The meaning of 3D shape and some techniques to extract it

In the context of information extraction, the question to begin with is: Which semantic information can a 3D model be expected to contain? The truth is that 3D data sets are used for conveying very different sorts of information. A 3D scanning process typically produces a number of textured triangle meshes, or maybe just a large set of colared points. So a single 3D scan is conceptually very much like a photograph; it is a result of an optical measuring process, only with additional depth information. One 3D scan may contain many objects at the same time, or a set of 3D scans may contain different views of the same object. The notion of an object is highly problematic in this context, of course, and must be used with care. For the time being, we define a 3D object pragmatically as a distinguishable unit according to a given interpretation or a given query. So the notion of what is regarded as an object may change as a function of interpretation and query context.