GEOMETRIC MODELING USING POINT CLOUDS 297Optical scanners are non-contact devices that use light to interact with the object. They are
probably the broadest and most popular with relatively fast acquisition rates. Optical methods can be
classified based on the principle of operation used:
- Triangulation uses location and angles between light sources and photo sensing devices to
deduce position. The light source is a high-energy laser and the sensor is a video camera. The
method can acquire data at a very fast rate and the accuracy depends on the resolution of the
camera. - Ranging methods measure distances by sensing the time of flight of the light beams. Popular
methods are based on lasers and pulsed beams. - Structured lighting involves projecting patterns of light upon the surface and capturing an image
of the resulting pattern as reflected by the surface. The image is then analyzed to determine the
coordinates of a data point on the surface. This method can acquire large amount of data with a
single image, but the analysis to determine the coordinates is quite intricate. - Image analysis and photogrametry methods are similar to the analysis of image performed in
structured lighting. The difference is that instead of structured images, stereo pairs are used to
provide the information to determine height and coordinate position. - Interferometry methods measure distances in terms of wavelengths using interference patterns.
These methods are very accurate, the accuracy being of the order of the wavelength (hundred of
a nanometer).
Acoustic range finders use sound reflected from the surface and magnetic methods use magnetic
field measurements using probes. These methods are prone to high noise and thus have not yet been
commercially used for engineering measurements. However, they are reasonably accurate for geological
survey where the tolerance on measurement accuracy is relatively high. Magnetic Resonance Imaging
(MRI) and Computed Tomography (CT) use images obtained by magnetic field/radiation for sensing
the internal geometry of the object being scanned. These methods are primarily apt for biomedical
applications and the like where data points need to be acquired for the internal geometry not accessible
to light.
Tactile methods contact the surface to be digitized using mechanical arms. Sensing devices at arm
joints determine the relative coordinate locations. Robotic arms like a coordinate measuring machine
(CMM) can be programmed to follow a path along the surface and collect accurate data. These
methods are among the most accurate but are slow in data acquisition.
To summarize, every acquisition method interacts with the surface of interest by some phenomenon.
The speed at which the phenomenon operates and the speed of response of the sensor determines the
speed of data acquisition. The amount of analysis needed to compute the measured data and the
accuracy are also determined by the sensor type selected. There are many practical issues when
acquiring usable data, the major ones being calibration, accuracy, accessibility, occlusion, fixturing,
multiple views, noise, incomplete data and surface finish. Nevertheless, it is possible to obtain
adequate point cloud data in reasonably short time period using methods appropriate to an application.
10.3 Surface Modeling from a Point Cloud
Shape characterization from the point cloud is a key step in converting discrete data into a set of
piecewise continuous surfaces. The organization of data and neighborhood information are important
issues to address at this stage. The procedure for constructing of piecewise surface patches depends
not only on the type of data (and hence the source) but also the type of model required. The type of
model to be created depends on the intended use. For example, some applications may only require