Nudged Elastic Band in Analysis of Range Image High-contrast Patches

Abstract

Range images are more interested in lately for several reasons, such as, range images could be used in object recognition; the scene geometry of the 3D world could be comprehended more effectively by using range images. As there is a distance between the laser scanner and the nearest object for each range image pixel, hence, every range image may be thought as a vector in a high dimensional space W. It is very difficult to study a set of range images X ⊆ W, because X has very high dimension and it is very sparse in W. An efficient way for analysing range images is to study the space of small range image patches. The nudged elastic band method is a main tool for searching minimum energy paths in computational chemistry. In this paper, Morse functions are created by the sampled data from range image high contrast small patches, and then one-dimensional cell complexes are built from the Morse functions by using the nudged elastic band method, topological features of range image data are detected by a sequence of cell complexes. Particularly, we experimentally show that there exist subspaces of high-contrast 3×3, 4×4, 6×6 and 7×7 range image patches whose homology is that of a circle.