least Squares: This mathematical technique is employed to solve the system of equations derived from the spherical fitting process, efficiently determining the sphere's center and radius.
- Position Invariance: This fundamental principle states that the physical position of the stylus tip center does not change regardless of the light pen's orientation. This invariance is the bedrock upon which the spherical fitting approach is built.
- Pose Domain Determination: The experiment involves systematically rotating the light pen around its three axes (pitch, yaw, roll) within specific angular ranges. For each degree of rotation, images are captured, and measurement errors are analyzed. This allows the researchers to identify the angular ranges where the system maintains acceptable accuracy, thus defining the "pose domain."
The process begins with camera calibration, followed by capturing multiple images of a measurement point with the light pen in different, known poses. The conversion matrix between image and world coordinates is established. Then, the stylus tip center's position is iteratively refined using the described methods.
Datasets
Not discussed in the paper. The experiments were conducted using a custom-built measurement system and light pens, with data generated from controlled experiments rather than pre-existing datasets.
Key Insights and Ideas
- Self-Calibration is Key: The ability to perform accurate stylus tip calibration in situ without external equipment is a significant advantage, making the system more practical for industrial environments.
- Spherical Fitting for 3D Position: Using the principle of position invariance to fit captured points to a sphere is an elegant way to overcome the challenges of varying poses and determine the true 3D location of the stylus tip.
- Pose Domain Matters: The paper highlights that the light pen's orientation is not just a minor detail but a critical factor affecting measurement accuracy. Defining a specific "pose domain" ensures that measurements are taken within reliable operational parameters.
- Accuracy Meets Industrial Needs: The experimental results demonstrate that the proposed method achieves measurement accuracy suitable for basic industrial requirements, validating its practical applicability.
Key Results
The paper reports that the self-calibration method stabilizes the absolute error of the probe's measured value below 0.0737 mm and the relative error below 0.0025 mm. Furthermore, single-point repeatability experiments showed that the probe repeatability of each light pen did not exceed 0.041. These results indicate that the method significantly improves measurement accuracy and satisfies basic industrial measurement requirements. The experiments also defined specific angular ranges for pitch (41°~159°), yaw (0°~180°), and roll (0°~45° and 136°~180°) where clear and recognizable images were obtained, contributing to the understanding of the light pen's pose domain.
Limitations and Future Work
- Limitations: The paper acknowledges that the processing error of the mechanical device used for rotating the light pen and the measurement error of the equipment itself can affect experimental results. While they mitigated this by using 1° increments, it's an inherent limitation of experimental setups. The accuracy is also dependent on the initial camera calibration.
- Future Work: The authors explicitly state that the next step of research will involve applying this method to actual industrial applications. This suggests a focus on real-world testing, integration into existing industrial workflows, and potentially further refinement for more demanding precision requirements.
Summary updated