Scientist successfully trained their crowd analysis model to track movements and recognize patterns that can lead to danger. Photo by Jesse Silverberg/Harvard University
CAMBRIDGE, Mass., Nov. 22 (UPI) -- The vast majority of mass gatherings come and go without major incident or serious injury. But there are risks lurking in crowds, and researchers at Uppsala University and Harvard University are working to tease them out.
Scientists from the two universities have developed a surveillance model to predict where and how trouble presents itself amid large gatherings of people.
The model can track and analyze large-scale collective motion in real time, pinpointing patterns that may yield stampedes, trampling, crushing pressure or other dangers.
To build their model, scientists first had to program it to accurately track the movements of individuals in a crowd. They used simulated crowds to train the model.
"From this data, we're able to predict the most risky collective motions that naturally arise in a dense shoulder-to-shoulder crowd," Arianna Bottinelli, a graduate student in mathematics at Uppsala, explained in a news release.
"The next step is to apply these techniques to real-time video data," added David Sumpter, a professor of applied mathematics at Uppsala. "If we can use computer vision to track people, then our analytical tools can warn event planners of potential hazards before they arise."
Researchers have previously studied the mechanics of crowd risks. One study looked at the movements of mosh pits at heavy metal concerts. The latest model is one of the first to identify precursor patterns capable of predicting trouble before it starts.
"It all comes down to way people gather into a randomly packed group," Bottinelli said. "Physical body-to-body contacts are the foundation for potentially dangerous collective motion. Our work shows how to identify the emergent risks based on which people are touching each other."
The new research -- detailed in the journal Physical Review Letters -- expands on previous mosh pit analysis, using the same concepts to understand movement within a variety of crowds.
"We were staring at the concert data when we realized there were direct similarities with rallies, protests, and Black Friday sales events," said Jesse Silverberg, postdoctoral fellow at Harvard University. "The more we dug, the richer the physics became. Pretty soon we found ideas from material science and field theory could be applied directly to human crowds in extreme situations."
Silverberg first became interested in crowd movements after taking his girlfriend to a heavy metal show and watching the mosh pit from the outskirts. He realized the movement was similar to way gas molecules bounce around in the atmosphere.
"It was basically just this random mess of collisions, which is essentially how you want to think about the gas in the air that we breathe," he told NPR in 2013.
Three years later and the realization has evolved to inform a new and improved understanding of mass gatherings.
For those planning a trip to crowded shopping malls and big box stores over the holidays, researchers say the best way to avoid injury is to remain and alert.
"If you're packed densely, then there's an inherent risk, and the best way to protect yourself and others is to spread out and move to an area with more physical space" researchers advised.