| Abstract / Kurzbeschreibung: |
For mobile robotics applications it would be beneficial to focus on visual sensory information only. In contrast to other types of sensory data, camera images offer additional information for range estimation, object classification, localization and navigation. Human vision proves that it is possible to realize these tasks without using laser scanners or ultrasonic sensors. In order to achieve good performance using visual data only, it is necessary to recognize, track and classify robust image features. Camera data representations of standard workspaces are usually simplified because the lighting conditions are constant at all times. In this type of environment, the use of color markings and maps is sufficient in order to achieve good results. In arbitrary environments, however, the lighting conditions vary. It is therefore necessary to extract additional information from the visual input data, such as natural and cognitive features. We propose a compact 1-dimensional qualitative symmetry operator which is capable of extracting horizontal and vertical symmetry from the image data. Throughout this article, the symmetry operator as well as different types of feature extraction are discussed. Additionally, a number of application fields is provided. |
