Issue: Volume: 24 Issue: 2 (Feb 2001)

No Bones About It

By Karen Moltenbrey

The Smithsonian National Museum of Natural History (NMNH) is trying to prevent history from repeating itself. More than 65 million years ago, the three-horned, plant-eating dinosaur known as triceratops became extinct; today its fossilized re mains are in danger of suffering a similar fate. After nearly 100 years of displaying the world's first triceratops skeleton, the museum was forced to disassemble the specimen and replace it with a cast replica to protect the fragile bones from further atmospheric wear and tear. While embarking on this major restoration project, the NMNH scientists also rewrote history by using a variety of digital technologies, including 3D scanning and stereolithography, to correct scientific inaccuracies in the original 1905 mounted skeleton.

For instance, the skeleton is a composite from at least 15 mostly similar-size triceratops dinosaurs. Using bones from different individuals is standard practice for even the most recently mounted specimens because finding a complete dinosaur skeleton is virtually impossible. Therefore, the museum's original preparators were forced to improvise when constructing the skeleton. Since no triceratops specimens had been found with feet, they substituted foot bones from a similar-size animal of a different species. The group also used a smaller skull as well as a mismatched left humerus (shoulder bone), which was smaller than the right, as were the left ilium (pelvis) and left scapula (shoulder blade) compared to their right-side counterparts.
By digitizing its triceratops skeleton with a high-resolution 3D scanner, the Smithsonian was able not only to build a precise physical replica, but also to create an accurate animation of the dinosaur's movements to better understand how it behaved.

Prior to the skeleton's disassembly in mid-1999, the museum contracted Scansite, a scanning service provider, to oversee the scanning of the fossils. This procedure would provide the scientists with a digital archive of the specimen before they replaced it with a replica made from casts of the original. Seizing a rare opportunity, the team also scanned most of the fossils individually once they were taken down from the mount.
The Smithsonian is using digital technology to save its triceratops skeleton-originally constructed in the early 1900s-from further deterioration.

According to Lisa Federici, cofounder of Scansite, the most ambitious step in the project, aside from the disassembly, was acquiring the 3D data. Once that was done, the door opened to a host of other digital applications, which were provided from various vendors free of charge or for a minimal stipend. "Scanning is the missing link that bridges the real world and the computer," she says. "And it's a perfect fit for dinosaur fossils, which have organic shapes that are difficult to measure by any other means without damaging the specimen."
After the overall skeleton was scanned for reconstruction purposes, each bone was then scanned at high resolution.

Toy maker Hasbro was the first to offer its services, creating one-sixth scale models of the original fossils from the scan data, using an SLA7000 stereolithography machine from 3D Systems. The plastic-like models enabled the museum's paleobiologists to easily manipulate the miniature bone joints for testing theories about the dinosaur's range of motion, and helped the group determine a more accurate and natural position for the new mounted skeleton. Other companies in the rapid-prototyping industry helped create correctly sized replacements for the skull and certain bones. As a result of these efforts, this May the museum will unveil a new triceratops-made from light, durable casts-that is more anatomically correct in structure and posture than the original skeleton.

Digital scanning is hardly a new concept for the museum or the world of paleontology. "Digitizing a bone is fairly simple now, if the accuracy of the scan is not an issue," says Ralph Chapman, a paleobiologist and head of NMNH's Applied Morphometrics Laboratory. "But for us, the data had to be scientifically accurate so we could produce exact replicas."

Scansite and its subcontractors used a Steinbichler Optotechnik OptoTrak scanning system to collect the overall dinosaur data, digitizing key points on the original skeleton before it was disassembled. By positioning approximately 100 reflective points on the skeleton as "landmarks," the group captured the dinosaur's posture in the old exhibit, so replicas of the 250-plus bones and bone fragments could be reassembled in the original position.

As the bones were removed from the mount, the group used high-resolution scans to collect data from more than 100 selected bones that were in good condition. For most of the smaller bones, the team used a Cyberware laser scanner, which creates point data from an object placed on a platter-like device. The larger bones that were too heavy or unwieldy for the Cyberware platform were scanned with a Steinbichler Comet, a portable optical scanning device that operates on a tripod.
Because scans of the dinosaur's huge skull were partially obstructed by the wood and steel supports on which it was mounted, extensive data editing was required to fill gaps in the point cloud data before a fully digital model could be obtained.

Digitizing the dinosaur's 6.5-foot skull, which contained a back frill and a complicated internal cavity, presented yet another challenge. Because of its fragility and weight (approximately 500 to 600 pounds with the wood and steel armature supports attached), the group had to scan the skull while it was still mounted, causing line-of-sight issues for the Comet.

