Not so long ago, we reported on biometrics driven smart environments, in which people can be tracked as they moved through an indoor environment. That was a little disturbing, but Czech inventor Zdenek Kalal is taking surveillance technology even further.
Utilizing the Tracking-Learning-Detection (TLD) technology he developed at the University of Surrey, Kalal’s Predator smart camera is programmed with a learning strategy that enables it to automatically detect its own tracking failures, and therefore get better and better at tracking its object of choice. It’s even smart enough to recognize you again if you leave its field of view, and later return; or to identify you among a group of people in a photo.
Predator face modelingHow does it do it? In short, patterns. It identifies consistent features as it tracks its victim subject, and it tracks its subject by distinguishing between probable and improbable movement, all the while correcting itself and learning more about what makes this particular subject distinct from any other.
It’s as impressive as it is scary.
But it isn’t just for tracking people (although it is very good at it). Kalal and his fans have devised dozens of potential applications, including autonomous navigation, human-computer interfaces, gaming, motion capture, augmented reality, even animal tracking.
Here is Kalal demonstrating the Predator. You need to watch this, because one day it might be watching you:
