Using eCognition to Detect Solar Potential

Geostellar is an online solar marketplace that aims to simplify the process of assessing and acquiring solar energy systems for residential properties. The company leverages geospatial technological advancements to provide a customized, real-time system that allows users to evaluate their property's solar potential, analyze financing options, and choose the most favorable solar option with just a few clicks. Geostellar's platform is designed to address the 'soft costs' associated with solar energy adoption, such as lead generation and customer acquisition, which can account for up to 50% of the total cost of solar installations. By automating the process of mapping building footprints and assessing solar potential, Geostellar aims to make solar energy more accessible and affordable for consumers across the United States. The company collaborates with various stakeholders, including local governments and utility companies, to provide comprehensive solar solutions that cater to the unique needs of different regions.

To achieve its mission to provide an on-demand, e-commerce system that could provide customized solar analytics, Geostellar first needed to have building vectors, the crucial data layer that would enable it to create precise maps and 3D models of all 3,143 counties in the US. Those foundational layers would then allow its system to map any rooftop’s solar potential in real-time. With millions of building footprints to map nationwide, a manual process of identifying and delineating buildings would not be feasible. Geostellar required an image analysis system that would almost completely automate the land classification process. And since each environment would present unique classification challenges, the solution needed the intelligence to quickly and accurately distinguish different structures from vegetative types and map out only buildings. In addition, although Geostellar would predominantly use LiDAR imagery for extracting buildings, not every county in the US has LiDAR data available. So the company also required a flexible system that could easily ingest and accurately classify imagery of variable quality.

With geographic targets in mind, Geostellar began scouring available data repositories for high-resolution spatial imagery such as satellite data and LiDAR point clouds, together with ancillary information such as local tax rates, utility rates, precipitation and temperature, and zoning regulations. To address the issue of automatically creating building vectors, Geostellar chose Trimble eCognition object-based image analysis software. 'With eCognition, I can quickly assess the quality of the datasets that I’m working with, tweak the rule sets accordingly, and the software does the rest,' said Dan Koopman, a spatial analyst with Geostellar. 'And it’s fast–depending on county size, it can take one minute to three hours on average to produce a building layer. That’s about 90 minutes of manual time for every one minute of eCognition time, which is significant time savings.' Although the process changes with the geography, once the available data is integrated, eCognition typically analyzes the information to first separate vegetation from impermeable surfaces. Then, based on height, it determines which vegetation is grass and which are trees, and identifies rooftops and roads. It also delineates building footprints and maps them. Those vector maps are then used by Geostellar’s proprietary solar simulation engine to create and provide on-demand rooftop assessments. Needing only a user’s address and average cost of their monthly electricity bill, Geostellar’s geomatics platform runs a 3D simulation to compute how much sun hits their roof annually. It then automatically layers in other data such as local utility rates, property values, and incentives programs and calculates the property owner’s financial prospects for transitioning to solar. In addition to the real-time assessment, it also provides a list of financing options, and vetted manufacturers and installers for consideration. Users then simply click on the most favorable offering.

• By resolving the key, automated-classification component of its solar e-commerce system, Geostellar was able to expand its reach into solar hotbeds across the US as well as offer its customers an efficient way to soften the hard costs of solarizing the country.
• Connecticut’s Clean Energy Finance & Investment Authority (CEFIA) is using Geostellar’s platform to efficiently connect the solar curious with solar contractors in Connecticut. Through CEFIA’s GoSolarCT web portal, more than one million Connecticut homeowners can obtain specific data regarding their home’s suitability for solar and compare energy and cash savings for solar leases and loans.
• Geostellar extended its solar reach to individual consumers. In December 2013 it launched a consumer-centric extension of its online solar marketplace called SolarMojo, a free mobile app that allows homeowners to compare solar installation and financing plans from industry leaders and purchase a system with the push of a button.
• By nearly automating the A-Z of 'going solar,' technological solutions like Geostellar’s are not only reducing manual labor to the physical installation of a system, they are helping to ultimately drive down the total cost of developing the solar energy market.

• Geostellar was able to map about 80% of the most valuable solar markets at one-meter resolution.
• The automated process saves about 90 minutes of manual time for every one minute of eCognition time.