Robust object segmentation and parametrization of 3D lidar data

Abstract

This article addresses the problem of robust signal processing of 3D lidar data prone to noise. After describing the characteristics of the lidar data given we describe how the data can be segmented in a robust manner. The approach is based on edge detection followed by region growing. We show how the segments can be described using parametric models. In the final step the segments are circumscribed using appropriate bounding objects. We motivate the individual steps of our approach and light up the mathematical background.