University of Michigan researchers have developed a method for improving a computer's human-tracking accuracy by more than 30 percent by examining where the targets are going and what they are doing.
"Our method reduces the computational complexity and makes it possible to solve the problem of inferring what a person will do based on their activities as an individual, their interactions with other individuals, and their behavior in larger groups," says Michigan professor Silvio Savarese.
He says individual motions provide information about specific interactions, and those interactions can predict a person's future behavior. The researchers taught the software to recognize interactions such as people walking together, standing in a line, or crossing the street. To build the program's knowledge base, the researchers feed it example videos with labeled targets and behaviors, which enabled program to recognize patterns based on the previous experience. Savarese says the researchers will continue to work to accelerate the process, and notes that a simplified version of the tracking software is approaching real-time operation.
From University of Michigan News Service
View Full Article