Abstract

In the present paper, we address the problem of recovering the true  underlying model of a surface while performing the segmentation. A novel  criterion for surface (model) selection is introduced and its performance for  selecting the underlying model of various surfaces has been tested and  compared with many other existing techniques. Using this criterion, we then  present a range data segmentation algorithm capable of segmenting complex  objects with planar and curved surfaces. The algorithm simultaneously  identifies the type (order and geometric shape) of surface and separates all the  points that are part of that surface from the rest in a range image. The paper  includes the segmentation results of a large collection of range images obtained  from objects with planar and curved surfaces.