V.Ravichandran
Deputy Director
Survey of India,
Email: firehon00@yahoo.com
V.K.Naidu
Officer Surveyor
Digital Elevation Model (DEM) has many potential real-life applications. Large Scale mapping in 1:2000 scale with 1m Vertical Interval was carried out from 1:8000 (B&W) Aerial Photographs for Tumkur town of Karnataka State. GPS Control Points and Levelled Heights were used for running AT in Leica Photogrammetric Software Suite. The RMS error was kept within one sigma. In this case study a small portion of Tumkur town that has undulating terrain (approximately 4sqkm) is taken for the study. DEMs were generated by various methods viz. automated DEM (0.19m cell size), DEM through mass points (20m interval) with break lines, DEM from manually generated contours (V.I. 1m), DEM from TIN. All these DEMs are compared against each other and checked with 106 randomly observed height points through stereo fusion.
The study indicates that DEM generated through mass points with breaklines closely agrees with the heights measured through stereo fusion. The minimum and maximum difference being 0m and 0.516m respectively with the average elevation difference of 0.156m and SD of 0.120m. The DEM generated from manual contours and automated DEM closely resembles with the values of minimum, maximum, average and SD more or less equal. Though manual contours take into account the breaklines while capturing them in 3D, the DEM through manual contours could not match that of DEM through mass points (with breaklines) as the intervening areas of the contours need to be interpolated. Editing of TIN is quite time taking and was not attempted and the DEM through TIN is comparatively poor with minimum and maximum difference of 0m, 2.113m with average difference of 0.220m and S.D. of 0.259m.
In terms of time taken manual contouring took 6hrs for 4sqkm, mass-points (with breaklines) took 10hrs for the same area and TIN generation took 0.5hrs. Automated DEM took 0.5hrs while generating DEM from mass-points, manual counters, and TIN took only 5minutes.
It is concluded that mass points with breaklines provides accurate DEM compared to other methods and especially for flat and densely built-up areas, it is a better option for accurate DEM. For hilly areas, measuring mass points at close interval is quite difficult and in this case contours with spot heights will yield sufficiently accurate DEM. Obviously, automated DEM is the quickest of all the DEM generation processes but there is some chances of error creeping in and its reliability is not assured unless, if there is a way to edit those selected points before DEM generation process. DEM through TIN may yield a better DEM, provided, it is edited but it is quite laborious.