Hyperspectral imaging eyed for machine vision
More than $2.8 million in state funding aims to kick hyperspectral imaging and other optical technologies into hyperdrive.
Two sets of grants will aid commercialization of technologies developed at Montana State University (MSU), fostering partnerships with several Bozeman-area businesses and even helping to launch a pair of university spinoffs.
“In this overall effort, what we’re trying to do is strengthen and broaden out the high-tech portions of the Montana economy, specifically optics and photonics … and start coupling them with some of the more traditional Montana economies like agriculture,” said Dr. Joseph Shaw, a professor and director of the university’s Optical Technology Center.
In one project, university researchers will develop new image processing technology for Resonon Inc., a manufacturer of hyperspectral machine vision systems used for automated sorting and inspection (among other applications).
Resonon’s hyperspectral imagers are already used on production lines to identify defects in nuts and berries. This is possible because the imagers capture 240 color bands rather than the three or four used by conventional cameras. But due to computer processing constraints, only a small portion of all that color data can be used in real time.
“There are only a few seconds (and in some cases, less than that) between the time the hyperspectral data are collected and when algorithms need to convert the data into useful information to direct the actuators (e.g., robots),” said Resonon President Rand Swanson. “Currently, we do not have the computational bandwidth to exploit the data to its fullest extent, plus we are imaging at relatively low resolution.”
Resonon is counting on field-programmable gate arrays (FPGAs) to ease the bottleneck and enable higher-resolution hyperspectral imaging. The FPGAs will be developed by researchers at MSU and Impulse Accelerated Technologies Inc. of Bellevue, Wash.
Current hyperspectral imaging systems operate with separate cameras, frame grabbers and computers, which constrains how quickly data can be transferred, according to Dr. Ross Snider, an associate professor of electrical and computer engineering at MSU. FPGAs would cut out the intermediate step and increase computational bandwidth by allowing cameras to perform their own image processing.
Faster processing of hyperspectral images has potential uses outside the food industry.
“This technology has surprisingly broad application,” said Brian Durwood, co-owner of Impulse Accelerated Technologies. “The same algorithmic architecture that visually sorts shells from almonds can also be used to more quickly scan the sea and ‘sort out’ a person overboard as differentiated from the surrounding sea, from an automated camera suspended under a Coast Guard plane.”
The MSU project, though, will have a narrower scope. It is focused on two application areas: food sorting based on shape detection using mid- and high-end FPGAs, and data recording for environmental monitoring using low-end FPGAs.
Last month the Montana Board of Research and Commercialization Technology (MBRCT) awarded the project a $100,000 grant. The project will also share in a $2.5 million grant awarded by the Montana Research and Economic Development Initiative and overseen by Shaw.
Source : http://www.photonics.com/Article.aspx?AID=57661