UAV-based remote sensing of the slow-moving landslide Super-Sauze

Transcription

1 Universität Stuttgart UAV-based remote sensing of the slow-moving landslide Super-Sauze U. Niethammer, S. Rothmund, M. Joswig

2 Motivation The four main objectives for our remote sensing campaign were: The development of a low-cost inhouse remote sensing system The acquisition of high resolution airborne photographs A GCP-coverage on the entire landslide Finally the generation of a high resolution ortho-mosaic

3 Super Sauze landslide

4 Super Sauze landslide The Super-Sauze landslide is located 100 km Southeast of Grenoble (Southern French Alps)

5 Super Sauze landslide The Super-Sauze landslide is located 100 km Southeast of Grenoble (Southern French Alps) The landslide is active since the 1970 s

6 Super Sauze landslide The Super-Sauze landslide is located 100 km Southeast of Grenoble (Southern French Alps) The landslide is active since the 1970 s The length of the landslide is about 850 m, the width about 200 m

7 Super Sauze landslide The Super-Sauze landslide is located 100 km Southeast of Grenoble (Southern French Alps) The landslide is active since the 1970 s The length of the landslide is about 850 m, the width about 200 m Its total volume is estimated to be m 3

8 Super Sauze landslide The Super-Sauze landslide is located 100 km Southeast of Grenoble (Southern French Alps) The landslide is active since the 1970 s The length of the landslide is about 850 m, the width about 200 m Its total volume is estimated to be m 3 Displacement rates range up to 0.4 m per day

9 Developed UAVs The airborne image acquisition was done by a quad-rotor system We developed this system especially for difficult alpine terrain

10 Digital camera Two digital cameras were tested on our quad-rotor system A high-end compact camera Fuji F30 (160 g weight) A low-budged compact camera Practica 8213 (110 g weight) We decided to use the lighter model because of windy conditions

11 Image acquisition

12 Image acquisition First tests of the quad-rotor system took place in September 2008

13 Image acquisition First tests of the quad-rotor system took place in September 2008 In October 2008 the entire landslide was imaged (1486 photographs)

14 Image acquisition First tests of the quad-rotor system took place in September 2008 In October 2008 the entire landslide was imaged (1486 photographs) Fligth altitude over ground was between 20 m and 250 m

15 Image acquisition First tests of the quad-rotor system took place in September 2008 In October 2008 the entire landslide was imaged (1486 photographs) Fligth altitude over ground was between 20 m and 250 m This leads to resolutions between 6 mm and 8 cm per pixel

16 Ground Control Points A reference was necessary for further image processing 199 GCPs were marked on the surface of the landslide

17 Ground Control Points The ground control points were measured using a DGPS-system The accuracy of the DGPS-measurements was at about 3 cm

18 Image processing Three image processing steps were applied: Correction of the optical distortion of each image Ortho-rectification of each image Image matching and color balancing in order to obtain an ortho-mosaic

19 Optical correction First the barrel distortion of each photograph had to be corrected

20 Optical correction Resulting photograph without optical distortion

21 Image rectification Airborne photographs naturally have perspective distortion caused by tilt of the camera This distortion had to be corrected for each photograph

22 Image rectification projective rectification polynominal rectification affine triangular rectification Three different methods could be performed for image rectification

23 Image rectification projective rectification polynominal rectification affine triangular rectification Most of the photographs were corrected using the projective rectification But this method is weak on steep slopes so that huge errors occured Affected areas were corrected using the triangular affine rectification

24 Rectification errors Errors of the rectification were taken into account Comparison of the GCP-location in the image to the DGPS measurement

25 Ortho-mosaic The resulting ortho-mosaic was combined using 59 suitable corrected photographs The maximum deviation reached 3.9 m, the mean error was quantified to 0.5 m The achieved resolution ranges between 3 cm and 8 cm

26 Morphometric analysis In May 2007 an ortho-photograph of the landslide was acquired by the Sintegra company The ground-resolution of this ortho-photograph is about 25 cm We compared our ortho-mosaic of 2008 to the ortho-photograph of 2007

27 Displacement analysis Displacement analysis was performed within a geographic information system (GIS)

28 Displacement analysis At the toe of the landslide displacements of 15 m are detectable

29 Displacements Displacements between 3 m and 55 m were detected from May 2007 to Oct. 2008

30 Displacements Displacements between 3 m and 55 m were detected from May 2007 to Oct Areas characterized by huge dynamics couldn t be measured (no visible features were left)

31 Analysis of fissures Fissures were compared between the mosaic of May 2007 and October 2008 In some areas fissures linger on the same location It turned out that this effect can be caused by burried crests

32 Conclusions Our low-cost remote sensing system provides high resolution information of the landslide The acquiered ortho-mosaic allows for morphometric and soil moisture analysis The projectiv image rectification is weak on steep slopes In further studies the stereographic photogrammetric analysis can be applied Areas characterized by huge dynamics couldn t be measured. This can be resolved by applying more frequent flight campaigns