Following three weeks of primary scanning, the group netted nearly 200 files of point cloud data totaling 20gb, which were given to Art Andersen, president of Virtual Surfaces, a company that works with point cloud data, for editing into a virtual dinosaur. "With the skull in particular, there were gaps in the data that I had to fill in by mirroring other pieces to create a solid model," he says. After struggling with the incomplete skull data, Andersen received help from Nvision, which used its portable ModelMaker laser scanner that attaches to a FARO mechanical arm, for capturing details in the hard-to-reach areas inside the back frill.
Digital scan data was used to mill a replica of the humerus bone, cut from a block of foam-like material called Renshape.

Andersen also deleted the scanned points associated with the steel mount supports that are permanently attached to the fossils, so the virtual bones more closely resemble the original state of the fossils.

To make the skull data usable for animation and prototyping, Andersen reduced the 30 million data points from the scanning to about 1.5 million points using SDRC's Surfacer software (formerly from Imageware) running on a DEC Alpha workstation. "This process of subsampling the point cloud data was a balancing act. [Andersen] had to maintain enough points to retain extremely accurate data for the specific applications but not so many that it would cripple the computers," adds NMNH's Chapman.
Defining the cutting paths to output each bone prototype (including the scapula above) required scripting an average of 40 milling machine programs.

With Surfacer, Andersen polygonalized the subsampled point data for each bone, then created a new, full-size version of the problematic bones-the left humerus, ilium, and scapula-by mirror-imaging the complementary bone from the opposite side of the specimen. Next, using the registration points from the scan data, he reconstructed the original skeleton posture in virtual space.

Using the original fossils-with the exception of the three incorrectly sized bones and the skull-the museum preparators made molds from which the cast replica bones for the new exhibit were formed. To create the prototypes for the right mates of the humerus, ilium, and scapula, Kelly Hand, president of Satellite Models, used Solid Concept's SolidView rapid-prototyping software to view the STL files provided by Andersen, and to create the mirror image of the complementary bones. He then exported the files into Vero International Software's Visi-Series CAD/CAM software, where an average of 40 programs per bone were scripted for generating the numerous tool paths used by the CNC five-axis milling machine.

Hand likens the milling process to that of initially excavating the fossils in that both procedures start out with a large block of material, and excess material is removed with progressively smaller tools until the final "bone" is revealed.
Museum paleobiologists plan to add muscle to the 3D triceratops model to see how the animated walk cycle is affected.

The 3-foot-long humerus and the scapula were fairly easy to cut from a block of Renshape (a foam-like substance) placed within the machine's 10- by 5- by 3.5-foot build envelope. In contrast, the 5-foot-long ilium contained overhangs and undercuts that were difficult to create, requiring Hand to cut the protruding data from the main CAD image, mill it separately, then glue the pieces together for further overall milling.

For Ron Jones, president of Shared Replicators, the issue looming large for digitally re-creating the triceratops head was figuring out how to accurately output the 7-foot skull and lower jaw from a 3D Systems' SLA7000 stereolithography machine with a build envelope of less than 2 feet. (The skull data was increased by 15 percent so it would be accurately sized for the skeleton.) Although a five-axis milling machine, such as the one used to reproduce the three other bones, could have accommodated the larger piece, the SLA process was far more accurate for replicating the complicated skull geometry, notes Scansite's Federici.

"For us, it was the biggest project yet, and I believe it is the largest SLA mold pattern ever produced," Jones says. "We wanted to prove to ourselves that we could develop a technical solution for this project, because the techniques that resulted from the work can be applied to other projects."
During the data editing process, missing data had to be filled in along the jaw line to create a complete surface. When the head was output using stereolithography, that area was much smoother than the rest of the head, which had a fissured look, so more

The approach taken by the team at Shared Replicators was to segment the data into 35 horizontal and vertical pieces using SolidView before exporting them into 3D Systems' Lightyear software for programming and setup. The files were output separately and then seamlessly joined together using a photo-bonding process commonly employed in the auto and aerospace industries for constructing larger prototypes.

