|
The paper describes a method, that allows to use a single terrestrial photograph, with unknown exterior and interior orientation parameters, as a source of geographic information. The idea is based on creating the orthoimage from a terrestrial photograph by means of photogrammetric and computer vision rules. The inputs are a digital image, Ground Control Points and a digital elevation model in raster format. GCPs are necessary to determine the camera orientation by means of Direct Linear Transformation. For non-calibrated cameras a large number of GCPs is necessary to obtain more accurate results. The next step is to eliminate invisible parts of terrain by applying the viewshed analysis. Finally, the terrestrial photograph is orthorectified using the extracted part of the digital elevation model. Output from this procedure is a georeferenced orthophotograph. The complete, ready to use algorithm is written in MATLAB. The method was tested using present-day digital images taken with a not-calibrated Nikon D80 and old postcards – both from the Polish Tatra Mountains. Comparison with an existing orthophotomap from aerial imagery gave satisfactory results, especially for deeply shaded areas.
|