Segmentation of discrete point clouds using an extensible set of templates

An evolutionary approach to the extraction of object construction trees from 3D point clouds

“…For each type of primitive, its parameters maximizing an objective function (see eq. (1) in [17]) are computed. The primitive best describing the data is then selected among all fitted primitives at this step (see section 5 in [17]) and the corresponding points are removed from the point-set.…”

“…These two steps are then repeated until the number of points left in the point-set is below some given threshold. An alternative approach for fitting primitives chosen from a list of candidate primitives is described in [17]. For each type of primitive, its parameters maximizing an objective function (see eq.…”

“…(1) in [17]) are computed. The primitive best describing the data is then selected among all fitted primitives at this step (see section 5 in [17]) and the corresponding points are removed from the point-set. These two steps are repeated until the number of points left in the point-set is below some given threshold.…”

“…There are various techniques available depending on the domain of application, see e.g. the section 1.1 of [17] for a more comprehensive survey of existing methods. Common techniques used in reverse engineering are described in [54,4,5,55] and references therein.…”

“…After the segmentation, we need to fit patches to each cluster. This fitting step can either be done separately from the segmentation step as for example in [54,32,4,53] or be done jointly with the segmentation as in [44,1,17].…”