Dave Kaeli, professor of electrical and computer engineering
Kaeli is designing crowd analytics platforms to automatically identify suspicious behavior in vulnerable gathering places, such as transportation hubs and concert halls—and he is the first in this field to solve the data processing issues by employing graphics processing units.
Continuous detection of individual threats in crowded locales requires comparisons of billions of data points, which even state-of-the-art surveillance systems cannot handle. But Kaeli’s GPUs can. Data from cameras, infrared devices, motion detectors, microphones, and even heat sensors, can be quickly processed in parallel using this technology.
Kaeli’s team has developed novel algorithms to define normal behavior in terms of these data variables. If something out of the ordinary is happening, the computer will recognize it. In turn, reasoning schema being developed by members of his team will enable the computer to decide whether the anomalous data indicate a significant—and potentially high-impact—event is about to take place.