Satellite images, with their zoomed-out view, have a way of reducing the world’s features to a series of shapes and lines. In a new project called Land Lines, artists Zach Lieberman and Matt Felsen partnered with Google’s Data Arts team to create an interactive platform that lets people explore these shapes with the drag of a finger.
Think of Land Lines as reverse Google image search, but for lines. Using computer vision technology, Liberman and co were able to build a tool that matches hand-drawn lines to those found in a database of more than 50,000 Google satellite images. Draw a line on your desktop or phone, and Land Lines will find a landmass, freeway, bridge, river—something—that follows the same contour. In another version of the tool called “drag” you can pull your cursor over the page, and Land Lines will create a continuous line that connects different landscapes with similar geometries.
On the technical page, Lieberman explains that developing Land Lines was a matter of figuring out how to distinguish similar lines from a diverse set of images. “It’s easy to take out a piece of tracing paper, throw it on top of a printout of one of these photos, and draw the lines that your eye sees,” he writes. “But in general computer vision algorithms for finding lines tend to not work well across very diverse images.” Lieberman wanted a tool that would generate matches, on its own, almost instantaneously. So he tracked down edge- and ridge-detecting algorithms, to help automate the line-finding process.
It’s hard to predict what image a gesture will surface. A semi circle might match to an island off the coast of Australia. Draw a similar curve and it will show you a crop circle in Saudi Arabia. Straight and diagonal lines match mostly to human-made features like roads and rows of crops—but nature still surprises, from time to time, with an uncannily straight shoreline, or outcrop in Australia’s Northern Territory. This unpredictably is the reason Land Lines is so fascinating. It simultaneously shows just how different—and how similar—Earth’s geometries really are.