Indoor Location Retrieval with Depth Images using 3D Shape Features

(3D Bild-Analyse und Synthese)
Betreuer: Anas Al-Nuaimi
Recognizing the location from captured images is a promising application of image retrieval algorithms. Matching the query images to an existing geo-referenced database like Google Street View enables mobile search for location related media, products, and services.
For the purpose of location retrieval, we are investigating the possibility of matching depth images (also called range images) representing the environment. The pixel values in a range image are proportional to the distance of the objects to the camera. Using a dataset captured at the LMT, we can render a representation of the environment made up of such depth images. Hence we plan to substitute a typical database made up of RGB images by one made up of depth images.
Depth images of buildings are mainly composed of large homogenous surfaces with little information inside them. The borders of these regions on the other hand, carry much more information. In this bachelor’s thesis, the student shall deal with the topic of depth image matching using 3D shape features.
Keywords: Computer Vision, Conent-Based Image Processing
Anforderungen: Matlab programming skills. Interest in image processing and computer vision. Preferrably C++ programming skills.