Abstract for presentation at Combined 5th Trans Tasman Survey Conference and 2nd Queensland Spatial Industry Conference 2006

Generation of DEM from Non-metric Digital Camera Images and GPS Data over Mountainous Areas

  • Dr Khairul Abdullah, Department of Geomatic Engineering, Universiti Teknologi Malaysia,, Malaysia
  • Ms Diana Muhammad, Department of Geomatic Engineering, Universiti Teknologi Malaysia,
  • Ms Aliza Sarip, Department of Geomatic Engineering, Universiti Teknologi Malaysia,

    • In November 2003, a series of landslides occurred at several spots along the only access road leading to Gunong Raya, the highest mountain in Langkawi Island, Malaysia cutting off communication to the summit. As part of the effort to identifying these areas, aerial photography was conducted using a 6.0 mega pixel non metric digital camera at a flying height of about 900 m over the affected mountainous area. In this paper, the generation of DEM from such non-metric images was investigated using a single model of the 1: 4000 scale photography derived from a pair of non metric images of size 1440 x 2160 pixels with pixel size of 10.5 μm. A well developed area with an altitude of 900 m above mean sea level was chosen as the test site for this investigation. The height differences between the highest and the lowest point within this site is about 60 meters. For the horizontal control, a total of 14 well distributed points were observed using the rapid static GPS approach. Some of these GPS points are later used as the GCP for the formation of the 3D model of the test site. Due to the unavailability of detailed height information of the site, a leveling survey over the test site was conducted resulting in point height accuracy of 5 mm. The DEM derived from this survey is used as the reference DEM for the performing analysis. The creation of the 3D stereo model of the test site was carried out using the DVP® software version 5.5. Two method of DEM generation were adopted; the auto-correlation method and the manual method. Since the images are of the non metric type with no fiducial marks, the four corners of the images were adopted for the interior orientation. In generating the DEM, grid spacing of 1m resulting in 125 points were used. Many combinations of the GCP points were tested in forming the 3D model with the final 3D model adopted for further analysis using 10 GCP points which are well distributed throughout the test site. The DEM generation method and its associated problems, test results and comparison with terrestrial measurements are described here. The test results indicate that the estimated accuracy (r.m.s) of 0.3m can be achieved in the DEM generation.

    Conference Organiser - ICMS Pty Ltd