Automated stereo-photogrammetric DEM generation at high latitudes: Surface Extraction with TIN-based Search-space Minimization (SETSM) validation and demonstration over glaciated regions

Figures & data

Figure 1. Schematic of the SETSM algorithm. The nth level of the image pyramid is automatically determined based on the ground sample distance (GSD) of the image.

Table 1. WorldView-1, 2 and Pleiades satellite specifications.

Satellite	WorldView-1	WorldView-2	Pleiades-1A
Launched data	September 2007	October 2009	December 2011
Panchromatic bands	400–900 nm	450–800 nm	480–830 nm
Ground sample distance (GSD)	0.50 m at nadir	0.46 m at nadir	0.70 m at nadir
	0.55 m at 20° off-nadir	0.52 m at 20° off-nadir	0.50 m (delivered products)
Dynamic range	11 bits per pixel	11 bits per pixel	12 bits per pixel
Swath width	17.7 km at nadir	16.4 km at nadir	20.0 km at nadir
Max contiguous area in stereo scenario	51 × 112 km with 3 pairs	63 × 112 km with 4 pairs	20 km × 280 km at stereo mode
	(30° off-nadir angle)	(30° off-nadir angle)	20 km × 120 km at tri-stereo mode (B/H = 0.6)
Revisit frequency (at 40°N Latitude for WorldView)	1.7 days at 1 m GSD or less	1.1 days at 1 m GSD or less	1 day at 43° off-nadir (at Equator)
	5.4 days at 20° off-nadir or less (0.55 m GSD)	3.7 days at 20° off-nadir or less (0.52 m GSD)
Geolocation accuracy	<4.0 m CE90 without ground control	<3.5 m CE90 without ground control	3.0 m ~ 10.0 m CE90 without ground control

Figure 2. Footprint of WorldView images and ILATM1 B data-set on west Greenland. Red, green, blue and cyan colors indicate 2010, 2011, 2012 and 2013 year, respectively. (a) Scene center coordinates of stereo pairs and strips connecting each other by lines. (b) Trajectories of LiDAR elevation data. The map axes are in polar stereographic projection with units of meters. For full colour reproduction, please see the online version of this paper.

Figure 3. Control points for calculating the translational offset. (1) and (2) are seventh stereo pair along WV01_20110605 strip and third stereo pair along WV01_20120422 strip, respectively. Parts (a) and (b) show the ILATM1 B data over SETSM DEM and control points over orthorectified image, respectively.

Table 2. SETSM DEM accuracy with ILATM1 B data-set through the co-registration algorithm. The name of Strip ID is composed of satellite type and date of imagery.

Strip ID	Co-registration translation parameters				Surface distance of control points, dD [m]
	Tx [m]	Ty [m]	Tz [m]	CE [m]	Before	After
WV01_20110605	3.272	2.482	1.275	4.107	1.083	0.117
WV01_20110903	2.945	3.916	0.924	4.900	2.301	0.194
WV01_20110910	3.596	3.777	1.590	5.215	1.095	0.403
WV01_20110920	1.607	3.018	2.405	3.419	2.805	0.331
WV01_20120210	1.401	3.176	2.741	3.471	0.718	0.146
WV01_20120422	2.193	1.597	2.634	2.713	0.675	0.105
WV01_20120531	2.392	0.347	0.935	2.417	0.730	0.221
WV01_20120708	1.237	1.719	3.148	2.118	1.105	0.167
WV01_20120716	1.239	0.830	1.593	1.492	1.843	0.142
WV02_20100710	1.989	3.355	0.408	3.900	0.421	0.089
WV02_20100805	2.800	3.671	2.164	4.617	0.497	0.223
WV02_20100819	2.886	3.374	1.588	4.440	1.488	0.083
WV02_20120724	3.090	3.207	2.594	4.454	1.612	0.133
RMSE	2.483	2.878	2.013	3.801	1.438	0.203
MIN	1.237	0.347	0.408	1.492	0.421	0.083
MAX	3.596	3.916	3.148	5.215	2.805	0.403

Figure 4. Outlet glaciers of (1), (2) and (3) are located in 66.94° N and 49.92° W, 67.07° N and 50.20° W, and 70.63° N and 50.65° W, respectively. The stereo pairs have been taken on 5 June 2011 by WorldView-1, 19 August 2010 by WorldView-2 and 22 April 2012 by WorldView-1, respectively. Part (a) shows the hill-shaded images of DEMs. Parts (b) and (c) show the cross section profiles (A) and (B), respectively. A polar stereographic projection is used with units of meters.

Figure 5. (1) and (2) indicate the Surpraglacial lake located in 67.95° N and 48.72° W and the rivers located in 69.24° N and 150.49° W, respectively. The stereo pairs have been taken on 28 November 2011 and 15 June 2012 by WorldView-1, respectively. Part (a) shows the hill-shaded images of DEMs. Parts (b), (c) and (d) show cross section profiles (A), (B) and (C), respectively. A polar stereographic projection is used with units of meters.

Figure 6. Mountainous region DEMs of Alaska are located in 68.05° N and 149.69° W. Stereo pairs have been taken on 10 September 2008 by WorldView-1. Arrows indicate shadowed regions. A polar stereographic projection is used with units of meters.

Figure 7. East of ice cap DEMs in Baffin Island, Canada (70.05° N and 73.39° W).

Figure 8. Dry valley DEMs of (1) and (2) are located in 77.74° S and 162.10° E, and 77.69° S and 160.58° E at Antarctica, respectively. Stereo pairs have been taken on 1 November 2009 by WorldView-1. Part (a) shows the hill-shaded images of DEMs. Parts (b) and (c) show the cross section profiles (A) and (B), respectively. A polar stereographic projection is used with units of meters.

Figure 9. Three-dimensional views with SETSM DEMs and orthoimages.

Figure 10. SETSM DEM of Astrolabe Glacier, Antarctica (66.77° S and 139.78° W) from Pleiades imagery. Stereo pair has been taken on 2 February 2013. Parts (c), (d) and (e) show the hill-shaded images of patches (A), (B) and (C), respectively. A polar stereographic projection is used with units of meters.

Figure 11. Surface elevation change with SETSM DEMs over large calving glacier. Parts (a) and (b) show the orthorectified images taken on 19 March and 3 April 2013 by WorldView-2, respectively. Dots on (a) are the control points for co-registration. Part (c) is the height difference map calculated as DEM(20130403) – DEM(20130319) over all area. Parts (d), (e) and (f) show the enlarged DEM1(20130319), DEM2(20130403) and height difference map over area (A) calculated as DEM2 – DEM1, respectively. A polar stereographic projection is used with units of meters.

Figure 12. Surface elevation change with SETSM DEMs over an area of land-terminating margin. Parts (a) and (b) show the orthorectified images taken on 5 June 2011 and 21 July 2012 by WorldView-1, respectively. Dots on (a) are the control points for co-registration. Part (c) is the height difference map calculated as DEM(20120721) – DEM(20110605) over all area. Large errors labeled B1, B2 and B3 are caused by frozen lake surfaces in the first DEM that were unfrozen in the second. Parts (d), (e) and (f) show the enlarged hill-shaded image of 20110605 and 20120721, and height difference map over area (A) calculated as DEM2 – DEM1, respectively. A polar stereographic projection is used with units of meters.