Welcome to RouteScout - a moving collection involving media-centric bits and pieces for Spatial Ground Imagery and Corridor Patrol interests
Friday, July 17, 2015
Wednesday, July 8, 2015
Google's DeepStereo makes Street View 'tourism' more realistic
Who here uses Street View to do some virtual traveling? It's not a shabby option if you want to look at places you might never visit within your lifetime, but Google thinks there's still a better way to simulate real-world tours. A team of the company's researchers has developed a technique that uses Google's experimental machine vision algorithm called DeepStereo to transform Street View panoramas into seamless virtual tours. See, Street View photos don't usually capture every detail of a location -- there are always a few frames missing in order to construct convincing digital reproductions of places like museums and houses. DeepStereo can synthesize those missing frames based on the ones that go before and after them, giving Google the ability to create realistic virtual tours.
In order to "train" DeepStero, the team had to feed the algorithm with images captured out of a moving vehicle; after that, it was able to start recreating images. It's still not perfect at this point -- some objects like trees or grass are hard to synthesize and details vanish from the recreated frame if the machine lacks pertinent details. In addition, it takes as long as 12 minutes on a powerful workstation to build a single synthetic frame. DeepStereo's obviously still a young technology, but the team believes it could be used not just to create virtual tours, but to generate environments for movies and virtual reality content in the future.
Monday, July 6, 2015
Mobile GeoInt
GEOINT to go: Optimizing data for mobile devices
Adam Stone, Contributing Writer1:24 p.m. EDT June 10, 2015
Every day, planners are finding new ways to bring vital geospatial data down to mobile devices at the tactical level. Still, tactical mobility is far from being a done deal when it comes to GEOINT.
Bonus: The Army's Dr. Joseph Fontanella discussed the challenges of disseminating geospatial data to disadvantaged troops during a June 3 C4ISR & Networks webcast. The on-demand version of the event is available now. Click here for more.
Despite the plethora of apps already available, technology gurus at NGA, as well as leaders in satellite data on the commercial side, agree there is much work to be done. But the results, if successful, could be substantial.
Exploring new devices
Before he'll even talk about diverse emerging mobility hardware, NGA Mobile Apps Team Lead Ben Tuttle needs to say a couple of words about whether battery life is adequate: Not yet.
The problem is a longstanding one. Mobile devices eat battery life, and the more powerful the devices get, the faster the battery drains. Today's handhelds can stream audio and download detailed imagery, but at what price? "Once we start taking advantage of these processors, we start draining the battery," Tuttle said.
Engineers can't fix it: The laws of physics keep getting in the way. It's a problem that will be around for a while.
That's the bad news. From there, things start to get better. The apps are here, and now the devices are getting better. They are uniformly cool, and the GEOINT field is exploring how to best take advantage oftheir new features.
Take, for instance, Internet-enabled eyewear such as Google Glass. They offer a version of augmented reality, layering data on top of images. Look at a building, see when it was built. Look at an enemy position, see at a glance all available GEOINT.
They're interesting, Tuttle said, because they could allow the war fighter to keep eyes on a scene, absorbing critical information without having to look away at some other device. That's the promise of mobility after all: to take it with you and put it easily to use. Such goggles could potentially take that a step further, integrating GEOINT seamlessly into the tactical experience.
Engineers are still working out the kinks. Early versions are a little delicate for tactical use. They're also too obtrusive for someone not eager to stand out in the crowd. At $5,000 they also are a pricey proposition as compared to the next up-and-coming wearable: smart watches. The Apple Watch models start at $350.
Tuttle likes the watch not for how much data it can pull down, but for how easily it can deliver. "I don't think you will look at a map on a watch," he said. What a user might do is glance down to see that a new message about changing scenarios on the ground has arrived. If it seems worth learning more information about, that user might pull out a phone or tablet to get the details.
It's a compact device and takes little power, easing the battery issue. Tuttle's office is looking into it.
Shrinking the data
New devices bring new questions, said Paul Millhouse, director of technical solutions at satellite imagery provider DigitalGlobe.
Today's wealth of satellite data has been described as a firehose problem. GEOINT is so incredibly rich, and uses bandwidth so intensively, it's impractical to point it at a mobile device and start blasting away. The device could never take it all in. "Any kind of solution you are looking at for GEOINT is inevitably going to have to deal with that," Millhouse said.
Managing the emerging landscape of mobile GEOINT, and its attendant bandwidth problem, will require tools that can bring only the needed information down to the soldier, rather than trying to cram the entire mass of intelligence down the pipe, Millhouse said.
DigitalGlobe's present strategy has been to thin the stream from terabytes down to kilobytes by buffering it in the commercial cloud, and then streaming out only the most relevant, actionable data. In large measure, this technique involves managing imagery. Suppose ground forces need an image of a helicopter landing zone. Rather than push down a map depicting multiple square miles, analysts can process data in the cloud and send out only the necessary, narrow sliver of data troops need to act on.
In some such cases automation helps to bring mobility to GEOINT. Cloud-based analytics can identify relevant objects and thin out data accordingly, making it more readily available for mobile devices.
Integration is another key. By processing data in the cloud, rather than on the device, it's possible to create analytic overlays that combine a landing zone, radio propagation and line-of-sight analysis. Without this preprocessing, mobile devices would not have the computer power to make such specialized use of GEOINT.
Compressing and combining the data serves a vital tactical function, Millhouse said, by giving soldiers ready-access to GEOINT even when they are outside the realm of mobile connectivity.
When data has been cleaned up and streamlined, it can be cached into a mobile device, "so now when they are disconnected, they still have access to the same set of information," he said. In tactical situations that ability to have continued access to maps, imagery and other data has long been the promise of mobility. It's also been a challenge, as engineers work out how to get significant volumes of data to reside on inherently limited devices.
Looking ahead
Running in the background of these efforts is the constant refrain of GEOINT: security. As far as military planners are concerned, nothing can ever be secure enough. When it comes to mobility, that goes double.
"There is still a learning curve," Tuttle said. "People have established how we define security for the old system, and they don't yet know what that is going to be for this [mobile] system. They are having to wrap their heads around it, and sometimes it's scary."
With no easy answer on that score, NGA is pressing ahead with its efforts. On any given day there are hundreds of people in the security, tactical and urban search-and-rescue fields running GEOINT on diverse mobile devices, prototyping future possible uses, Tuttle said.
Wearables aren't actively in the mix yet, but their moment of deployment is likely not far from now. "We are exploring them. We see some value proposition and we are looking at the strengths and weaknesses," Tuttle said. "We are hearing from the consumer: This is something we want to use."
Not that long ago, Tuttle was hearing the same thing about phones and tablets. Even as the military moves toward the next generation of wearable mobility, GEOINT still is settling into a rhythm when it comes to the use of these more conventional devices.
Certainly those in the field like the ready-access to information, Tuttle said. But it's hard to carry a rifle and a tablet, just as it is hard to read a map on a phone. So there are physical limitations. Still, the prospect of near-real-time information sharing in the field continues to make these products enticing.
Even as issues of form factor get worked out, Tuttle predicted, functionality in the mobile space will continue to drive interest. When data is not just mobile, but interactive, then the GEOINT community will really be making headway — be it on tablet, phone or any other device. "People want their maps, but they also want to create stuff on the maps," he said.
Attach a voice or video recording to a map. Draw on the map before sharing it. Now mobility starts to get interesting.
Subscribe to:
Posts (Atom)