3-D Scanning Machines

(Last update June 11, 1997)

We have built several machines over the years and have plans of continuing to do so in the future. The machines can be divided into three categories based on their methodology.

The first group of machines we worked with were meant to be low cost and were based on TIME MODULATED structured illumination. These devices were built as part of our design courses as well as our research. The emphasis on these machines was in the signal processing since the time modulation allowed the usage of matched filtering to achieve extremely high SNR. Most of these devices are in disrepair or archived in our labs. So far they have only been documented in conference papers. Hopefully, we will find the time to document these machines further, produce journal papers describing their performance and maybe even resurrect them for further study.

The second group of machines are based on the CLASSICAL SINGLE LIGHT STRIPE techniques. For the most part, these devices are open art. However the software used to extract information from the raw data was heavily influenced by what we learned from the time-modulation techniques. Three of the most advanced features of these devices were their low cost, high accuracy and self calibration. We have documented one of these machines as a technical report and have plans to continue using the Hb1009 unit as an active component of our research. We have built four machines so far, two of which are shown below.

The third group of machines are based on successive light striping and the use of HIGH SPEED SLMs. We have only constructed temporary setups in our labs to produce experimental verification of these techniques. Our research has yielded some interesting insights into the capabilities of these systems and we are in the publication process with some of these insights. We expect that in the next couple years we will construct more general purpose systems of this type.

Figure 1: The Structured Light Illumination Technique (SLIT) machine represents our first machine based on the classical single light stripe technique. We constructed it with a $13,000 grant in about 14 weeks. It was one of those rare projects where everything worked right, fit well together, etc.. Its measurement accuracy for a flat surface was a standard deviation of 1.56 microns or 0.0000625 inches. The surface under measurement was a black foam "rubber" surface which was machined to be flat. One of the problems with measuring such a surface was that it had a mixture of textures. The basic surface texture was corrupted by large pockets or "crater" like shapes. The Central Limit Theorem was limited in this case since the mixture of measurements led to high standard deviations. We developed some vary sophisticated nonlinear filtering techniques to separate the two surface textures which allowed for the high accuracy. The machine was also made more robust by using 3 levels of measurement combined into an AI like state process.

Figure 2: The machine that we are presently using is referred to as the Hb1009. As a result of the SLIT machine we continued to develop the classical structured light devices. One of the problems with the SLIT machine was that whenever we changed a lens, a slit reticle or an axis length, we had to manually re-calibrate the system which could take several hours. For the Hb1009 we incorporated our pattern recognition techniques to perform self calibration/determination of axis lengths, light stripe projection angle, light stripe offset, light stripe intensity variation, working distance and magnification (at variable distances). Our reconstruction algorithm incorporates non-teleconcentric properties of lenses so we can use standard microscope lenses. For research applications this automatic calibration aspect has been very practical. The price is right too, $1000 for the camera, $1000 for the frame grabber, $600 for the 3-axis table (that wasn't easy), $400 for the electronic driver/interface hardware (we build our own in the PCB Lab), $225 for the laser projector and $1600 for the computer (486 or pentium). Although we could obtain the resolution of the SLIT if we replicated the SLIT set up, the existing machine set up yields: a 4" x 4" x 12" scan volume, step resolution of 1/4000 of an inch, and 0.0008 inch height resolution. For a spatially dense sample grid, the motor motion has less effect on scan time, so the scan speed is limited primarily by numerical processing. With an ISA bus frame grabber the scan rate is 110 point samples/sec while with a PCI bus frame grabber it would be 255 points/second. Both scan speeds were measured with a 66 Mhz 486. The machine hardware also has capability for color imaging but we have not included that in the software yet. We can see from our experiences with these devices that it will not be long before we will be able to buy a $1000 scanner at the local computer store. If you look at a 2-D scanner, you will notice that the only thing missing is the light stripe. The rest would be software. Maybe with the multimedia market developing, a 3-D rendering will be desired by enough consumers to justify a market large enough to support a $500 to $1000 3-D scanner unit. We will see.

Figure 3: In Figure 2 we got a little carried away with our photo studio software. This image is probably a better view of the Hb1009.

Figure 4: The key designers and developers, Laurence G. Hassebrook (left) and Raymond C. Daley (right) of the Hb1009 (shown), Structured Light Inspection Process Prototype 1 (SLIPP1), SLIPP2, Structured Light Inspection Technique (SLIT) and Structured Light Amplitude Modulation Process (SLAMP) systems.