Communications of the ACM
Machines That Can See
Computer vision or object recognition has progressed by leaps and bounds to the degree that machines can in some instances surpass humans. Applications being developed and/or implemented include monitoring systems in which video footage relayed to a computer can determine whether an elderly resident is in trouble or having difficulty so the system can take remedial action; software that can rate a facial expression's authenticity by measuring the difference in the duration of its formation and disappearance; computer-vision systems on product assembly lines that can spot faults with more accuracy than humans; and digital signs that use cameras and software to measure various factors about passersby — gender, age, etc. — and tailor ads for them.
The University of British Columbia's Jim Little is working to improve robots' precision by linking them wirelessly to the Internet, allowing them to search for pictures online so that they can quickly learn to recognize objects in close proximity. Crime prevention applications of computer-vision systems include software that can recognize the make, model, and color of moving vehicles and use data from roadside cameras to alert authorities that a car's license plate has been switched. The Technion-Israel Institute of Technology is working on roadside vision systems for hazardous junctions. If the system spots a potential collision between approaching vehicles, street signs are flashed to warn motorists.
From The Economist Technology Quarterly
View Full Article
No entries found