<XML><RECORDS><RECORD><REFERENCE_TYPE>3</REFERENCE_TYPE><REFNUM>7487</REFNUM><AUTHORS><AUTHOR>Bock,M.</AUTHOR><AUTHOR>Bowman,A.</AUTHOR><AUTHOR>Bowman,J.</AUTHOR><AUTHOR>Siebert,J.P.</AUTHOR></AUTHORS><YEAR>2002</YEAR><TITLE>Data extraction from dense 3D surface models</TITLE><PLACE_PUBLISHED>Proc. 15th Int. Conf. on Computational Statistics (Compstat 2002), Bremen, Germany, 2002. </PLACE_PUBLISHED><PUBLISHER>N/A</PUBLISHER><LABEL>Bock:2002:7487</LABEL><KEYWORDS><KEYWORD>Shape</KEYWORD></KEYWORDS<ABSTRACT>Stereo-photogrammetry involves the use of pairs of cameras to extend standard photographic methods by recovering depth information through triangulation (Ayoub et al. 1998). The resulting data consist of a point-cloud of threedimensional observations, providing a digital representation of the surface of the object of interest. The number of observations can typically be many thousands, providing a rich source of information on object shape. This can be represented efficiently as an irregular triangular mesh, in Virtual Reality Modelling Language form. This technology is currently being used in a study of the growth of children's faces. There have been very few quantitative studies of growth, especially in three dimensions. This study involves an additional longitudinal component in that repeated measurements are taken over time, from 3 months to 2 years for one cohort of children and from 3 years to 5 years for another. Stereo-photogrammetry provides a practical means of gathering these data. In particular, the very fast capture time involved is important in minimising the effects of movement in the subject. The clinical aims of the study are to contrast the facial shape and growth of healthy children with those from children who have been born with a cleft lip and/or palate and who have subsequently undergone surgical repair. This short paper describes the interface between the computer science, which has produced informative and realistic 3-D models of real surfaces, and the statistical analysis, which attempts to answer the research questions. In particular, several types of data, each extracted from the 3-D stereophotographic models, are described and methods of analysis appropriate to each type are briefly outlined. The types of data follow a natural progression of increasing complexity and information, namely: 1. point data consisting of between 30-40 anatomically meaningful landmarks; 2. curve data obtained by interpolating anatomical points and 3. surface data represented by a dense regular mesh of points, anchored at the anatomical landmarks and conformed to the surface of the 3-D model. </ABSTRACT></RECORD></RECORDS></XML>