Calibration of Robotic Manipulator Systems for Cone-Beam Tomography Imaging

Abstract

Image reconstruction of tomographic data relies on the precise knowledge of the geometric properties of the scan system. Common tomography systems, such as rotational tomography, C-arm systems, helical scanners, or tomosynthesis scanners, generally use motions described by a few rotational or linear motion axes. We are interested in applications in nondestructive testing, where objects might have large aspect ratios and complex shapes. For these problems, more complex scan trajectories are required, which can be achieved with robotic manipulator systems that have several linear or rotational degrees of freedom. For the geometric calibration of our system, instead of using an approach that scans a calibrated phantom with markers at a known relative position, we propose an approach that uses one (or several) markers with unknown relative positions. The fiducial marker is then moved by a known amount along one degree of freedom, thus tracing out a “virtual” phantom. Using the assumed spacial locations of the markers together with the locations of the markers on the imaging plane, we use a nonlinear optimization method to estimate the orientation of the linear and rotational manipulator axes, the detector and source location, and the detector orientation.