In this case plane photogrammetry regularly would have been sufficient. But due to extremly bad road conditions – the surface was bumpy and the street had many significant potholes – the plane photgrammetry was very distorted. The mathematical model of plane photogrammetry is based on the assumption of a ideal plane surface. This assumption is sufficiently satisfied by most real road surface, but not in this case. Due to the vertical perspective of the drone this distortion is reduced significantly. Making marks on the road surface, which are essential for plane photogrammetry, is a safety relevant task, if done without road blocking. So using the aerial photos for computing point clouds saves oneself working on the road. With respect to the flat surface a vertical camera perspective has been choosen. For this flat surface this perspective is optimal, but vertical surface near the boundaries of the crossing, hedges and fences, are computed with significant lower accuracy. For computing these surfaces as well as the road surface an inclined perspective of the camera would have been choosen. This might be useful if the line-of-sight is relevant.At this crossing an accident happend between a left-turning car and a car drive straight forward. The final positions of the cars had not been measured. But images from the final positions exist.The cars involved had been to old respectively to exotic, so proper 3-D models of the cars have not been available. But the tarmac had been tinkered again and again and additionally the surface had several cracks, so the contact points of the wheels could be identified by the pattern of the surface. So the final positions of the cars could be estimated with sufficient accuracy. So the accident case could be solved. Without the detailed point cloud of the surface one had to try to find the surface pattern on-site on the road itself. That would habe been much more time-consuming and much more dangerous, too. A 3-D presentation of this point cloud is available. On most up-to-date hardware – desktops and laptos as well – the presentation runs in your web-browser. Older hardware and most of the smartphones and tablet will not be able to display the point cloud, because the point cloud is to detailed for that hardware. Examples with reduced point density, which should run on weaker hardware, too, will be prepared in future.