The complex and in-depth process has been undertaken by data company EarthDefine, located in Redmond. The fully detailed national building footprint dataset carefully maps out over 137 million buildings, and is operated by deep neural networks. The accuracy has been assessed at 98 percent.
The output is the National Building Map. As well as detailed outlines of homes and premises, the map includes accurate locations and addresses. Such data can be used for initiatives like routing and rooftop-accurate geocoding for millions of properties across the U.S. The level of detail is such that building aspects like roof form, building elevation, eave height and so on can be extracted from the data and visualized.
The map is also interactive and it can be updated very quickly; by deploying AI, the software can continuously extract building footprints from newer imagery, adding any new facilities or design changes, and also remove older buildings. This leads to a regularly updated system containing a close to real-time snapshot of the built environment from across the U.S.
The data will also be used by planners, plus improving safety and for more informed decision making processes like insurance risk assessment and responses by emergency services.
Speaking with Smart2Zero, EarthDefine CEO, Vikalpa Jetly says: “Our goal is to create and maintain state/national level building data with a unique and consistent attribute set…We are achieving this best-in-class building footprint dataset through applying advancements in artificial intelligence to high resolution aerial imagery.”
In related news, the geospatial analytics startup company Allvision has raised $3.2 million in Seed funding. The money will be used to push ahead with the development of high-definition digital tools to be used for mapping, mobility and analytics applications. IBM is also experimenting with extracting value from geospatial-temporal big data. One application of this is with developing a precision irrigation system that utilizes a cloud-based communication network, hundreds of sensors and actuators and satellite imagery.
