This paper proposes a new tracking algorithm within a 3D-SLAM framework that takes segmented range images as observations. The framework has two layers: the local layer tracks the view path and correspondences across the image sequence, using ambiguous landmark surfaces, and provides multiple hypotheses. The global layer conjectures and closes loops. We analyze the effect of different local trackers on the global layer. An interpretation tree (IPT) is compared to a new algorithm, orthogonal surface assignment (OSA), which attempts to track a building coordinate system. OSA is specialized to man-made work spaces. But our indoor experiments, performed with a rotating laser scanner, show clear advantages over the more general IPT: OSA covers the more relevant portion of the solution space, avoids the accumulation of rotation errors, and can estimate a unique translation in cases where IPT fails

Reliable mapping and self-localization in three dimensions while moving is essential to survey inaccessible work spaces or to inspect technical plants autonomously. Our solution to this 3D SLAM problem is novel in several respects. First, a new rotating laser-scanning setup is presented for acquiring point clouds and reducing them to surface patches in real time. Second, the SLAM algorithms work entirely on highly reduced, attributed surface models and in 3D. Third, we propose a novel system architecture of an extended Kalman filter (EKF) for 3D position tracking, cooperating with a 3D range image understanding system for matching, aligning, and integrating overlapping range views. The system is demonstrated by an indoor exploration tour.

This paper presents a novel solution for integrating range views into 3D geometry models. Unlike previous approaches, sparse surface and relation attributes are fused directly, without relying on the original range data or a dense mesh of data points. Corresponding surface patches in overlapping views are found and registered automatically. The main focus is on the integration of registered views. Two algorithms for merging partial boundary polygons, one following the 'explained' and the other following the 'enlarging' contour, are presented and methods for updating surface areas, quadric approximations and other attributes are provided. The technique is relevant for 3D map making in autonomous vehicles: the current surface map is used for symbolic action planning while being extended by incoming sensor data. Preliminary experimental results indicate that the prototype system copes well with large measurement and segmentation errors.

The automatic segmentation of scattered 3D points into few, meaningful surfaces is a key step in range image processing, both in matching a scene with stored model objects and in reconstructing model objects from sensor data. Few segmentation algorithms exist for range images that are sufficiently robust, predictable, general, and efficiently parallelizable. The main contribution of the paper is a new split-and-merge approach with dynamic top-down triangulation that supports hierarchical resolution and backtracking. Feedback from model driven matching to data driven segmentation is provided in order to meet the recognition goal in a joint effort. Border control, quality criteria for surfaces and for partition points as well as unsupervised clustering help to achieve robustness and independence of image content. Results from a simple, high-noise range image are presented.>