Hardner Matthias, Liebold Frank, Wagner Franz, Maas Hans-Gerd
Institute of Photogrammetry and Remote Sensing, TUD Dresden University of Technology, 01069 Dresden, Germany.
Sensors (Basel). 2024 Aug 8;24(16):5139. doi: 10.3390/s24165139.
Micro-Computed Tomography (µCT) systems are used for examining the internal structures of various objects, such as material samples, manufactured parts, and natural objects. Resolving fine details or performing accurate geometric measurements in the voxel data critically depends on the precise calibration of the µCT systems geometry. This paper presents a calibration method for µCT systems using projections of a calibration phantom, where the coordinates of the phantom are initially unknown. The approach involves detecting and tracking steel ball bearings and adjusting the unknown system geometry parameters using non-linear least squares optimization. Multiple geometric models are tested to verify their suitability for a self-calibration approach. The implementation is tested using a calibration phantom captured at different magnifications. The results demonstrate the system's capability to determine the geometry model parameters with a remaining error on the detector between 0.27 px and 0.18 px. Systematic errors that remain after calibration, as well as changing parameters due to system instabilities, are investigated. The source code of this work is published to enable further research.
微计算机断层扫描（µCT）系统用于检查各种物体的内部结构，如材料样本、制造零件和自然物体。在体素数据中解析精细细节或进行精确的几何测量关键取决于µCT系统几何结构的精确校准。本文提出了一种使用校准体模投影的µCT系统校准方法，其中体模的坐标最初是未知的。该方法包括检测和跟踪钢珠轴承，并使用非线性最小二乘法优化来调整未知的系统几何参数。测试了多个几何模型以验证它们对自校准方法的适用性。使用在不同放大倍数下捕获的校准体模对该实现进行了测试。结果表明，该系统能够确定几何模型参数，探测器上的剩余误差在0.27像素至0.18像素之间。研究了校准后残留的系统误差以及由于系统不稳定而变化的参数。这项工作的源代码已发布，以便进行进一步研究。
