An algorithm for keeping track of overlapping critters
To better study the interactions and movements of whales and birds in the wild, researchers tag them with GPS tracking devices. But tracking critters in more crowded environments, such as a Petri dish, can be challenging—especially if, like fruitflies, they zig and zag in overlapping movements; or, like bacteria, they wriggle and glide while in contact with one another.
“If you have multiple objects far from each other, it’s easy. But when they get closer, the tracker gets confused,” says Joshua Shaevitz, PhD, associate professor of physics at Princeton University and the Lewis-Sigler Institute for Integrative Genomics.
To remedy that problem, Shaevitz and his colleagues built on an approach called an active contours or “snake” algorithm, which treats the bright parts of an image as a kind of energy that needs to be maximized. “You put wiggly lines around and they try to dock where the energy is,” Shaevitz says. But, typically, the snakes are allowed to merge, break, fork, and recombine.
In Shaevitz’ team’s innovation, the algorithm knows how many objects are present and assumes not only that they don’t disappear or get generated, but also that they repel each other—preventing them from collapsing into one another if they overlap.
The algorithm can track any objects that get close to each other, and is already widely used for tracking Drosophila melanogaster, the fruitfly commonly used for research.