Updating satellite imagery using telemetry data

Telemetry data coming from our SDK allows our satellite team to collect new imagery, faster and smarter.

The 3 million sq km of high resolution imagery that we just bought is almost fully processed and we’re gearing up for our next round of imagery acquisition. In classic Gundersen fashion, my meeting with Eric last week started off with, “We’re going shopping. What are the priority refresh areas you see for Q2? How much can the team update and process by the end of the year?” My response, “Let’s shop where our users need it most.”

The satellite team is now working with our telemetry team to analyze anonymized, aggregated latitude and longitude points, looking for user activity where our imagery is old or has clouds. This means that as people look at the map, we check to make sure it’s the best. And when we can make it better, we flag the area as a priority collect. This creates a system where developers using the map SDK will get the most updated imagery specifically where their users need it.

Anonymized telemetry data shows us where we may have out of date or cloudy imagery. Points on known streets are filtered out of this view, so untraced ways pop. With updated imagery, these point collections reveal streets and ways that were previously not visible.

This might feel like the opposite of what other satellite teams do. We don’t use satellites to count the cars parked in front of Walmart to glean consumer behavior before earnings reports. We see the activity in the parking lot, or the highway, or the walking trail and synthesize data of where people are using our SDK to purchase priority imagery so developers get better and fresher imagery for their apps.

All of this depends on super high resolution imagery. DigitalGlobe’s 30 cm imagery from WorldView-3 has been streaming live for over year now. Our team is now anxiously awaiting the additional capacity coming with the launch of WorldView-4, currently scheduled for September out of Vandenberg.

With knowledge from our anonymized telemetry data, we are able to replace old imagery with new, immediately trace any missing features, and improve our satellite imagery and streets layers simultaneously.

This not only means better imagery coverage, it also means better streets coverage. When we find a new road from our telemetry data, we trace the imagery into OpenStreetMap. The process builds on itself. Better imagery makes better maps. Better maps help more users, and more users using the SDK generates more telemetry data. And more data streaming from our SDK makes the map better and shows us where we need to buy fresh imagery. Repeat.