Towards Automatic Feature Line Modelling from Terrestrial Laser Scanner Data

For the representation of objects or scenes different representation forms are utilised. While for some tasks a point cloud representation or a triangulation of the point cloud is sufficient, a boundary representation or a wireframe model is necessary for other applications. For these the determination of feature lines is essential. In the past the generation of feature lines was a manual task. Nowadays, many research activities for the automated extraction of feature lines from digital images (e.g. roads and other infrastructure networks) can be found. In contrast to these activities, the extraction of feature lines based on terrestrial laser scanner data is rather at the beginning. This paper presents, after a short summary of the status of research, an approach for the automated feature line modelling. This approach consists of two main steps. In the first step feature line segments are automatically detected, whereas in the second step the modelling of the whole feature lines is performed. Within this approach also, the determination of quality parameters for the extracted feature lines is possible. Practical results of the presented methods are presented. They demonstrate the feasibility of the automatic determination of feature lines from TLS data and provide insight into shortcomings of the presented method. * Corresponding author.