Volker Rehrmann

Fast and Robust Segmentation of Natural Color Scenes

We present a fast and robust system for color-based segmentation. The system is based on hierarchical region-growing on a special hexagonal topology. In contrast to common region-growing techniques it is independent of the starting point and the order of processing. It is generally applicable in natural color scenes and algorithmically efficient. The use of local and global information and a new color similarity measure contribute to the robust segmentation results. The system is successfully applied in two difficult applications from the field of autonomous vehicle guidance.

Traffic Sign Recognition Based On Color Image Evaluationion

This paper introduces some principles of an evaluation system, CSC (color structure code). The authors describe how to apply the CSC and its evaluation for a real-time traffic sign recognition (TSR) in vehicles. Several images are presented that provide an idea of the quality and reliability of this approach.

Ideogram identification in a realtime traffic sign recognition system

A robust system for the automatic detection of traffic signs has been developed at the Image Recognition Laboratory of the University of Koblenz. This traffic sign recognition (TSR) system was originally designed to localize traffic signs and to recognize their classes, e.g. prohibition signs, danger signs, beacons, etc. The exact identification of traffic signs is added. Traffic signs are identified by the interpretation of their ideograms realized by different modules in our TSR. The first module detects the position and direction of arrows. A second tool recognizes numerals and interprets them as reasonable speed limits. A third one is a general nearest neighbor classifier applied to three classes of ideograms (prohibition sign ideograms, speed limits, arrows on mandatory signs). The fourth module is based on neural nets and applied to two of these classes. Some of these components are used competitively in our realtime TSR. The use of several results from different tools increases the safety and provides high recognition rates.

On hierarchical color segmentation and applications

A hierarchical color segmentation technique is introduced. It combines the advantages of local (simplicity and quickness) and global (robustness, accuracy, avoidance of chaining mismatches) region growing methods. The method is implemented for a traffic sign recognition system.<<ETX>>

Object Oriented Motion Estimation in Color Image Sequences

This paper describes a color region-based approach to motion estimation in color image sequences. The system is intended for robotic and vehicle guidance applications where the task is to detect and track moving objects in the scene. It belongs to the class of feature-based matching techniques and uses color regions, resulting from a prior color segmentation, as the matching primitives. In contrast to other region-based approaches it takes into account the unavoidable variations in the segmentation by the extension of the matching model to multi matches. In order to provide extended trajectories, color regions that could not be matched on the feature level are matched on the pixel level by the integration of a correlation-based mechanism. The usage of color information and the combination of feature-based and correlation-based matching leads to robust and efficient algorithms. The system was applied to a motion segmentation task in vehicle guidance. Experiments on more than 1000 natural color outdoor images, taken from a moving car, show promising results.

Industrial Raman mapping spectroscopy for mining applications

A Raman mapping system for detecting and discriminating minerals such as dolomite, marble, calcite and pyrite is demonstrated. The system is built from components that are suitable for industrial conditions. Together with a signal processing and a classier the system was shown to be capable of discriminating between several important classes of mineral. The technique is a potential alternative to sensing methods currently used for mineral sorting.

A Fast Generator for the Hierarchical Structure Code with Concurrent Implementation Techniques

The sequential generator for the Hierarchical Structure Code (HSC) developed by G. Hartmann can be parallelized on a processor farm with an improved run time of a few seconds. To achieve a further improvement we suggest exploitation of its inherently asynchronous, parallel (i.e. concurrent) structure resulting in a concurrent OCCAM-HSC-generator.

Asynchrone Parallelisierungsstrategien zur Generierung des Hierarchischen Strukturcodes (HSC)

Im Fachbereich Elektrotechnik der Universitat-Gesamthochschule-Paderborn wurde unter Leitung von Professor Hartmann der sogenannte Hierarchische Strukturcode (HSC) zur Bildverarbeitung entwickelt (vgl. [Hartmann 1987]), der ein von einer Kamera aufgenommenes Grauwertbild in Strukturinformationen uber helle und dunkle Objekte und deren begrenzende Kanten zergliedert und diese als hierarchische Datenstruktur verwaltet. Die im Hexagonalraster abgetasteten Bildpunkte werden in sich uberlappende Inseln eingeteilt und in fest vorgegebenen Inselgruppen (auf verschiedenen Hierarchieebenen) jeweils auf zusammenhangende Strukturen untersucht, woraufhin Codebaume des HSC entstehen. Da diese Operationen lokal ausgefuhrt werden, kann man die Verarbeitung jeder einzelnen Inselgruppe als eigenstandigen Prozes auffassen. Auf dieser Basis wurden von den Autoren verschiedene Parallelisierungsstrategien entwickelt und in OCCAM2 auf Transputern implementiert, um die HSC-Generierung zu beschleunigen. Dabei wurde die Laufzeit von mehreren Minuten auf einer SUN4 in den Bereich weniger Sekunden reduziert.

Automatische Verkehrszeichenerkennung mittels Echtzeit- Farbbildanalyse

Prof. Dr. Lutz Priese ist Professor für Theoretische Informatik an der Universität Koblenz-Landau und Leiter des Labors Bilderkennen. Seine Forschungsinteressen liegen in der Theoretischen Informatik in Grundlagenfragen paralleler und nebenläufiger Systeme und in der Praktischen Informatik in Computer Vision. Adresse: Universität Koblenz-Landau, Abteilung Koblenz, Fachbereich Informatik, Rheinau I, D-56075 Koblenz. Tel.: (0261)9119-414, Fax.: (0261)9119-496, E-Mail: priese@uni-koblenz.de