Deploying the Lowest Earth Orbit Satellite to Edit OpenStreetMap
One of the challenges we’ve seen with OpenStreetMaps over the year is access to current remote imagery for creating and editing vectors. iD and JOSM are marvelous tools, but fueling them with up to date imagery can, at times, be challenging. For this post we’ll break down a few of the challenges with access to satellite/aerial imagery. Then propose a complimentary alternative based on work the team at Pixel8 has been doing where we generate a NADIR GeoTIFF from just commodity video.
The Challenge of Acquiring Overhead Pixels
Launching satellites into space and flying airplanes with expensive megapixel cameras is expensive. There is no way around it. This creates two challenges 1) licensing and 2) frequency of updates. When you have significant overhead costs like rocket launches and airplanes it is hard to give your imagery away for free. This means a cadre of lawyers draws up licensing terms for your imagery to protect the business. Lawyers are exceptionally good at this task. The downside is it can often make doing creative deals — like donating imagery to OSM difficult. There is probably a whole series of posts to be had about the legal challenges, but we’ll leave that to more qualified writers. Suffice it to say getting anything through legal that revolves around licensing is a herculean task that should be appreciated and not under estimated. Thank you Maxar, Mapbox, Microsoft and ESRI!
The second challenge with acquiring overhead pixels for OSM editing is how frequently they are updated. Each aerial flight is expensive and updates tend to lag. Combine this with imagery companies wanting to save their freshest imagery for paying customers, and we get a time lag for the pixels that come to OSM. On the other hand we have satellites that are collecting all the time, so should make freshness way better. The problem is satellite firms most often provide mosaics for mapping. Mosaics use a variety of satellite images from different days to stitch together a cloud free map of a large geographic area. Typically mosaics are updated a couple of times of years. Both aerial and satellite imagery are amazing resources that are invaluable to OSM, but are there opportunities for ad hoc imagery creation by the community.
Less Remote Pixels
As we’ve been working on creating 3D point clouds from commodity video at Pixel8 we started to think about how the data might help OSM. Using just a 360 camera, like a GoPro Fusion or Max, we’ve been able to generate point clouds with 50cm absolute accuracy. The problem was iD and JOSM don’t support 3D objects like point clouds. PDAL to the rescue! Ends up there is a super cool PDAL pipeline that will convert your point cloud to a NADIR GeoTIFF.
Now that we have a GeoTIFF we can post it to a cloud service and drop the URL into the excellent iD editor. First, let’s see the GeoTIFF without vectors and get a sense of the detail in the image.
Next let’s see how well the imagery lines up with existing vectors in OSM. Boulder is very well mapped so we should get plenty of edits to compare our image alignment with.
The alignment of results exceeded our expectations and it is really encouraging to see that a simple GoPro video can give us a solid image to edit OSM with. This new approach certainly doesn’t obviate the need for satellite or aerial imagery, but it is a nice compliment when traditional imagery falls out of date. The added bonus is the imagery will be under an open data license and you can generate it yourself with minimal effort. If you are interested in being updated on the Pixel8 project and joining the community drop us an email address here.
