BOSTON – The father of an open source computer vision library put out a call to ARM developers at
DESIGN East here. Gary Bradski, a lead developer for OpenCV, said the library sorely needs optimization for the ARM architecture that dominates mobile computing and is on the rise in embedded systems.
[Get a 10% discount on ARM TechCon 2012 conference passes by using promo code EDIT. Click here to learn about the show and register.]OpenCV has been moving rapidly on many fronts, said Bradski in a talk at a daylong meeting hosted by the
Embedded Vision Alliance here. But the effort is resource constrained and getting help with ARM optimizations is one area where developers could lend a much-needed hand, said Bradski, chairman of the
non-profit foundation that manages OpenCV.
The library is used as the underpinning of a wide range of computer vision applications. They include Google Maps and Street View, a mine safety app used in China and an automated lifeguard that can detect people drowning in a swimming pool.
“I happen to know it’s used by most militaries, so in the next robot wars we will be on both sides,” said Bradski who is known for his work on Stanley, the robot that won the $2 million Darpa Grand Challenge for navigating an unmanned vehicle across the desert.
The OpenCV community is now generating a new release every quarter. It is accessed at a rate of 110,000 downloads a month and has been downloaded five million times to date.
OpenCV includes image processing routines, transforms and support for computational photography, an app engineers are starting to build into mobile devices. The group is increasing its support for 3-D vision apps and recently upgraded its driver for Microsoft’s Kinect.
The library supports most popular languages and environments. They include Android, C++, Cuda, iOS, Java and Python. Nvidia recently contributed image stabilization code to the library and others contributed iris recognition software.
The Khronos group is considering OpenCV is a possible basis for its poending OpenVL computer vision library. If it is adopted, the Khronos standard could become the basis for future hardware acceleration blocks in a wide variety of chips.
“It will take a year at least for this to solidify,” Bradski said.
Related stories:Introduction to embedded vision and the OpenCV libraryThe next killer app: Machines that see DARPA seeks breakthroughs in computer vision