PITTSBURGH, Sept. 9 (UPI) — Computer vision systems can better understand an image if programmed to make assumptions about the physical constraints of the scene, U.S. researchers say.
Carnegie Mellon University scientists say that like a child using toy building blocks to assemble something that “looks like” a building, a computer could analyze an outdoor scene by using “virtual blocks” to build an approximation of the image based on parameters of volume and mass, a university release reported.
“When people look at a photo, they understand that the scene is geometrically constrained,” Abhinav Gupta of CMU’s Robotics Institute said. “We know that buildings aren’t infinitely thin, that most towers do not lean, and that heavy objects require support.
“It might not be possible to know the three-dimensional size and shape of all the objects in the photo, but we can narrow the possibilities,” he said.
“In the same way, if a computer can replicate an image, block by block, it can better understand the scene.”
Automated scene analysis could eventually be used to understand not only the objects in a scene, but the spaces in between them and what might lie behind areas obscured by objects in the foreground, Alexei A. Efros, associate professor of robotics and computer science at CMU, said.
That level of detail would be important, for instance, if a robot needed to plan a route where it might walk, he noted.
Copyright 2010 United Press International, Inc. (UPI). Any reproduction, republication, redistribution and/or modification of any UPI content is expressly prohibited without UPI’s prior written consent.