Contour Recognition and Similarity Search in Mountain-Image Databases
Computer based information systems become increasingly important for our all-day-live. In particular the touristic sector uses these solutions to augment sights with location based information. This thesis deals with retrieving several concepts and implementation strategies for such a system. It should be capable to automatically recognize mountains on images, making it possible to show additional context based metainformation. For processing purposes, reference images of mountains are needed. These can be created by a manual or semi manual process. Both approaches will be discussed by analyzing actual available height models. Additionally the main part of this thesis deals with various approaches for implementing a pattern recognition system in respect of the special mountain recognition application domain. In particularly a dual feature approach, using texture and contour based features, is used to overcome certain problems occurring by the use of a single feature type. Finally some concepts for combination of both approaches are discussed and based on all of the gathered information, a
recommendation for a complete software prototype is made.