Study On Mapping And Optimization Of Facade Texturesfor Realistic Three-dimensional Models Of Buildings

With the rapid development of city information and digital city, to create realistic 3D model of city buildings become a hot research topic in computer vision, photogrammetry and other fields. In recent years, with the rapid development of 3D laser scanning technology, the buildings’ 3D spatial data can be efficiently obtained to reconstruct 3D model with accurate geometric information, and in order to reconstruct realistic 3D model, texture mapping on 3D building model is an important step. Handheld digital camera can flexibly take images with low cost, which can obtain clear facade textures and provide a good source for texture acquisition. However, these images could be acquired at any angle without initial value of exterior orientation elements, or some of them could have tree occlusion and illumination difference, some others do not contain all the texture. This thesis studies the texture mapping and texture optimization method of building facades based on images obtained by handheld digital camera.As digital images acquired with digital cameras do not have initial value of external orientation elements, and images have large dips because of close distance shooting, this thesis uses an initial value un-dependent method-- unit quaternion approach to solve exterior orientation elements of images. Large number of images are taken in order to select one with good texture information. This thesis takes three factors, the texture area, tilt angle and tree occlusion into consideration and sort image scores to automatically select the best image from a large number of images. Some images could unavoidably obtain tree occlusion, which will affect the integrity of the texture, this thesis recognize trees occlusion areas by hue information and compensate for the occlusion areas by acquiring adjacent images, which do not have tree occlusion in the same area. As building facade has large span and are usually taken with close distance, the entire building facade may not be concluded in one image. More than one image needs to be taken and mosaiced to generate complete facade textures. This thesis utilizes ant colony algorithm to optimize seam line and automatically mosaic textures, then it adjusts the brightness information for image dodging. At last, it maps optimized facade texture on 3D building model.According to the proposed methods, this thesis uses digital camera to take images of buildings on campus, and maps textures on 3D building model with optimized facade texture. The obtained realistic 3D building model verifies our methods.