Home
Resources
Case Studies
Pléiades
Reveal uses OneAtlas Imagery t...
U.S. Census Bureau: Automated Detection and Classification of Nationwide Residential Construction
Machine Learning algorithms from Reveal analyse high-resolution Pléiades imagery from OneAtlas to count residential construction projects and assess their current stage across the US. The information derived from satellite imagery is analysed by Reveal and provided to the U.S. Census Bureau to help monitor areas under construction.
Challenge
The U.S. Census Bureau tracks new home construction activity in approximately 20,000 jurisdictions and unincorporated areas every month. This includes all types of residential structures - single family homes, townhomes, and condominium buildings. The Census Bureau tracks these projects in three stages: New Starts, In Progress, and Completed. These status reports are used to calculate monthly investments in residential construction, which is one of the key economic indicators measuring the health of the US economy.
The Census Bureau has traditionally gathered construction data by manually monitoring building permits and conducting surveys. Unfortunately, this information is often out of date, incomplete, or inconsistent because construction permitting and reporting processes vary from one jurisdiction to the next.
Our solution
Reveal has developed an automated process that uses machine learning to analyze 50cm Pléiades optical imagery to identify when land has been cleared for residential construction, determine the type of structure being built (home, townhouse, etc.), assess monthly progress, and verify completion.
Because building types vary greatly across the country, Reveal leveraged the extensive Pléiades archive from OneAtlas to train the algorithms to recognize the many building features and differentiate them from other construction projects. The algorithms learn over time to improve the accuracy of classification of construction stages, so the Census Bureau can better calculate economic value.
With approximately 90% accuracy, Reveal’s advanced process automation (APA) solution is identifying residential construction and classifying each phase of activity from groundbreaking through to completion.
Residential construction progress reports are more accurate and reliable across the nation because building sites are monitored directly by satellites that acquire imagery that can be analyzed by machine learning in an automated process that is consistent from month-to month and region-to-region.
Organisation involved
Reveal Global Consulting, LLC is a data analytics, visualization, and modeling firm located in Fulton, Maryland. Learn more at http://www.revealgc.com/
The U.S. Census Bureau Economic Indicators Division, Construction Programs, tracks building activity every month as a key indicator of U.S. economic health.
Want to know more?
Our sales team will be happy to provide you more information about this case study and how it can meet your business needs.
See all our case studies
Oil palm trees health in Malaysia
MYSA, the Malaysian Space Agency, is developing indicators to monitor the health of plants to support agriculture and forestry. A major priority is the health of oil palm trees, essential to Malaysia’s position as a leading palm oil producer.
FrieslandCampina trusts Starling for their sustainable palm oil sourcing
The production of commodities such as palm oil, cocoa, soy, and pulp and paper is responsible for over 70 percent of tropical deforestation . To combat this, it is important for agro food industries to have detailed knowledge of their supply chains to address deforestation development risks effectively.