This item is available under a Creative Commons License for non-commercial use only
The creation of detailed 3D (three-dimensional) building models has become an area of considerable research over the last couple of decades. The accurate modelling of buildings offers LBS (Location Based Services) applications in planning, cultural heritage, tourism and e-commerce among others. The approach taken by the majority of contemporary modelling systems that use terrestrial imagery taken from arbitrary locations requires the user to carry out manual correspondences across the image set. These correspondences are used for two purposes. Firstly, the correspondences are used to determine the exterior orientation parameters (position and orientation) of the cameras used to capture each image. Secondly, (and more importantly) the correspondences are used to group highlighting primitives (points or lines placed in the images of the image set) that highlight the same building feature. Requiring the user to carry out these correspondences manually is a very time consuming process, which greatly limits the scalability of such systems. This thesis investigates techniques that reduce the amount of user interaction required to model a building. SAMATS, a Semi-Automated Modelling And Texturing System, has been created to demonstrate these techniques. Modelling systems that automate the modelling process generally add restrictions and place constraints on what building types the system is capable of modelling. These restrictions and constraints make such systems less flexible, greatly reducing that usefulness. SAMATS demonstrates that it is possible to automate the steps that have traditionally been carried out manually while maintaining the system’s flexibility to produce geometrically accurate photorealistic 3D building models of arbitrarily shaped buildings. SAMATS does not require the user to carryout correspondences manually. SAMATS makes use of georeferenced terrestrial imagery so that the cameras’ exterior orientation parameters are provided. Also, while contemporary modelling systems have required the user to carry out the correspondences manually in order to group highlighting primitives, SAMATS demonstrates how these correspondences can be determined automatically using the georeferencing provided, This makes it possible to eliminate the manual correspondence step from the modelling process completely, reducing user inter action substantially. Although obtaining accurate positional information for the imagery is still a bottleneck, as this information becomes more readily available with the use of GBS (global positioning system) enabled cameras, digital compasses, and gyroscopic sensors, the reduced user interaction offered by SAMATS’ approach will likely outweigh the burden of providing this additional georeferencing information.
Hegarty, J. (2006). SAMATS: Semi-automated Modelling and Texturing System. Masters dissertation.. Dublin Institute of Technology. doi:10.21427/D75P79