AUTOMATIC 3-D POINT CLOUD CLASSIFICATION OF URBAN ENVIRONMENTS

Abstract

Figure 1. Urban environment classification with our approach. This paper is best viewed in color. Unless otherwise noted, the same color code labeling is used throughout the paper: brown for ground, red for facade, green for scatter, dark blue for pole/trunk, skye blue for wire. This paper addresses the problem of assigning a la-bel to three-dimensional data points collected from laser scanners. We are specifically interested in the application of environment modeling for autonomous robot navigation in natural and urban terrains. To capture contextual in-formation, we choose to work within the Markov Random Field framework. The approach used in this paper is a vari-ant of the Associative Markov Network (AMN), extended to learn directionality in the clique potentials, resulting in a new anisotropic model that can be efficiently learned us-ing a gradient-based method for non-differentiable func-tion. We validate the proposed approach using data col-lected from different range sensors. 1.