Fully Automatic Optical Surface Determination in Computed Tomography

摘要

Computed tomography (CT) has become a widely used method for different industrial applications where inner information about the object is needed, e.g. material and defect analysis or dimensional measurement. However, a priori knowledge about the outer object surface can be useful for the automation of the acquisition and reconstruction of the three-dimensional volume or further volume analysis steps. In the article two approaches will be presented to determine the outer surface of arbitrary objects fully automatically and fast using an optical camera. The methods are well suited for objects in the range from the sub millimeter space up to several centimeters. They are integrated into a typical industrial CT-system with an X-ray tube and a detector and a rotatable object table. Only the object shape or the outer object hull is of interest. Hence, backlight is used to determine the contour robustly out of the resulting shadow images. Furthermore, homogenous backlight avoids specular highlights of the object or ambiguities due to the material transparency of the object. The object contour can be directly extracted out of the shadow images. For each height inside of an image the shadow radius is determined. The maximum over all radii of each height defines the maximal size of the object shadow in relation to the rotation axis for all regarded rotation angles. Since the camera has been previously calibrated, the maximal object size can be determined based on the found maximal radius for each image height. This approach can be further extended. It is possible to determine not only the maximal radius per height but each radius per angle and height. This results in cylindrical coordinates (angle, height, and radius) as surface representation of the object. The article will be concluded by evaluating the accuracy and performance of both approaches. It will be shown that the proposed methods are even more accurate than using a standard filtered back projection approach. The surface representation can be used for various steps of a CT working pipeline, e.g. during object navigation inside the cabin of the CT-device or as prior information for a volume reconstruction.