The World in 3D

Imagine a GEOINT enterprise in which 2D images and GIS data are routinely merged with full 3D models, not just of Earth’s surface, but of roads, towers, power lines, trees, and even the exterior and interior of buildings. We are witnessing this transformation now as high-resolution 3D imaging technologies, robust data structures, and commodity applications enable the easy viewing and manipulation of 3D models.

Surface Models vs. Terrain Models

Some digital elevation products model the top of Earth surface features such as forest canopies and buildings, while others represent the so-called bare-Earth surface, in which trees and other features have been removed. Though there is no formal scientific agreement regarding terminology, it is generally accepted that Digital Elevation Model (DEM) is a generic term for any digital record of earth’s surface. DEM subsets include Digital Surface Models (DSM)—DEMs that model the top of surface features—and Digital Terrain Models (DTM), which represent the bare-earth surface with most features removed from the data.

The First Digital Elevation Product

More than 30 years ago, the military mapping organizations that preceded today’s National Geospatial-Intelligence Agency (NGA) developed Digital Terrain Elevation Data (DTED), the original digital product for terrain modeling. DTED is a discrete matrix of points in which elevations are assigned to a grid of posts set at regular intervals. The smaller the post interval is, the greater the DEM’s accuracy. For example, dropping a pole every football field length will likely produce less accurate data than with spacing equal to the length of a basketball court. The Shuttle Radar Topography Mission (SRTM), an 11-day space shuttle mission in February 2000, produced the first DTED on a near-global basis. SRTM used a technique called radar interferometry (INSAR), which employs a pair of antennas and makes precise radar phase measurements to produce elevation data. DTED 1 posts are approximately 90 meters apart, while DTED 2 posts are roughly 30 meters apart. More recently, the DTED specification has been extended to include smaller, much denser grids that can accommodate more high-resolution techniques.

Grids vs. TINs

DEMs can be represented as a raster grid, such as the method used for DTED, or by an irregular arrangement of connected triangles with X, Y, Z coordinates at each vertex. Unlike regimented grids, these Triangular Irregular Networks (TINs) vary in density as a function of terrain roughness. The triangles are small when the terrain surface varies and larger when the terrain is smooth. Since the point distribution of TINs is not limited to a fixed grid, they can better model height, such as the peak of a mountain or topographic breaklines and ridges. TINs also allow slope and aspect angles to be more easily derived.

Multi-Surface Point Clouds

The traditional formats for digital elevation data assign one elevation to a particular horizontal location. For this reason, a DEM of a mountainous forest canopy produces a surface model that looks like a sheet draped over the land and trees.

A point cloud is an unstructured format of raw X, Y, Z measurements of many surfaces of an object, where multiple elevations can be assigned to one location, and which includes a massive number of points. Point clouds are produced by LiDAR and other sensors, and are used routinely in Computer Automated Design (CAD) applications. They have become valuable for large-scale GEOINT applications to better represent the many levels of land surface structures.

Applications and Prospects

DEMs have numerous applications, including terrain visualization, mapping, landscape fly-throughs, line-of-sight calculations, and flood and hydrological modeling. However, the emergence of very high-resolution point clouds from LiDAR, radar, and optical sensors; the use of a wide range of platforms such as aircraft, spacecraft, large- and micro-UAVs, ground sensors and hand-held sensors; and commodity rendering software and hardware spearheaded by the gaming machine industry, enable new capabilities with staggering implications.

At the GEOINT 2013* Symposium in April, Airbus Defence and Space released WorldDEM, which is a global elevation model at 12-meter density, pole to pole. This commercial, open product is billed as far more dense and accurate than any other global DEM. Consider as well that it is now possible to purchase, for a few hundred dollars, a laser scanner that can be clipped to an iPad to create and manipulate 3D models of our local environment. It is even possible to add the fourth dimension of time by using LiDAR or video stereo to capture and reconstruct 3D motion with dynamic point clouds.

Elevation modeling now extends to so many GEOINT applications that they are difficult to enumerate. These include security preparation using 3D models of an event space, historic preservation that models the exterior and interior of buildings and cultural heritage sites, 3D crime scene reconstruction, modeling the interior of mines, and virtual tourism. For example, the Smithsonian uses 3D modeling to record the location and structure of bones and fossils at a paleontology site in the Atacama Desert.

In the past, the world was catalogued in 2D with photos, maps, and blueprints, and first attempts to create 3D earth models, while notable, were quite limited by today’s standards. GEOINT is becoming 3D, globally and locally, and not just for DEMs used for traditional applications, but for any application dealing with our world in 3D.

Featured image: This view of Antelope Valley, Calif., was generated by draping a Landsat satellite image over a preliminary topographic map from the Shuttle Radar Topography Mission. Courtesy of NASA JPL