From CAD to 
Reverse Engineering

A Technical Note on 3D Digitizing and Surfacing
for MicorScribe an MicroScan Users

This technical note will explain the differences and backgrounds of shape reconstruction. We will compare two popular approaches - the measurement of characteristic points and curves verses the capturing of dense point clouds. Immersion’s MicroScribe family of desktop digitizers has been available on the market for several years - over 7,000 MicroScribe's have been shipped to users around the world. 3D measurement and object recon-struction using a touch probe digitizer has become popular and affordable to almost everyone working in 3D applications.

There are three steps which lead to a 3D computer model from an object under reconstruction:

1. The preparation step:

Set up a digitizing strategy based on the characteristics of the object (contours, freeform surfaces, geometric primitives such as lines, curves, corners, planes, cubical, spherical, or conical surfaces, etc.), subdivide the object surface into a number of individual patches, if required mark a grid of helper curves for improved treatment of freeform shapes

2. The digitizing step:

Open a 3D software application or/and the MicroScribe utility software (MUS) to capture points along the contours and grids relative to the patches defined before, adapt the point distance to the surface curvature to allow for high precision surfacing in the final step

3. The modelling step:

Use the 3D software’s modelling tools to convert the digitized data to CAD surfaces, work on all patches and combine them to create a closed and watertight surface or solid model, export model in a suitable format for downstream applications, if MUS is used stand-alone (e.g. when a 3D software is not supported) first save all digitized points in ASCII or text file format and then import it into the target application for final modelling.

The above method is very similar or even equivalent to traditional CAD in that it reflects the principle of object design based on some sample points and lines which, according to the underlying design intent, have to be entered by the user. In the context of this paper we therefore call it CAD reconstruction. Clearly this method is very easy to learn and follow by 3D designers and CAD engineers. Also, since respective software applications deliver mathematical surfaces in NURBS (Non-Uniform Rational B-Splines) representation the results are compatible to almost every downstream application (shape modification or remodelling, reverse engineering, rapid prototyping, visualization, physical analysis, etc.).

However, same as for every other technology as well CAD reconstruction has its limits. If e.g. an object is too complex (natural or biological objects, sculptures or artwork pieces, freeform-ruled or arbitrary shapes, a.o.) it would require too much effort or even make it impossible to reconstruct it as described. The same holds for objects having soft or sensitive surfaces which can not be touched by the MicroScribe stylus tip. In these cases RSI’s new MicroScan laser sensor will considerably enlarge the MicroScribe’s capabilities. MicroScan is a compact miniature sensor head which easily attaches to the MicroScribe stylus, ready for operation in a minute without the need for other external devices. MicroScan adds dual field laser scanning to the MicroScribe (two different working space sizes at different distances to the object surface).

3D scanning with MicroScan is of course much more complex in comparison to capturing isolated points but the user is not concerned with that process. Two basic steps are required:

1. The scanning step:

Similar to spray painting, sweep the sensor’s laser line across the object (surface profiling) until the surface portion accessible from the MicroScribe’s current position is captured (watch model growing the computer screen, to increase point density in high curvature regions scan in different directions) - if all surfaces of interest can not completely be scanned reposition the object start another scan (repeat this if needed until shape is closed)

.

2. The processing step:

Use the MicroScan software tools to convert the measured point cloud(s) to a quality polygon model), involves clean-up, triangulate (create the initial wireframe mesh), fill holes, smooth and decimate, in case of multiple scans register and merge them, and export model in polygon format (.stl, .dxf)

Remark: Postprocessing the raw point data coming directly from digitizing is required with every scanner because there are always missing or outlying points and noise or distortion superimposed to measurement. The essential is to have a powerful and comfortable software at hand to streamline processing.

As opposed to sampling points with a standard MicroScribe to directly create a CAD model (NURBS surface), MicroScan models are polygon models (wireframe meshes). This is true for many other optical or laser based scanners as well. Polygons are sufficient for many applications including rendering and visualisation, rapid prototyping (stereo lithography, nc milling), physical analysis (FEM in stability, thermo and fluid dynamics).

Popular formats for polygon data are AutoCAD DXF and STL. In terms of quality polygon models are restricted by the scanner’s accuracy and resolution only and not by user-specific factors like measurement strategy or personal care. The scanner does not know about the object topology but captures the entire surface with every detail down to the resolution limit. Since points and polygons serve as the basic information in industrial reverse engineering, let us call MicroScan digitizing the reverse engineering method of object reconstruction.

In its full meaning reverse engineering also implies a process known as surface reconstruction. Although often named as surfaces because they can be rendered to appear as closed shapes, polygon models only consist of points interconnected by lines or curves in triangle or quadrangle structures. The areas inside the polygons do not exist in that representation so there are limited possibilities only to modify or manipulate polygons using CAD tools. The purpose of surface reconstruction is to convert polygon meshes to NURBS surfaces so they are fully compatible to objects created in CAD. NURBS are built from sets of mathematical functions which allow to access every incremental position within an object surface, opening the way to almost unlimited flexibility in shape modification and surface treatment. Among popular NURBS formats are IGES, STEP, and VDA. Some major CAD software packages already provide surface reconstruction capabilities. However, they typically transpose traditional forward engineering (CAD) to reverse engineering whereas today’s independent and dedicated surface reconstruction packages use considerably different approaches. These packages which are also available from RSI are by far more productive (faster, easier to learn and use, highly automated, workflow-oriented) than proprietary tools.

Since they behave the same as CAD models, NURBS objects can directly be used in subsequent CAD applications.