Digital Elevation Modeling oflnaccessible Slope by Using Close-range Photogrammetric Data

Digital Elevation Model (DEM) currently is extensively used extensively in vanous
applications such as for natural hazard assessment and monitoring of high risk areas. DEM
data source of inaccessible areas can be collected by using several methods, but mostly are
costly and requires sophisticated instruments. Due to these conditions, close-range
photograrnmetry offers a low cost alternative solution. Materials presented in this thesis are
based on the experiments to explain the application of close-range photogrammetry with the
aid of commercial digital pocket camera as DEM data collection tools, applied on
inaccessible slope areas. The analysis covers calibration of the camera and surveying
instruments, DEM data collections, data processing and visualization, together with DEM
quality measures. The data collections are accomplished on several study areas with different
topographical characteristics by using close-range photograrnmetry technique. The sampling
points were selected on stereo model, by using three types of sampling methods. The DEM
quality measures are assessed by following elevation interpolation error and volumetric
difference error analyses. The representation of the DEM is generated using TIN-based
(Triangular Irregular Network) approach. The result shows that the method is able to be
applied for three dimensional (3D) modeling of potentially unstable slope areas, with
accuracy of less than 15 em in RMS for elevation error and is less than 1% in volume error.
The result has indicated that topographical condition has not affected the accuracy of
generated DEM. Improvement of point density radically enhances the DEM's quality, up to a
certain level of point density beyond which the increment of the accuracy is not significant.
The difference setting of focal length has also influences the quality of captured images, and
drastically affects the accuracy of the DEM. If the accuracy of the DEM is a matter of
concern, the preferred sampling method is selective sampling, while if accuracy and DEM's
time generation are the concern the most effective sampling method is regular sampling
method. Since there was no permanent points on the observed slope surface, velocity and
direction of landslide could not be accurately determined. However the distribution of massmovement
and elevation changed on the slope surfaces can be modeled through spatialcalculation
of overlaying DEMs together with profiling of cross-section and longitudinalsection
of the generated DEMs.