Enhanced Affordable Computed Tomography Solutions

High cost and complexity of commercial X-ray computed tomography systems act as a barrier to the adoption of such technology by small and medium sized companies. The work presented here attempts to address these issues through the development of affordable solutions for both CT reconstruction and X-ray projection data set acquisition. Introduction The Precision Measurements Group at SIMTech has been providing X-ray inspection services to local industry for over 10 years mainly for hidden defect and internal structure visualisation of components and materials from the electronics, semiconductor and precision engineering sectors. One recurring comment from our interaction with local industry has been the relatively high entry cost for X-ray and computed tomography based inspection. We approached this concern from both a hardware and software perspective. In the first part of this paper we discuss the initial development, and subsequent further refinement of a software package to provide simplified access to 3D tomographic reconstruction of 2D X-ray projection images. The package incorporates projection image pre-processing, user-defined area-ofinterest selection, fan-beam and cone-beam reconstruction with variable reconstruction filtering and SIMTech’s patented central ray determination techniques. Performance enhancement using customised implementations of the back-projection process on both host and GPU devices is evaluated. Input data can come from any suitable X-ray inspection machine by providing a definition of that system through various configuration options. Output data is made available in a range of standard formats or can be examined directly by an accompanying visualisation software tool. In the second part of this paper we discuss the upgrading of a low cost manually controlled 2D X-ray inspection machine by adding a rotation stage suitable for the acquisition of projection image data sets. A computer-controlled rotation stage was mounted on a removable platform within the inspection machine. Scanning and alignment software was developed to operate the system whilst following a series of defined procedures. Methods for correction of distortion introduced by the image intensifier have been developed which do not require prior knowledge of the intensifier geometry. The upgraded system can provide input data to the reconstruction and visualisation package described in the first part of the paper. 2nd International Symposium on NDT in Aerospace 2010 We.1.A.1 Licence: http://creativecommons.org/licenses/by-nd/3.0 1 1. CT Reconstruction Software Package Development The objective in our development of a CT reconstruction software package was the creation of a hardware independent solution for reconstruction and visualisation that was useable on general purpose PC hardware. The software was required to process projection images and associated calibration and control information to yield a series of reconstructed slice images perpendicular to the axis of rotation during projection image acquisition that represent the 3D model of the object under examination. Simple visualisation of the reconstructed slice images was required to be integrated with, or closely coupled to, the processing required to generate them. The visualisation allows the user to view sections through the reconstructed volume along the axial, frontal or sagittal axes. The user was also to be provided with sufficient control over the process to allow the software to operate with projection images generated by any suitable X-ray inspection machine. 1.1 Theoretical Background As illustrated in Figure 1, computed tomography is an X-ray imaging technique that makes an image out of thin cross-sections of a part. Only a thin X-ray plane moving through a part is detected. X-ray data are collected through all points of the X-ray plane, and from many angles. This process produces many thin images ‘slices’ which are combined by mathematical operations. In performing computed tomography, the system must scan around an entire part to collect all the data needed to create a CT image. In most X-ray inspection systems for industry applications, the scan is performed by simply rotating the part on a part manipulator turntable. The most straightforward form of tomographic reconstruction is based on the fanbeam algorithm [1]. Suppose a parallel beam projection on the j detector cell at an angle θ is denoted as ) , ( θ j p and its Fourier Transform is ) , ( θ w P , the reconstructed image is then obtained as ∫ ∫ +∞ ∞ − + = dw e w w P d y x f y x w j ) sin cos ( 2 0 ) , ( ) , ( θ θ π π θ θ A detailed description of the computer implementation of this equation can be found in [2]. When used in conjunction with an area detector the fan-beam algorithm operates on single lines of pixels extracted from the image acquired at each projection angle. Each slice is individually reconstructed and the effect of the axial position (relative to the X-ray source) is not generally considered. In order to achieve an improved and more consistent quality of reconstruction over the field-of-view of an area detector it is necessary to make use of the cone-beam or FDK algorithm [3]. The simplest way to obtain the cone-beam reconstruction algorithm is to start from the fan-beam reconstruction algorithm and calculate the contribution of each ray to an object point based on the geometry as shown in figure 2. The reconstructed image is obtained from ds v s D D s s h v s q d y D D y x f m