From Imagery to Map: Digital Photogrammetric Technologies 13 th International Scientific and Technical Conference From Imagery to Map: Digital Photogrammetric Technologies The new methods of dense DTM creation in PHOTOMOD Andrey Sechin Scientific Director, Racurs September 2013, Fontainebleau, France
DEM, DTM, DSM, nDSM DEM, DTM have different definitions in different countries. In Russian (по-русски) ЦМР, ЦМП DEM & DTM - bare earth terrain. DSM include tree canopy & buildings. nDSM = DSM - DTM
PHOTOMOD. Different models (depending on algorithm) Automatic DSM (old cross-correlation algorithm) with filtering buildings and trees Automatic DSM (old cross-correlation algorithm) 3D semi-automatic model (with manual stereo vectorization) Different models for Novokuznetsk city GeoEye-1 stereopair (GSD 0.5m)
Local algorithms of DTM creation Local algorithms Memory efficient Fast Subpixel accuracy in “smooth” regions Problems with periodic structures and poorly textured regions Big problem with discontinuities on images
Global algorithms of DTM creation Global algorithms Global energy minimization Take into account discontinuities and hidden surfaces Not memory efficient Still require filtering and smoothing in the end of algorithm E = E(data) + E(smooth) Semi Global Matching (SGM) Graph-cuts Simple Tree Iterative- deformation method (RACURS)
PHOTOMOD: iterative deformation method (IDM) All images are taken into account simultaneously Memory efficient Image pyramid hierarchy is used for speed and reliability Image resection is used to calculate occlusions Still requires filtering and smoothing on the final step
Local vs Global method SGM and Iterative deformation methods CrossCorrelation
SGM example Stereopair and disparity map
IDM: GeoEye example Cross Correlation IDM
IDM: Pleiades example
IDM: DMC aerial example GSD = 6 cm
IDM: DMC aerial example GSD = 6 cm
IDM UltraCam example GSD = 7 cm RMS difference between LIDAR and DSM – 21 cm
IDM UltraCam example
Thank you for attention