NGA plans to turn 2D images into 3D

The agency is looking for ways to create automated 3D point clouds from satellite images.

The National Geospatial-Intelligence Agency is looking for research on existing capabilities for automatically deriving georeferenced three-dimensional point clouds from commercial satellite imagery, according to a Request for Information released today.

Point clouds, which comprise sets of data points in a coordinate system, can be used in geographic information systems to make digital elevation models of terrain or generate 3D models of urban areas. The data points in a point cloud typically represent the X, Y, and Z geometric coordinates of the scanned objects or areas.

These data points can be collected in several ways. Light Detection and Ranging (LIDAR) and other laser scanners — including the Xbox Kinect — can be used to map out a 3D point cloud. The current solicitation is looking to get that data from multiple commercial satellite images.

At the minimum, NGA is looking for a fully automated system that would be able to ingest at least two commercial satellite images and output a true 3D point cloud — capable of showing multiple elevations at a single horizontal location —  in a standard format. Each point within the cloud would be required to have a spatial location and pixel intensity.

The agency is hoping to use the technology to create spatially accurate, high-fidelity, dense 3D point clouds, according to the announcement. This includes accommodating different types of terrain such as urban and rural areas, and adjusting to fluctuating solar illumination.

NGA provides the military and other agencies with geospatial intelligence, and is developing a comprehensive Map of the World containing all of its geo-intelligence, multisource content, analysis and reporting. 

The system would preferably be able to ingest up to hundreds of overlapping images from differing angles and process up to several hundred square kilometers of Earth. Another desirable capability would include the production of uncertainty estimates for each point using the Universal LIDAR Error Model.

Information obtained from the solicitation will be used to decide future funding opportunities and acquisition decisions. NGA is interested in working with working with any organization on the North American continent.

The final response date for the RFI is Aug. 28.