.

Geoinformatics_05-2009
article fig. 8: graph showing resulting deviations dh, where c1 and c2 is the designation of cameras. was used to calculate the building s height, which was subsequently compared with the measurement acquired from pixoview. as expected, the tests confirmed that there was no dependence on the accuracy of the exterior orientation parameters, meaning that we will get comparable results regardless of whether we use direct georeferencing or aerotriangulation. the main downside will be our ability to actually interpret where we were measuring the height to and from. the deviation (dh) from the geodetically measured height was on average 10 cm, although if the start and end points of the building and its given relative height are defined incorrectly, this error can be several times greater. for measuring, it is therefore preferable to choose images where the building is closer to the projection centre, since the pixel size is smaller and greater detail can be seen. however, a problem may occur in terms of good visibility of the given points necessary for measurement, because the nearer the structure is to the nadir of the image, the more the advantage of measuring in oblique images is lost. tab. 2: accuracy characteristics at with gcp, where c1 and c2 are oblique cameras. conclusions when testing positional measurements in pixoview, it is important to take great care with the quality of the input data (image resolution, exterior and interior orientation parameters, dtm), which has a direct effect on the degree of accuracy of the individual points determined in the application. we will get significantly better results by improving the quality of the exterior orientation parameters. with direct georeferencing, average deviation in the position at signalized points was 0.45 m. using the analytical aerotriangulation method, we improved accuracy and reduced the average deviation in the position at signalized points to 0.20 m. the altitude accuracy of the individual points then depends primarily on the quality of the dtm used, but this does not influence the accuracy when determining the relative height of structures. overall accuracy characteristics for at with gcp were divided into four distance intervals (d), according to the aforementioned high correlation between the accuracy achieved and the distance from the projection centre, and are shown in the following table (tab. 2). the achieved accuracy approximately corresponds to 1.5 times the size of pixels for all selected intervals of distance. one of the greatest advantages of oblique photography is that it allows us to measure the height of buildings, bridges, and in fact any kind of landmark contained in a photograph. previously, stereoplotting has been the only way to determine height proportions. but working in stereo requires special equipment and trained operators. now, using oblique photography and appropriate software, this information can be acquired without extensive knowledge of photogrammetry, with any computer using average hardware, and to a relatively high degree of accuracy. therefore, this technology has become attractive to a number of sectors - from town planning to the coordination of integrated rescue system teams, where it can be used to ascertain preliminary information on affected areas. last but not least, oblique images can also be used to create realistic photo-textured 3d city models. karel sukup, managing director, division of geoinformatics, geodis brno ltd., karel.sukup@geodis.cz jan sukup, application engineer, division of geoinformatics, geodis brno ltd., jan.sukup@geodis.cz patrik meixner, production manager, division of geoinformatics, geodis brno ltd., patrik.meixner@geodis.cz for more information, have a look at www.geodis.cz 40 july/august 2009
Powered by fluidbook tm