DTM DEM Generation

DTM/DEM Generation

SUG541

Introduction → Map/GIS LayersSpatial information layers require a base comprising a DTM/DEM & imagery

The ortho-image layer: a critical element, which is draped over a DTM

Introduction → DEM & ortho-imagery

DEM provides height, referenced to position

Ortho-image provides metrically correct map for feature extraction

Ortho-image draped over a DTM

Modern photogrammetric process for geospatial information collection

DTM Generation: Contents

• Terminology• DTM Acquisition• DTM Editing• DTM Representation• Accuracy Issue• From DSM to DTM/DEM• DTM Manipulation

Terminology• DEM (Digital Elevation Model)

– Refers to regular array of elevations• DTM (Digital Terrain Model)

– More complex concept involving elevations and other GIS features– (e.g., rivers, ridges, break lines, etc.);– Encompasses terrain relief, planimetric, and derived data (slope,

aspect, visibility, etc.)• DHM (Digital Height Model)

– Similar as DEM, but less commonly used terminology• DGM (Digital Ground Model)• More emphasis on digital models of the solid/continuous surface

of the earth (used in the UK)

Terminology

• DTED (Digital Terrain Elevation Data)– Term used by the US Mapping Community; comes

from military specs.– Usually refers to gridded/regular arrays– DTEDs come in different ‘levels’

• DSM (Digital Surface Model)– Refers to digital model including features above

surface of the earth (e.g., trees, buildings)– Very important for orthophoto generation

DTM Acquisition

• Photogrammetric data capture (passive sensor)• Aerial photography• Digital satellite imagery

– Image matching used to automatically extract a dense– point cloud of 3D surface points from stereo image pairs and

potentially multi-image coverage; DTM derived via ‘regularization’ of the point cloud.

• RADAR: RAdio Detection And Ranging (active sensor)• LIDAR: LIght Detection And Ranging (active sensor)• Digitized contour maps• Ground surveying

DSM to DTM

DTM Editing

Modification and refinement of DTMs, and derivation of intermediate models.

• Editing: correcting errors and updating DTMs• Filtering: smoothing, enhancing, compression

and resampling• Merging and joining DTMs: – combining DTMs from several sources (possibly from

different dates)– Converting DTMs from one data structure to another

DEM Representation• Raster DEM– Elevations are available at equally spaced grid points

• TIN (Triangulated Irregular Network)– Elevation data at irregular points that are formed into triangles– The TIN is generated in such a way that the summation of the

lengths of the triangle legs is a minimum

DelaunayTriangulation

Triangulated Irregular Network (TIN)

Raster DEM & TINs

RASTER: Effect of Grid Size on Surface Representation• Sampling interval will affect

– Amount of detail captured (accuracy)– Amount of storage (redundancy, efficiency)

• Optimum sampling interval depends on the nature of the terrain• Raster DEM & TINsTINS:• Each vertex must have the following information

– Height– Connectivity information– Surface normal

• The triangle legs can be forced to coincide with the break lines

Raster vs. TINs

Shaded DSM

Source image (1 of stereo pair) Shaded DSM

Accuracy of Photogrammetric DSM from IKONOS

DSM to DTMMorphological operations

Morphological operations: removal of above-ground features

DSM

DTM

DSM to DTMMultiple height bin method for object detection & removal

Multiple height in method for object detection

image DSM Detected above-ground objects

DTM Manipulation

• Nearest neighbour assignment• Linear interpolation• Bilinear interpolation• Cubic convolution

Nearest Neighbour Assignment

• Assigns the values of the nearest grid point to the output grid cell

• No actual interpolation is performed based on values of neighbouring points

• Does not create a continuous surface

Bilinear Interpolation• Determines the height of a point based on a

weighted average of the 4 grid points• Generated surface is continuous, but not smooth

Bilinear Interpolation

Cubic Convolution

• Determine the Z value of a point using 16 neighbouring grid points

• Convolve with a smooth curve function