OSM buildings from ordinary photos

Vladimir Elistratov

Details about the talk:
I discuss my experiments with a number of photogrammetry techniques to reconstruct building outlines from ordinary photos. High resolution satellite imagery is not available for all areas. Even if it is available, some buildings appear vague or are hidden by clouds. That was my motivation to provide photogrammetry tools for the OSM community. I started from the predecessor to Microsoft Photosynth called Bundler, a tool developed in the University of Washington. In combination with another tool from the University of Washington called Clustering Views for Multi-view Stereo (CMVS), one is able to obtain automatically more or less dense set of points in 3D space from a number of photos. The number of photos could be from 5 to several hundred with processing time ranging from minutes to hours. I came to the conclusion that this method is more suited for researchers than for the most of OSMers. By that time I learnt Google Sketchup and its Photo Matching tool that allows reconstructing manually a whole 3D model from 1-2 photos. This tool is based on projective geometry laws. Eventually I have developed a plugin that could import OSM files and GPS tracks to Google Sketchup and export an OSM file from Google Sketchup. I expect that this is to be a useful tool for OSMers mapping areas without high resolution imagery.

Details about the speaker:
Vladimir Elistratov joined OSM in February 2008 and has been actively participating in the work of Russian OSM community since its formation in March 2008. He has been mapping Tula, a regional center 200 kilometer to the south from Moscow. He has been experimenting with application of photogrammetry methods for mapping in OSM. He works as a software developer of mapping and charting applications.

Theme: map