Modeling Sparse Range Data

Abstract

To represent more concisely the spatial layout of objects in the environment, the three-dimensional data produced by the cooperative sensing process must be segmented into meaningful clusters. But which data points “belong” together? And what are “meaningful” clusters? The clustering problem (described in Chapter 1) is to partition objects, represented by measurements in a multidimensional space, into groupings reflecting the “natural” structure of the data, without the utilization of any a priori information about class membership.