One problem the group faced was retaining the part's integrity. "When the scan work was originally done, there was no technical solution for reproducing it," says Jones. "[The scanning technicians and Andersen, who edited the data] knew the scan data was going to be used to make a miniature model, and the resulting polygonal file was fine for that one-sixth scale model, produced by Jason Dickman [Shared Replicators' vice president of laboratory operations] while he was at Hasbro. But when we blew it up to create the full-size skull, some of the polygons became too pronounced. Instead of a smooth contour, we had scalloped sections that required a good deal of handwork to smooth and fill."
Using the digital scan data, the Smithsonian produced these detailed 3D images of the dinosaur's thoracic vertebrae and ribs (left) and cervical vertebrae (below). From the scans, the museum has constructed a fully articulated digital triceratops that wil

Shared Replicators encountered similar problems in areas where Andersen filled in missing scan data to create a solid, complete surface. "A dinosaur skull is nothing but one wrinkle after another. You've got all these fissures for the blood vessels and nerves that create complex geometry," Jones says. "So we'd have a highly textured surface that suddenly mated into the mouth's gum line, which was very smooth because Andersen had to fill in that area with data he didn't have." Working with the NMNH scientists, the modelers tweaked some of these areas by adding more contour data to the SLA file. More complicated changes were addressed by the museum prior to casting.

The computerized scan data holds a wealth of information that researchers are only beginning to tap. For instance, the museum is using the 3D data to conduct primary re search into how the animal actually moved. "The real bones are too large and fragile to manipulate by rotating the joints in your hands," says NMNH's Chap man. "But we can do this with the miniature and CG models."

Virtual Surfaces' Andersen, recognizing the potential of the 3D data, met with Hans Larsson, a paleontologist and post-doctorate fellow at the University of Toronto's Department of Zoology. Together they used Autodesk's 3D Studio software to create an animation cycle of the dinosaur walking, based on the original posture.

This application was taken a step further by Rebecca Snyder, an intern at the NMNH, who imported each scan file into Discreet's 3D Studio Max running on a Dell PC, and built the virtual triceratops bone by bone to achieve the proper articulation of each joint. Using the 3D file, the paleontologists, with input from colleagues throughout the world, have tested various theories about the dinosaur's posture and movement. This information was then used to determine the new posture for the mount.
Using 3D Studio Max, the museum created a walking cycle of the triceratops based on accurate anatomy, not artistic models such as those used to produce walking dinosaurs in movies.

Animating the model, though, has been a slow process because of the large file size (without the animation parameters, the file is 300mb). "Yet we are slowly but surely making it walk, which is an end product of several years of hard work," says Snyder. To speed up the progress, the group hopes to migrate to a more powerful animation package such as Alias| Wave front's Maya.

"It's not like Jurassic Park or other movies that have walking dinosaurs. We are building the animation from actual bones, not artistic sculptures," Chapman says. "For the first time we can get an accurate feel for how these animals really functioned, thanks to computer graphics."

In particular, the paleobiologists observed that the triceratops's rear knees and front elbow joints could lock in place, similar to horses and cows. Using the 3D model, the group also calculated the volume of the dinosaur's skull, which is now believed to have weighed about 400 pounds-one third lighter than previously thought. In the near future, Snyder will use a fossilized line of tracks, believed to have been made by a triceratops, as the basis for an animated walk cycle. This will help determine whether it was capable of a trot or gallop, based on its center of gravity.
The museum is using the 3D data to study the dinosaur's physical capabilities. For instance, the paleobiologists are planning to produce a walk cycle using fossilized tracks to determine its locomotor limitations.

"The animation will give us the most reasonable hypothesis of triceratops, and the most testable, when compared to other hypotheses of locomotion," says Larsson. "Using actual trackway data and a scanned skeleton, a fairly realistic locomotor style can be narrowed down, as the bones can only fit into the tracks so many ways." Also, by looking at how the bones are forced to hold the body, the researchers can examine possible ranges of motor styles such as walking versus galloping. "With the help of modern analogs of these possibilities, we could start to make hypotheses of physiology (cold- versus warm-blooded)," he adds. As the work progresses, Chapman expects to uncover more revelations about the dinosaur by using the 3D data.

To illustrate the importance of the technology to the field of paleontology, the museum is displaying the models and information used in the triceratops reverse engineering process in the current exhibit display. According to Richard Benson, chairman of the NMNH's paleobiology department, using digital technology may enable museums to overcome space limitations by displaying scaled-down replicas of certain treasures. It will also allow scientists to share information digitally to further scientific advances in the field. "We're now at the point where we just found the tip of the dinosaur bone sticking out of the ground," analogizes Benson. "There's still a lot of work ahead before we see what has been hidden to us in the past."

Karen Moltenbrey is a senior associate editor at Computer Graphics World.

To see how the triceratops is believed to have walked, click on the following URLs. Your type of Internet connection will determine the image download speeds; please be patient.

Click here for a side view:

For a front view, click here:

The animations are provided courtesy of the Smithsonian Institution, National Museum of Natural History, Department of Paleobiology, in Washington, DC.

3D Systems ·
Autodesk ·
Cyberware ·
Nvision ·
Solid Concepts ·
Steinbichler Optotechnik ·
Vero International Software ·