|
I would like to know if pictures taken from the roof of tall buildings to the ground can be used to map in OSM. The pictures won't be pointed vertically to the ground but they will have an angle, and that is my question, if those pictures can be transformed into something flat that can be used as imagery for JOSM or id. |
|
If the building is high and the terrain is fairly flat and you don't use data near the horizon, i feel you could use http://wiki.openstreetmap.org/wiki/MapWarper to warp or transform the photo to the existing osm map and then add the image to JOSM as background and add further data to the osm. You would need to identify a minimum of 4 locations on the photo surrounding the area you want to map to do the warp. You could also use a josm plugin to stretch your image to match the osm and edit - https://josm.openstreetmap.de/wiki/Help/Plugin/PicLayer |
|
Even professional aerial photographs intended for use in cartography are generally constructed from fairly wide strips, if only to keep the cost down, so at least part of the image is taken at a significant angle from the vertical. The process of correcting the resulting distortions is complex, and can only be done if the height of the camera above the land is known. That gets particularly difficult if the land is not flat. It is also necessary to correct for distortions introduced by the camera and its lens. Professional surveyors know the distortion profiles of their cameras (no lens is perfect), and some DSLR systems correct for the known distortions of their own lenses. Even the professionals don't always get it exactly right (see https://help.openstreetmap.org/questions/13320/satellite-image-accuracy and several similar questions), which is why we should not map directly from aerial imagery without first aligning it with multiple GPS traces or features known to be mapped accurately. Unless you really enjoy a challenge, I would support nevw's suggestion that you try to adjust the image to match four or more precisely known locations. The locations should be fairly well spaced, but not too close to the edges of the image, where distortions are likely to be greatest. |
