Personal compact drones are becoming more popular around the world, despite jumbled government regulations. While that trend has so far resulted mostly in some cool aerial photography and video, hobbyist drone pilots are quickly expanding the capabilities of their craft. Take this trio of programmers in Brazil, who recently whipped up a simple real-time object tracking algorithm for their homebuilt drone's video camera.The algorithm allows the drone to pinpoint a designated object in crosshairs and keep it in view while hovering.While it seems pretty basic right now, that's kind of the point: it was written in just 50 lines of open source Python code, meaning anyone else with a drone can use it. Obviously future iterations should only improve.It's hardly the first example of hobbyist tracker drones though. Back in 2011, then-graduate student Mike Mogenson created a facial recognition and track algorithm for the Parrot AR drone, which he also open sourced.Other enterprising grad students have developed similar algorithms over the years, allowing their drones to follow people around autonomously.Perhaps the spookiest example of these types of algorithms is the appropriately named 'Predator,' an object-tracking system developed by another graduate student that quickly "learns" a specific object, recognizing it even if it moves out of frame and returns, and can track multiple objects at once.While they may pale in comparison to the types of tracking abilities found on larger drones increasingly used by government and industry, the growing number of open source algorithms points to a future where autonomous piloting abilities are readily available to the common hobbyist.
Advertisement