Aerial images are rich with data just waiting to be interpreted and converted into knowledge, perhaps even actionable knowledge. Given the unimaginable amount of imagery being collected by satellites and other aerial devices each day, even relying on automated pattern deciphering by computers doesn’t provide the capacity to keep up with both the supply and demand. As smart as software algorithms have been designed to be, there are still times when the best combination for speed and accuracy in image interpretation is a pair of human eyes — or, thousands of pairs — which is one reason why we have seen an exciting growth in applications for crowdsourced remote sensing, one dimension of the citizen science trend.
Citizen science today sits at a sweet spot with benefits from several converging factors: an abundant usage of hand-held devices capable of returning geographic location information, a rich supply of software developers looking to apply their skills to engaging projects on mobile devices, constantly expanding Wi-Fi access, faster and more robust bandwidth for sharing images and a ludicrous amount of data being constantly collected. It’s like a dog that’s caught a flailing fire hose of data and can barely hold it in its mouth, much less gulp and swallow the stream. Processing even a fraction of the images being gathered would be impossible without the volunteer contributions of people around the world.
In this article, we’ll consider programs that require the visual cognition skills that people bring to the table, ones that humans can do better than machines, at least at this point. This can mean looking at an aerial image and tracing what you see, comparing two images or reading a pattern in an image. Practice with these now and you’ll be all set for when winter arrives. There’s nothing cozier than gathering together with the family around the coffee table while each person’s laptop keeps them warm!
Tracing and drawing
Though it’s only been around for about a decade, OpenStreetMap is a grandfather in the world of crowd-produced geospatial data. Its premise is simple: provide raster images of the world over which people trace shapes and create vector data, then add descriptive information; then people can download the vector data. Having a computer pick out a square building top or a road from the rest of a scene might not seem so difficult a task, but knowing that the structure is a school, or that the road is dirt and cannot support significant weight, is the human-provided value add.
Moreover, for most of the world, having current and freely available geospatial data files of buildings and roads is still uncommon, yet disasters are often likely to strike in locales that have not been well mapped, often because they are remote and/or impoverished. Haiti’s 2010 earthquake was the inspiration that brought the potential for OSM to the world’s attention and inspired the launch of the Humanitarian OpenStreetMap Team, a widely popular application of OSM.
The maturing of OpenStreetMap has meant they’ve been able to figure out how to best support the efforts of their volunteer contributors and increase the likelihood that the energy exerted will produce good results. They’ve produced training materials such as LearnOSM that explain the basics of heads-up digitizing. The global community has developed affiliated programs to support and facilitate the editing process, such as Java OpenStreetMap, and they’re organized enough to have a Tasking Manager in place.
The U.S. government has followed their lead and recognized the power of local expertise to inform and enhance their geospatial data sets. The U.S. Geological Survey, primary stewards of The National Map, has established The National Map Corps to coordinate volunteer data contributions, particularly for updating information on structures. According to Elizabeth McCartney, lead coordinator of the project, volunteers are welcome to edit anywhere in the U.S., Puerto Rico and the U.S. Virgin Islands, but periodic Map Challenges are sponsored to target efforts to areas needing special attention. Recent Map Challenges have focused on prisons (think federal and state penitentiaries) and law enforcement (think sheriff's offices, local police departments, highway patrol) in Kentucky and Tennessee, so that those data sets will be improved prior to the next print run of topographical maps scheduled for those states in 2016. They expect to emphasize post offices and schools in the future, but McCartney notes that all structural features would benefit from attention.
Categorizing and classifying
Using imagery as a visual background for creating new data sets is a fairly high level of active participation on the part of volunteers. Another useful cognitive task we can offer the community is our ability to categorize and classify information based on the patterns we see in an image. In each case, some training is provided to guide the viewer through the image interpretation process. Such efforts could result in identifying the movement patterns of wildebeests, noting evidence of fracking or types of underwater plankton from countless underwater images — who knew there were so many types?! If astronomical space is more your thing, there are galaxies and dust-surrounded stars just waiting to be found!
Making an interpretation decision can be made easier by comparing two side-by-side images. With CycloneCenter, users decide which of two storms is stronger and then proceed to classify the event further. SunSpotter asks for help in classifying sun spots by their complexity.
Platforms and progress
If some of the interfaces start to look familiar, there's good reason. Organizations have now sprung up to help initiatives launch image-based, crowdsourcing platforms. The State Department's MapGive program relies on the HOT Task Manager to support data entry for their humanitarian projects. Geo-Wiki once hosted a popular project to classify images by their land cover and is now available for others to launch related activities. Not all of the research projects that Zooniverse supports are geo-related, but all are “people-powered.”
Editing structures or tagging galaxies may seem like endless tasks but projects do reach an endpoint, and we don’t often enough appreciate the satisfying conclusions. Collaborative efforts contributed to a complete set of global forest cover maps, and the iCoast project allowed almost 8000 images to be tagged following Hurricane Sandy. Any one user’s contributions seem small until the cumulative effort is considered, like the before- and after-earthquake maps of Kathmandu. Every little bit really does help.