We match photogrammetric point clouds with 3D city models in order to texture their wall and roof polygons. Point clouds are generated by the Structure from Motion (SfM) algorithm from overlapping pictures and videos that in general do not have precise geo-referencing. Therefore, we have to align the clouds with the models' coordinate systems. We do this by matching corners of buildings, detected from the 3D point cloud, with vertices of model buildings that are given in CityGML format. Due to incompleteness of our point clouds and the low number of models' vertices, the standard registration algorithm ``Iterative Closest Point'' does not yield reliable results. Therefore, we propose a relaxation of a Mixed Integer Linear Program that first finds a set of correspondences between building model vertices and detected corners. Then, in a second step, we use a Linear Program to compute an optimal linear mapping based on these correspondences.
CITATION STYLE
Goebbels, S., Pohle-Fröhlich, R., & Kant, P. (2018). A Linear Program for Matching Photogrammetric Point Clouds with CityGML Building Models (pp. 129–134). https://doi.org/10.1007/978-3-319-89920-6_18
Mendeley helps you to discover research relevant for your work.