Jan Böhm

A multi-sensor system for positioning in urban environments

Within the article a low-cost system for the provision of georeferenced terrestrial images in urban environments is presented. Based on an image with approximate exterior orientation from a low-cost GPS and a digital compass and a 3D CAD model of a visible model as provided from a 3D virtual city model, the exact location of the building in the image is detected automatically and used for a refined orientation of the image. The work is part of a project aiming on the development of a mobile device, which enables access to location-based services in a complex urban environment. The intuitive access to object-related information is realized by so-called telepointing. For this purpose, a spatial model of the users environment is mapped to an oriented image, allowing for the access to object-related information by pointing to the respective image sections. Since the provision of location-based services currently is one of the most promising markets for the use of spatial data in urban areas, these applications will also be discussed.

Orientation of Image Sequences in a Point-based Environment Model

In this paper we propose a method to determine the exterior orientation of each frame of an intensity image sequence using prior knowledge of the scene stored in a point-based environment model (PEM). The orientation is performed by tracking landmarks across the image sequence acquired with a calibrated camera. The landmarks are intensity features, which are automatically extracted from the PEM. The PEM can easily be acquired by long range 3D sensors, such as terrestrial laser scanners. The orientation procedure of the imaging sensor is solely based on spatial resection.In this paper we propose a method to determine the exterior orientation of each frame of an intensity image sequence using prior knowledge of the scene stored in a point-based environment model (PEM). The orientation is performed by tracking landmarks across the image sequence acquired with a calibrated camera. The landmarks are intensity features, which are automatically extracted from the PEM. The PEM can easily be acquired by long range 3D sensors, such as terrestrial laser scanners. The orientation procedure of the imaging sensor is solely based on spatial resection.

Towards Online Pose Measurement for Robots

We present a photogrammetric system for on-line pose measurement of a robot. The system is based on photogrammetric measurement techniques, namely resection. We describe the theoretical foundations of our approach as well as early details of our implementation and hardware set-up. The results achieved are compared to those of a commercial ball-bar system.

Automated interpretation of dense range data

We address the problem of automating the processing of dense range data, specifically the automated interpretation of such data containing curved surfaces. This is a crucial step in the automated processing of range data for applications in object recognition, measurement, re-engineering and modeling. We propose a two stage process using model-based curvature classification as the first step. Features based on differential geometry, mainly curvature features, are ideally suited for processing objects of arbitrary shape including of course curved surfaces. The second stage uses a modified region growing algorithm to perform the final segmentation. The results of the proposed approach are demonstrated on different range data sets.

Camera pan-tilt ego-motion tracking from point-based environment models

We propose a point-based environment model (PEM) to represent the absolute coordinate frame in which camera motion is to be tracked. The PEM can easily be acquired by laser scanning both indoors and outdoors even over long distances. The approach avoids any expensive modeling step and instead uses the raw point data for scene representation. Also the approach requires no additional artificial markers or active components as orientation cues. Using intensity feature detection techniques key points are automatically extracted from the PEM and tracked across the image sequence. The orientation procedure of the imaging sensor is solely based on spatial resection.

CAD-basierte Objekterkennung für ein multisensorielles Meßsystem

An der Universitat Stuttgart wird im Rahmen eines Sonderforschungsbereichs an einem System zur explorativen Erkennung und Vermessung von Werkstucken geforscht. Ein Ziel der Arbeiten ist der Aufbau eines Gesamtsystems, bestehend aus Beleuchtung, Aktoren, optischen Sensoren und Auswertealgorithmen, welches in der Lage ist, ein eingelegtes Werkstuck zu erkennen, dessen Position und Orientierung zu ermitteln und anschliesend fur dieses Teil spezifische, vor definierte Mes- und Prufaufgaben durchzufuhren. Ein Hauptaugenmerk liegt dabei auf der Automatisierung und Robustifizierung der Objekterken-nung, welche den Einsatz des Messystems fur verschiedene Werkstucke ermoglicht.