Semi-automatic Landform Mapping at Steep Slopes

The traditional mapping on digital aerial photos and elevation models has its limits at steep slopes where the surface is only very sparsely sampled even in high resolution data sets. As a consequence, there is too few information available to perform a detailed mapping. In this paper we present a framework for augmenting terrain data with additional photos and show how to utilize this additional information in the mapping process. The photos are matched to a textured elevation model by marking correspondence points through a simple point-and-click interface, which are then used to solve for the camera projection matrix. To compensate for different lighting conditions during acquisition, a histogram matching is applied in order to align the color distributions of the photo and the data set. The photos are rendered onto the terrain using projective texture mapping. If multiple overlapping photos are available, the best views are determined and blended together to avoid visible seams. The photos are incorporated into the segmentation process by computing a patch-wise, adaptive parameterization of the terrain geometry which defines a mapping of the surface into the plane with minimized distortion. The resulting image patches are connected appropriately at their boundaries and used as input for the segmentation algorithm.