Object modeling from observation is the process of automatically or semi-automatically constructing 3D computer-aided design (CAD) models of real objects based on physical observations of the object.
Observations of the object may come in many forms. A video sequence taken of a moving object or taken while moving around an object provides 3D geometric information which can be used to determine the shape of the object. More direct observations can be obtained using a laser range finder, a device that sends out brief pulses of invisible light and then times how long it takes for the light to return. This gives the distance information that makes 3D shape reconstruction easier in some ways, yet doesn't provide any color or texture information.
Once the basic shape information has been computed, planes and curved surfaces need to be "fit" to the basic geometry to produce a reasonable CAD object containing familiar objects such as flat surfaces, spherical surfaces, and arbitrary Non-Uniform Rational B-Spline (NURBS) surfaces with clearly defined edges.
These CAD models can then be modified with standard CAD tools, used in animation or computer-assisted movie productions, or put on the Web in VRML format so that they can be interactively viewed by web readers.
Stellar Science's Chief Scientist, Dr. Conrad J. Poelman, who has published numerous technical reports and papers on the subject of modeling from observation, is spearheading this research and development project.