Fitting an unknown number of lines and planes to image data through compatible cluster merging

摘要

A compatible cluster merging algorithm is presented that is specially designed to find the optimum number of linear, planar and hyperplanar clusters (i.e. clusters that lie in a subspace of the original space), when an upper bound on the number of clusters present is known. This algorithm is shown to be superior to more traditional validity-measure-based approaches. The effectiveness and advantages of the proposed technique in 2D and 3D applications is demonstrated with both synthetic and real data. The proposed applications include character recognition, obtaining straight-line descriptions of intensity edge images and obtaining planar approximations of 3D (range) data.