Sometimes, in surface deformation studies, the use of fixed control points to recover surface coordinate systems is not practicable. In such cases, surface fitting techniques can be employed to locate the best surface coordinate match position in three dimensions. In computer vision, the 'iterative closest point' (ICP) concept has been a popular basis for numerous versatile surface fitting algorithms since 1992. However, it is suggested here that, when surface models are often only 2 1/2 -dimensional, and when they have good initial orientation estimates, (as occurs in many cases encountered in surveying), the relatively complex ICP algorithm may be bypassed in favour of a simpler least squares technique. Recognizing the success of the digital photogrammetry concept for automatic point correspondence determination, the writers have registered a series of surface models for a dental erosion study by least squares minimization of surface differences at corresponding points. However, as with many deformation problems, differences between the surfaces were to be quantified. This provided an additional challenge to the surface matching, as difference detection has not been relevant in image matching. Match results have been analyzed to gauge the algorithm's soundness. Use of the simple image matching concept has been vindicated by the success of the dental project.
|Number of pages||10|
|Publication status||Published - 1999|