HAL Id: hal-01496754
https://hal.archives-ouvertes.fr/hal-01496754
Submitted on 30 Mar 2017
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Farmland terrace slope detection from Pleiades digital elevation models
Jean-Stéphane Bailly, Giulia Sofia, , Paolo Tarolli, Florent Levavasseur
To cite this version:
Jean-Stéphane Bailly, Giulia Sofia, , Paolo Tarolli, Florent Levavasseur. Farmland terrace slope de-tection from Pleiades digital elevation models. EGU General Assembly Conference Abstracts, 2015, Vienne, Austria. pp.10021. �hal-01496754�
Farmland terrace slope detection
from Pleiades digital elevation models
Jean-Stéphane Bailly, Giulia So a, Nesrine Chehata, Paolo Tarolli, Florent Levavasseur
[email protected], giulia.so [email protected], [email protected], [email protected], [email protected]
Introduction
Mediterranean farmers have adopted from Roman time soil conservation practices. By decreasing slope intensi-ties and lengths, terraces are one these soil conservation practices. The spatial pattern of terrace slopes is hetero-geneous and depends on natural terrain slope. In erosion and hydrological studies, maps of terraces slopes may be needed [1]:
• How to map these landscape elements from re-mote sensing ?
• Is a Pleiades derivated DEM suitable and accurate enough to map these landscape elements ?
Materials and Methods
Reference terrace data, Pleiades images and a LIDAR 1 m DSM-DTM were acquired on the Peyne Catchment (Languedoc-Roussillon). A DSM was build from a stereo pairs of Pleiades images (January 2013, 29 degrees in-cidence) using the Mic-Mac IGN suite with low regu-larization and resampled at 1 m resolution. DTM was generated from DSM after closing operations.
−4 −2 0 2 4 6 8 42 44 46 48 50 0 1000 2000 3000 4000 ● study area ● Paris ● Montpellier ● Toulouse 726000 726500 727000 727500 728000 6265000 6265500 6266000 50 60 70 80 90 100 110 120 Terrace wall height
0m< <1m 1m< <2m 2m< <3m 3m< <4m 4m< <5m 5m< <6m > 6 m
Terrace slopes consisting in an abrupt linear change in elevation on plot lattice.
725800 725850 725900 725950 6264900 6264950 6265000 6265050 6265100 6265150 62 65200 Original DEM 95 100 105 110 115 725800 725850 725900 725950 6264900 6264950 6265000 6265050 6265100 6265150 62 65200
Filter + Gaussian contrast
−2 −1 0 1 2 725800 725850 725900 725950 6264900 6264950 6265000 6265050 6265100 6265150 62 65200 Gradient 10 20 30 40 50 60 725800 725850 725900 725950 6264900 6264950 6265000 6265050 6265100 6265150 62 65200 Aligned segments (LSD)
Figure 1: Terrace walls detection process
The DEM process relies on:
1. a fast line segment detector algorithm to detect aligned abrupt change from gradient image [2]. 2. or a curvature thresholding technique to detect
terrace slopes [3] followed by a vectorisation pro-cess.
3. a post-processing ltering step using plot lattice distances.
LSD: looking for aligned regions with equal gradient angle c
http://ubee.enseeiht.fr/vision/ELSD/
References
[1] Francesc Gallart, Pilar Llorens, and Jerome Latron. Studying the role of old agricultural terraces on runo generation in a small mediterranean mountainous basin. Journal of Hydrology, 159(1-4):291 303, 1994.
[2] R Grompone Von Gioi, Jeremie Jakubowicz, Jean-Michel Morel, and Gregory Randall. Lsd: a line segment detector. Image Pro-cessing On Line, 2012.
[3] Giulia So a, Giancarlo Dalla Fontana, and Paolo Tarolli. High-resolution topography and anthropogenic feature extraction: test-ing geomorphometric parameters in oodplains. Hydrological Pro-cesses, 28(4):20462061, 2014.
[4] Paolo Tarolli, Federico Preti, and Nunzio Romano. Terraced land-scape: from an old best practice to a rising land abandoned-related soil erosion risk. In EGU General Assembly Conference Abstracts, volume 15, page 3355, 2013.
