.

Geoinformatics_05-2009
article fig. 7: measuring tools in the pixoview application. for each variant. the measured coordinates were then compared with the coordinates determined by terrestrial (surveying) methods. and as is usual in aerial photogrammetry, the accuracy achieved was divided into an altitude variable, computed as the difference between the given altitude and the measured altitude (dz), and a positional variable, which was calculated as the mean deviation in the position for each measurement from the formula. the measured values are greatly affected by the horizontal distance from the projection centre, or the viewing angle at which we see the point in the image. the mean deviation dependence in the position on either one of the unfavorable factors can be statistically expressed using a correlation to describe mutual relations between variables. the result is a value or correlation coefficient ranging from -1 to 1, where -1 indicates indirect dependence, 0 indicates that there are no statistics between the variables to determine dependence, and values towards 1 show direct dependence. in our case, this means that the greater the distance from the projection centre, the less accurate our results are when determining point positions. the following graph (fig. 4) shows the mean deviations in the position at the individual distance intervals (d), and also compares all three methods of eo parameter determination. determining the altitude variable is somewhat easier. the correlation coefficient s dependence on the horizontal distance is practically zero for all three variants, so the height measurements have the same degree of accuracy as the digital terrain model used (fig. 5). so far, we have only taken into account signalized points which are clearly identifiable on the images. with other non-signalized points, such as the vertices of pavements and roads, the resulting deviations will be affected far more by interpretational errors. this means that especially in remote areas of the photographs, where the pixel size can be up to 20 cm, compared to 10 cm in nadir areas, our chances of identifying the measured points are much worse. in the case of building corners, this uncertainty is even greater. in the second stage of testing we also measured 450 non-signalized points, and the resulting coordinates were compared with the coordinates of these points on a digital cadastral map (dkm). a comparison of all the results achieved from testing the positional accuracy of pixoview measurements is shown in the graph below (fig. 6). when evaluating these results, it is necessary to take into consideration that we are dealing with monocular measuring; the dkm was used for the calculation of the accuracy, and the dkm s accuracy is also not entirely homogeneous. testing height measurements tab. 1: table showing correlation coefficients for cameras c1 and c2. within the testing of the pixoview application, a tool for measuring the height of structures was also examined. for this purpose, geodetic measurements of selected buildings were first taken, and the oblique length july/august 2009 38
Powered by fluidbook tm