|M.Sc Student||Hershko Nir|
|Subject||3D Augmentations of 2D Maps|
|Department||Department of Computer Science||Supervisor||Professor Gershon Elber|
Maps of the world around us are used daily for many purposes such as navigation. 3D maps, with terrain elevations and 3D objects, allow to model and simulate natural phenomena such as water flow or line-of-sight, and assist in orientation.
This dissertation presents several methods for augmenting 2D maps into 3D, each focusing on a different aspect of the map. The recurring theme in these methods is that the acquisition of the data can be done using a modern smartphone, and that they allow to use crowdsourcing (obtaining the data using a large number of people). In all three methods, the phone’s GPS sensor is used to acquire the global position of the user. However, the elevation reported by it is not as accurate as the horizontal location and this must be taken into account when using this data. In the methods’ implementations, we use data from the OpenStreetMap (OSM) project. We use the map provided by this project as the 2D map to augment to 3D, and the repository of GPS traces uploaded by OSM contributors.
The first method presented in this work reconstructs a roads interchange in 3D using a 2D map of the interchange and GPS traces recorded while driving. To overcome inaccuracies in the elevation data of the GPS traces, we only use the relative elevations along the trace. In the result, any road segment without GPS coverage is interpolated between the endpoints of that segment, where the relative elevation between those endpoints is inferred based on other paths between these points.
The second method reconstructs a digital terrain model (DTM) from GPS traces recorded in an area. This method can augment an existing DTM by refining it and integrating the data from the GPS traces (even in case the terrain has changed). To overcome errors in the elevations of the GPS traces, we add an offset factor to each trace, and calculate the offsets that achieve the best correspondence between the traces.
The third method calculates heights of buildings. The user selects a face of a building and captures the direction vectors between the user and the four corners of that face, using the smartphone’s sensors. This data, together with the 2D location of the building, is used to estimate the building’s height. As part of this method, we developed an Android application that allows the user to collect the data.
In order to visualize the results of all these methods, a web-based application was developed to display an interactive 3D map, including terrain, roads and buildings, using WebGL technology. It is accessible in http://osm3d.cs.technion.ac.il/. This application was written in part by Eli Sherer as a part of a student’s project. The data displayed by this interactive map is based on OpenStreetMap data, with an existing globally available DTM, and augmented using the results of the methods presented in this work, that were uploaded to the server in advance.