I’m presenting on our imagery processing pipeline and how we deal with the enormous amount of imagery that we see every day, from satellites to drones, at the Bay Area GeoMeetup, this Thursday the 11th at 6:30 at the Mapbox office. Our pipeline is a set of customizable and modular image processing components that handle everything from data calibration to declouding, mosaicking, and color adjustment. At a scale of tens of trillions of pixels, it all has to work smoothly and efficiently. The pipeline is what lets us hook up to any imaginable imagery source—ultra high-res commercial data, microsat frames, weather satellite streams, aerial and drone surveys, you name it—and get consistent, beautiful maps.
A very simple pipeline for Landsat 8: low-resolution color data (left, including the Golden Gate Bridge) is combined with high-resolution panchromatic data (center, including Alcatraz) to prepare a true color image (right, including SOMA and Mapbox SF, where we’ll see you on Thursday).
I’ll talk about the challenges we’ve found while building and refining this system, and how we worked through them. I’ll also mention some details of basemap philosophy, for example how we prioritize image recency versus image quality. And I’ll point out some resources like free imagery and open-source tools that anyone can use to get started working with satellite data. All illustrated with lots of pretty pictures, of course.