Researchers have developed a method for tracking hundreds of individuals at once, giving new meaning to the phrase “big brother is always watching”. While it may sound like something out of a futuristic spy movie, this technology could help to spot suspicious behavior and predict the likelihood of dangerous crowd movements, potentially saving countless lives.
Previously, computer-based studies have only been able to track one person at a time in recorded videos and rely heavily on appearance, making distinguishing someone from above in low-resolution video challenging to say the least.
In order to address this issue, the new research, to be published in IEEE Transactions on Pattern Analysis and Machine Intelligence, focuses on using a mathematical function to analyze five factors to predict where an individual will go next. The software observes appearance, target motion, neighbor motion, spatial proximity and grouping to predict how someone will move in a crowd and what their likely next course of action will be.
Mubarak Shah and Afshin Dehghan, the developers of the algorithm, tested their method by analyzing nine crowd videos ranging from 57 to 747 people. The program’s accuracy at tracking every person ranged from 67 percent to 99 percent, a performance that either matched or surpassed the results of five other algorithms which all track people one at a time.
Though the method currently does not work in real time, the researchers are hopeful that the algorithm can be improved upon in the near future. Being able to track crowds of people at once could prove to be a major step forward in ensuring crowd security in public spaces and identifying potentially dangerous behavior.
Lauren Leising is a freelance writer based in Athens, Georgia.