Planet's "best-pixel mosaicking" process

One of our main goals at Planet Labs is to make our imagery easily accessible. This means providing not just the raw images from our satellites, but creating ‘mosaics’ — large overviews of area from many satellites. Building imagery mosaics with the traditional approach involves a lot of human input, and they can take months to produce. At Planet, we knew we needed a different approach to take advantage of the frequency of imagery we’ll be collecting. So we’ve developed a fully automated mosaicking process, using a technique of ‘compositing’ the imagery from our archive together pixel-by-pixel. We call this “best-pixel mosaicking.”

Our mosaics are also a bit different than what you might see on popular web mapping sites or in map APIs. Looking at a map online an image could be two months old, but it also might be two years old. There’s no way to know though, as consumer mapping providers don’t provide any information on when the image was taken. Our mosaics will be ‘traceable’ — we track where every pixel came from — and we’ll soon enable functionality to enable you to go back to the original scene and know exactly when it was taken.

We recently released four mosaics for California, all of them easily searchable, with information about the scenes that we used to create the mosaics.

Soon we will also create mosaic ‘slices’, where you can see all the data that has been gathered over a certain time frame — a season, a month, a week or a day.

We are excited to bring the notion of timely and traceable mosaics to the world, and will continue to improve upon them based on feedback from our users and customers. If you are interested in the compositing techniques we’re using, we’ve even open-sourced our tools for building seamless and cloudless image mosaics from deep stacks of satellite imagery. We welcome pull requests!