 |
Reverse Engineering: |
||||||
|
Steps in 3D reconstruction of |
Image3 LLC |
|
Many designers integrate 3D into their design process because they are looking for time-to-market improvements for better links to manufacturing. Unfortunately, 3D data is not always available for many commercially available products. Reverse engineering is the term used to describe the process of recapturing lost or damaged 3D design data for these products. Most reverse-engineering approaches involve imaging or digitizing an object and then creating a computerized reconstruction that can be integrated in 3D into the particular design environment. Relying on volume visualization technology, a fundamental technique for interpreting and interacting with large 3D data sets, it is possible to establish reverse engineering as an integrated step in rapid production and agile manufacturing. Image3's innovative approach uses Velocity2TM and computed tomography (CT) data sets to reconstruct objects. Velocity2, a modular software application for 3D reconstruction, rendering, and rapid prototyping, is Image 3 LLC's showcase product. This package supports file output from volumetric image data sets as well as 2D-contour data sets that describe 3D objects. This proprietary software package, developed and owned by Image3, was specifically designed for medical and industrial markets with emphasis on rapid prototyping. In CT imaging, a 3D image of an x-ray-absorbing object is reconstructed from a series of 2D cross-sectional images. An x-ray beam penetrates the object, and transmitted beam intensity is measured by an array of detectors. Each such "projection" is obtained at a slightly different angle as the scanner rotates about the object. The 2D image is computed from the projected images using the approximate method of "back projection" or the more accurate method of inverse Fourier transformation. CT was introduced in the early 1970s as a neurological examination technique, and later extended to industrial applications. It is a radiographic examination technique used whenever the primary goal is to locate and size planar and volumetric detail in three dimensions. Current industrial CT systems can provide dimensional measurements at an accuracy competitive with coordinate measuring machines (CMMs). Of the existing methods for generating a CAD model of a physical part, only CT can nondestructively dimension internal as well as external surfaces. CT has the unique ability to detect and quantify defects. Additionally, it is indifferent to surface finish, composition, and material, and it can measure part coordinates as fast as a laser scanner -- and orders of magnitude faster than CMMs. |
|
| Because of the relatively good penetrability of x-rays, as well as the sensitivity of absorption cross sections to the density and atomic number of matter, CT permits the nondestructive evaluation (NDE) and, to a limited extent, chemical characterization of the internal structure of materials. Also, since the method is x-ray based, it applies equally well to metallic and nonmetallic specimens, solid and fibrous materials, and smooth and irregularly surfaced objects. |
Graphic created from CT-based 3D |
|
3D reconstruction of a volumetric data set is accomplished by extracting a region of interest (ROI), closing the boundary or defining the edge(s), and reconstructing surface from ROI to ROI throughout the image set. The resulting 3D reconstruction consists of a highly accurate 3D surface comprised of triangles. Output to rapid prototyping devices is accomplished directly via a Velocity2 module. When working with scan data, it is fairly common to produce models that have large numbers of surface polygons. Very large files can be difficult to export to rapid prototyping systems. When this occurs, Velocity2's polygon reduction program, PolyMerge, can be used to selectively reduce the numbers of surface polygons by collecting small triangles into larger ones in regions of the surface that are relatively flat. With PolyMerge, you specify this "surface flatness" as the deviation in the local surface normal vector, the "delta value," in units of angular degrees. For example, a perfectly flat surface, i.e., one with a delta value of zero, will have no variation of the surface normal vectors from one triangle to the next; whereas, in regions of high surface curvature the delta value will be large. Typically, delta values of 20-30 degrees provide reductions in numbers of triangles of 30 percent or more in flat areas of the model without significantly affecting surface detail. In many cases the reconstruction may have surface irregularities simply due to noise in the original image set. In these cases, it is advantageous to smooth the surface prior to polygon reduction to remove local surface roughness. The smoothing algorithm used in Velocity2's PolyMerge (and in Display as well) recalculates the locations of triangle vertices as the average of a given vertex and its immediate neighbors. Significant file-size reduction can be achieved, which greatly improves the ability to export reconstruction files in the rapid prototyping STL file format to RP systems. With good technique and data, CT scan accuracy generally falls within ±20 percent of the slice data. For a 1-mm slice this would equal ±0.2 mm. Slice or scan spacing is critical for 3D model reconstructions, and should not be confused with slice thickness. Anything over 3 mm is not acceptable for most complex structures. The accuracy in the Z axis is determined by the spacing. ARACOR's CTM 500 industrial scan system reports an accuracy of ±0.001 in. with a resolution of ±0.007 in. and a tracking speed of 100-300 slices per hour. The ARACOR-built ICT-1500 CT system at Hill AFB, Utah, employs a 9-MeV linear accelerator and achieves a maximum resolution of 1 mm and a minimum scan time of 1 minute per slice. For more information, contact Alair Griffin, CEO of Image 3 LLC, Salt Lake City, UT, and the author of this article;
(801) 466-9176; fax: (801) 466-5817; E-mail: javelin@lonepeak.com.
The author acknowledges the following sources: Yancey, R. et al., "Integration
of Reverse Engineering, Solidification Modeling, and Rapid Prototyping
Technologies for the Production of Net Shape Investment Cast Tooling,"
(1996), Advanced Research and Applications Corporation (ARACOR), Dayton,
OH; Haystead, J., "Computed-Tomography-Based Medical Imaging,"
Vision Systems, Vol. 2, No. 7 (July 1997); Cyberform International,
Inc., Richardson, TX, marketing literature (1996).
|