The Research Of Photos Mosaic Arithmetic And Application In Three Dimension Sight Modeling

Digital close-range photogrammetry, an important branch of photogrammetry, has been broadly used to take photos of varied targets for measuring their figure, dimension, location and deformation etc. besides surveying. Now digital close-range photogrammetry is developing from static state to dynamic, from 2D to 3D, also it will realize real-time and high precision.In information times, the application of digital close-range photogrammetry and some correlated sciences in 3D city modeling has becoming one of the research hotspots due to the greatly increased requirements of 3D city sight model especially 3D building model. 3D city model give a true and lively expression of city information, express city construction situation and planning design in a very different style, replace old 2D maps and 3D real objects with computer-aimed 3D diorama, provide user or decision-maker a more profundity and more intuitionistic knowledge, scientize the planning and designing of city, so it is very important to the sustainable development of city.In both geomatics and virtual reality, there are 3 kinds of 3D city modeling method: model-based modeling, image-based modeling, combination of these two. In model-based modeling method, 3D city geometry model need to be abstracted firstly. Then we can render the polygons using computer prototype or texture them with photos. Finally the realistic sight model is produced. In image-based modeling method, after changing the basic datum (dispersed images or series videos) into panorama, we can combine several panoramas into virtual panorama spacethrough appropriate space model to get the whole picture of city. The third method synthesizes the merits of those two. We can use model-based modeling in large simple area and image-based modeling in small complex area.In all methods, whether as texture-mapping or as modeling datum, images are very important. For improving modeling efficiency and shortening modeling periods, vast research and studying have been made in obtaining images using non-metric camera these years.Because of the limitation of general digital camera, more than two images are required to perfectly characterize a building or buildings. Also there exists move, zoom and rotate etc. between these photos. So establish a relation between them and project their coordinates to the same coordinates system must been solved.This thesis mainly studies mosaic method of series images. By totally analyzing the theory of image matching existed, this paper studies matching method fitting for digital close-range photogrammetry. For need of reconstructing independent building, put the point-based, line-based and region-based photos mosaics methods into forward. By introducing the plenoptic function, we apply multi-viewpoint concentric mosaics and cylinder panorama mosaics methods to reconstructing buildings.According as object-oriented program designing theory, this thesis also actualized point-based, line-based and region-based photos mosaics using Visual C++.