Next: Performance Acceleration
Up: Range Data Registration
Previous: Pose versus Correspondence Search
When searching for corresponding features in the two data sets, there
are a variety of criteria that one can use for classifying points
as being similar or interesting or key points, such as colour, local surface
normals, local curvature shape, edgeness, texture, etc.
(22,73,66,62,38,78,26,14,36,8,70,37,1). All such features
are to varying degrees sensitive to noise and other conditions
such as occusion and thus, the extraction of such features is also
a challenging task. This contrasts with using the points directly
(19,5,21) with the shortcoming that such
algorithms often require good initialisation.
Bob Fisher
2003-07-13