Results: Pleaides DEM vs Lidar DEM
726000 726500 727000 727500 728000 6264800 6265000 6265200 6265400 6265600 6265800 6266000 6266200 60 80 100 120 726000 726500 727000 727500 728000 6264800 6265000 6265200 6265400 6265600 6265800 6266000 6266200 60 80 100 120 726000 726200 726400 726600 6265500 6265600 6265700 6265800 6265900 80 90 100 110 120 725700 725800 725900 726000 726100 726200 6264900 6265000 6265100 6265200 95 100 105 110 115 120 726000 726200 726400 726600 6265500 6265600 6265700 6265800 6265900 70 80 90 100 110 120 725700 725800 725900 726000 726100 726200 6264900 6265000 6265100 6265200 90 95 100 105 110 115 120 0 100 200 300 400 500 100 105 110 115 120 ele v ation (m) Lidar Pleiades Terrace wall location
Figure 2: Lidar (06-2001) & Pleiades (01-2013)
726000 726500 727000 727500 728000 6264800 6265000 6265200 6265400 6265600 6265800 6266000 6266200 −20 −10 0 10 20 726000 726500 727000 727500 728000 6264800 6265000 6265200 6265400 6265600 6265800 6266000 6266200 −10 −5 0 5 725700 725800 725900 726000 726100 6264900 6264950 6265000 6265050 6265100 6265150 6265200 −15 −10 −5 0 725700 725800 725900 726000 726100 6264900 6264950 6265000 6265050 6265100 6265150 6265200 −6 −4 −2 0 2
to LiDAR Conf. Conf. Median DEM 95% (m) 68% (m) dev. (m) Pleiades [-2.91,1.54] [-0.45,0.58] 0.35
Aerial [-4.76,1.37] [-0.38,0.68] 0.46
Table 1: Deviation statistics on elevation
Results: Terrace walls extraction from DEMs
Method DEM TP TN FP FP QI Type 1 2 (1) LSD dtm 0.95 0.05 0.20 0.33 0.79 LSD dsm 0.75 0.25 0.40 0.61 0.54 Curv dtm 0.98 0.02 0.36 0.68 0.72 Curv dsm 0.76 0.24 0.31 0.70 0.58
Table 2: LiDAR performances
Method DEM TP TN FP FP QI Type 1 2 (1) LSD dtm 0.62 0.38 0.23 0.68 0.50 LSD dsm 0.44 0.56 0.20 0.58 0.37 Curv dtm 0.65 0.35 0.26 0.79 0.52 Curv dsm 0.50 0.50 0.22 0.87 0.41
Table 3: Pleiades performances
726000 726500 727000 727500 728000 6265000 6265500 6266000 726000 726500 727000 727500 728000 6265000 6265500 6266000 TP FP TN FP (before post−t) 726000 726500 727000 727500 728000 6265000 6265500 6266000 726000 726500 727000 727500 728000 6265000 6265500 6266000 TP FP TN FP (before post−t)
Figure 3: Maps of the detected terrace slope of LiDAR DTM using the LSD method (left) and from
the Pleiades DSM when using the geomorphometric method (right). Resulting T P elements in black, T N
elements in red, ltered F P elements in blue and non ltered F P elements in light blue (down)
Conclusion
Detection of terrace slopes was performed using two meth-ods based on the linearity of features (LSD) or their con-vexity (geomorphometric). Four major results were ob-tained from this study. The automatic mapping of terrace slopes from Pleiades DEMs is reliable for terrace slopes that are higher than 2 m and. The LiDAR detection performances are systematically higher than the Pleiades performances, especially when using a LiDAR DTM. The DSM to DTM ltering processes used for the Pleiades data do not result in better terrace slope detection. The ge-omorphometric algorithm is more robust than the LSD algorithm when applied to noisy digital elevation models.
● ● ● ● ● ● ● Height (m) TP ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0< <=1 1< <=2 2< <=3 3< <=4 4< <=5 5< <=6 6< <=7 0.2 0.4 0.6 0.8 1.0 Pleiades−DSM−lsd Pleiades−DSM−Curv Pleiades−DTM−lsd Pleiades−DTM−Curv LiDAR−DSM−lsd LiDAR−DSM−Curv LiDAR−DTM−lsd LiDAR−DTM−Curv
Figure 4: Detection rate as a function of terrace
height
Acknowledgements
The authors would like to thank the French Space Agency CNES and Airbus Defense & Space for providing the Pleiades Images within the RCT Pleiades program. In addition, the authors would like to thank IGN-Espace, especially Mr. Nicolas Champion, for helping generate Pleiades DEMs from image pairs.
