(c) Copyright 1998 VR News
One of the most time consuming aspects of creating virtual worlds is the creation of the geometric models of objects. This can be particularly frustrating when you have a real physical version of the object. Fortunately there a variety of commercially available technologies that can be used to digitize objects from the molecular scale on up to multi-story buildings. Many of the commercial products are in use at service bureaus, so world builders with smaller budgets or infrequent demands can avoid the capital and time investments. A fairly long list of commercial vendors that offer sensors, software, and/or complete integrated systems is provided in a table that accompanies this article. This is not an exhaustive list, although collecting it certainly was exhausting! Many of the vendors provide links to related sites on their web pages and the table should provide an excellent starting place for your own exploration.
The process of 3D digitizing basically consists of a Sensing Phase followed by a Reconstruction Phase. The sensing phase collects or captures the raw data generates the initial geometry data, usually as a 2D boundary object, or a 3D point cloud. Sensing technologies are based on Tracking , Imaging, Range finding or a combination of these. The reconstruction phase is the internal processing of this data into conventional 3D CAD and animation geometry data such as NURBS and polygon sets. Sophisticated reconstruction software packages are available from scanner vendors and 3rd party software houses.
Tracking systems digitize by positioning a probe on the object and trigger the computer to record the location. The simplest tracker is a mechanical linkage or pantograph. Computerized Measuring Machines (CMM), such as those made by Faro Technologies and Romer Inc. are robust 3D mechanical trackers for manufacturing applications. Immersion Corporation's MicroScribe is a similar device marketed more towards the animation industry. Electromagnetic, ultrasonic, optical, gyroscopic and inertial trackers are also used in some commercial 3D digitizers. Trackers can suffer from interference problems either mechanical or electromagnetic. Object and environment space and materials need to be considered. Manual tracking systems requiring a large amount of patient, skilled labor, but they can digitize an object directly into polygonal models, eliminating the need for the reconstruction phase. Automated probe tracking systems produce point cloud data that will require reconstruction. One form of automated tracker is the Scanning Probe Microscope (SPM), which can be used to create 3D models of molecular scale objects.
Imaging starts with the capture of one or more 2D images that are then used as input to image processing algorithms to create the initial geometry data. Some imaging methods create a point cloud, while others use feature extraction to create an initial topology model. Active imaging systems project a moiré or grid pattern on the objects to provide known reference points and simplify the image processing. Passive imaging systems collect available light. General Dynamics Information Systems and VirtuoZo Systems International offer software that can use stereoscopic satellite and aerial images to create terrain elevation and building models. EOS Systems PhotoModeler uses manual feature digitizing in disparate viewpoint images to create object models. Geometrix introduced a product at Siggraph 1998 that applies automated photogrammetry to video images. This technology holds significant promise for rapid creation of 3D models using conventional and digital video cameras.
Another image approach to 3D scanning relies on a using a series of slices through the object. These can be obtained by actually cutting the object and taking optical photographs of the ends or by using advanced sensors such as Ultrasound, Magnetic Resonance Imaging (MRI), X-Ray Computed Tomography (CT), Confocal Microscopes. The slices can be used to produce volumetric data (voxels) or feature extraction might be used on the images to produce contour lines. Both forms of data can be readily converted to polygonal and surface models. Microscopic and volumetric systems are generally very expensive and proper treatment of the technologies is not possible in our limited space here. There are some excellent resources available on the web, such as Microworld's image analysis guide http://www.mwrn.com/guide/image/analysis.htm, the Confocal Microscopy pages http://www.cs.ubc.ca/spider/ladic/confocal.html, and the Stanfordís 3D reconstruction page http://biocomp.stanford.edu/3dreconstruction.
Range finding is related to Imaging as it usually results in a 2D array of range data (think Z buffer) that is then processed as an image. While 3D Pipeline Corporation has developed an ultrasonic range finder system to digitize caves and other large structures, optical systems are the dominant range finding technology. Both laser and white light range finding are available. Some optical range scanners also capture object color into an image for use as a texture map. An in-depth discussion of the technologies used for optical scanners can be found the article "Optical 3D Digitizers: Bringing The World To Life" in the June 1998 issue of IEEE Computer Graphics and Applications (CG&A).
All imaging technologies require multiple images from different views in order to create a complete object scan. Some scanning systems acquire these by rotating the object while others combine an imaging (or range finding) sensor with a tracker device allows the sensor to be moved around the object. Many of the laser or moiré scanner vendors market their sensors for attachment to CMM mechanical arms. The Polhemus HLS is a hand held laser scanner mated with their magnetic tracker. Tricorder Technology PLC has introduced a laser scanner using micro-electronic inertial tracking. These systems allow the scanner to be moved over a much larger area and even inside the object.
3D digitizing and especially automated scanning technologies can provide a world builder with highly accurate models. Currently the system costs and scanning/processing time can be an impediment to their use in world building, especially where the accuracy and detail is not required. An experienced modeler can often create low detail models faster and cheaper. However system costs have been dropping and recent innovations such as the MetaCreations RTG group's Galatea, discussed in the IEEE CG&A article, offer very fast model generation from point clouds providing on-the-fly level of detail generation.
This will be an interesting technology area to watch.