News

‘Best-kept little business secret’- Visual Learning Systems comes out of "stealth mode" in Missoula

Missoula company may be on the verge of revolutionizing the use of satellite and aerial photos

A satellite takes a picture of, for instance, a remote corner of Afghanistan.

By MICHAEL MOORE of the Missoulian

Armed with the current technology, the American military begins working on the picture, trying to
identify trails, caves, people in an effort to root out the last al-Qaida strongholds. It’s painstaking work
on computers, and the results can be decidedly imperfect.

Interestingly enough, a small Missoula company may have created a project that will make the
military’s work easier and more precise. The product could also revolutionize the way cities, towns
and private companies use satellite and aerial photographs.

The company is called Visual Learning Systems Inc. http://www.vls-inc.com and the product is Feature Analyst. VLS is the
brainchild of David Opitz, an assistant professor of computer science at the University of Montana,
and Stuart Blundell, a geographic information systems specialist who recently was profiled in the
industry journal Space Business International.

The company incorporated in 1999, but has been "the best-kept little business secret around" as it
developed its product. The company is small – waffling between nine and 13 employees – but stands
on the verge of massive potential growth with its new software.

"Once people see the software, the light just goes off," Blundell said Thursday, as the company held
a coming-out party.

What Feature Analyst does is crack the nut that has boggled the GIS industry for years – how to get
a computer to automatically extract geographical features from digital images.

"Until now, GIS technicians had to either accept bad results from hard-to-use image processing
software or hand-digitize the features by hand," the company’s literature explains. "This is an
expensive, painful and time-consuming process. As a result, billions of research dollars have been
spent, with little success, on trying to automate the problem of feature extraction from digital
imagery."

Companies as big as Lockheed and Boeing have tried to solve the problem, but have thus far been
stymied. Enter Opitz, whose interests growing up were computers and math. Opitz is a "machine
learning" specialist, which is to say he "teaches" computers. Actually, the machine learns from
experience.

Feature Analyst works like this: Digital images are really nothing more than a conglomeration of
pixels, which a computer can read as numbers. Essentially, the computer learns to recognize
patterns in images by recognizing numerical patterns.

For instance, a swimming pool looks like a swimming pool in a picture, but the computer recognizes
its makeup as a pattern of numbers. What the program does is let the user quickly identify a feature
in an image. The program then searches the image for that feature and outlines it. It does so nearly
165 times faster than any other system.

During Thursday’s demonstration, Opitz highlighted several areas of trees – the program then found
and highlighted all the trees in the image, making sure to distinguish between the foliage and the
shadows it casts.

"If I wanted it to find all the swimming pools, we could show it a swimming pool and it will identify
them," Blundell said.

It would not, on the other hand, find other bodies of water. That’s because it places the feature in its
proper context. A river, with its banks, looks different. For the same reason, the program could, if
asked, identify all the lines painted on the middle of roadways, but not those painted on the edge of
the road.

"This software can be very, very specific," Blundell said.

The possible uses, of course, are endless. A municipality could use the software for any number of
things – look for damaged roads in need of repair, find every house with a swimming pool, identify
green space. The U.S. Border Patrol is using the program to map trails along national borders. The
Forest Service is using it to map fires, including distinguishing between burned and unburned foliage.

VLS also has completed the first phase of advanced research for the U.S. Department of Defense
and the federal National Imagery and Mapping Agency, known as NIMA. Visual Learning Systems
Inc. is one of 11 U.S. companies picked by NIMA to investigate new approaches for feature
extraction and is looking forward to NIMA’s second phase of research. Money for that program is in
limbo because of the war efforts stemming from the Sept. 11 terrorist attacks, but Opitz said VLS
could receive as much as $750,000 per year in research funding for three years if the second phase
moves ahead.

"We have a technology that they need more than ever," Blundell said.

A contract at a top government level is precisely what VLS needs to take off. Because it’s a small
company and located in Montana, VLS struggles to battle the so-called "Beltway Bandits,"
companies that locate in the Virginia, Maryland, Washington, D.C., area to get better access to
federal contracts.

"We don’t have a big sales force, but what we do have is good access to our congressional
delegation," Blundell said. "You can just sort of walk around outside the Pentagon as a small
business and never get in. We really need their help to get inside."

Representatives of both Montana senators attended Thursday’s gathering and Blundell has briefed
both senators on VLS and its products.

Despite the difficulty of doing business on the Beltway, Opitz, Blundell and VLS are committed to
Missoula.

"We’re doing what we want to do in Missoula," Opitz said. "We don’t want to move."

Reporter Michael Moore can be reached at 523-5252 or at [email protected]

http://missoulian.com/display/inn_news/news03.txt

Posted in:

Sorry, we couldn't find any posts. Please try a different search.

Leave a Comment

You must be logged in to post a comment.