Advances in computer vision help robots see

New Scientist : Researchers at Cornell University have developed an algorithm to help robots identify objects in homes and offices. In their system, Hema Koppula and Abhishek Anand used a low-cost Microsoft Kinect sensor, which perceives scenes in three dimensions with two cameras and an IR sensor. Their algorithm uses the data to determine the color and shape of individual objects. The researchers also included instructions to look for certain objects in certain locations; for example, computer monitors are normally found on top of a desk or table, rather than underneath. To find out how the algorithm would perform in a real-world setting, the researchers mounted a Kinect on a mobile robot and asked it to find a keyboard. The robot examined its surroundings, spotted a computer monitor, and moved in for a closer look, knowing that keyboards are often found nearby. The work could revolutionize computer vision, said Daniel Huber of Carnegie Mellon University.