Früh and Zakhor developed efficient techniques for close range modeling of urban environments [1-3]. Combining the close range 3D model with a
Far range modeling of urban environments presents some distinct challenges due the operational disparity in the sensors that acquire information. To fuse the information from different sensors such as an airborne laser scanner and airborne camera, efficient techniques are being developed for fast and accurate registration, segmentation, and polygonalization. Some of the issues in far range modeling are: low sampling density of airborne laser scanner and aerial images, sampling inconsistency between camera images and laser scanner, loss of boundary and edge detail while rebinning the aerial scans, difficulty in segmenting the aerial scan based digital surface map (DSM) to identify buildings and vegetation, registration between aerial scans and images, etc. Our research concentrates on rebinning the scan points using edge preserving resampling techniques for laser scan points and combining information from aerial images to rectify the boundaries and improve the segmentation performance. These steps lead to an improved polygonal approximation of building rooftops, which are then texture mapped using the aerial images for accurate 3D models of building roofs and the terrain shape.