First we told our devices how to locate themselves, then we gave our computers the power to parse the profusion of data those devices generate. Now, those devices are returning the favor by providing useful information about the world around us. But many of us have only begun to realize all the possibilities these changes have opened after creeping up on us from multiple directions.

“We sort of slouched into it,” said Dr. Todd S. Bacastow, a professor of practice at USGIF-accredited Pennsylvania State University. “It’s certainly been within the last five to 10 years that we’ve begun to see this massive amount of data and all the opportunity within it.”

Around 25 years ago, only approximately 15 percent of the information collected in the world was geo-tagged, observed Dr. Steven D. Fleming, a professor of spatial sciences with the University of Southern California’s Spatial Sciences Institute, which is also accredited by USGIF to grant academic GEOINT Certificates.

Now? “Most of the world’s data is geo-tagged—I think it’s 85 to 90 percent,” Fleming said. “We know where a banking transaction starts and where it ends. We can track digits. We can certainly track where people are.”

That’s the story of how geospatial intelligence (GEOINT) has generated new perspectives on the natural and built environment. But the next chapter—how companies take these possibilities and turn them into new products and services—includes many plot twists.

These four companies illustrate only a few of many ways the commercial world is leveraging the power of GEOINT.

Teaching Cadillacs to Drive Themselves

 
The form of GEOINT many people know best is the digital map—for example, the latest Geography 2050 conference in New York City focused entirely on mobility. The ability of a phone to locate itself and then offer directions customized to traffic conditions was the stuff of science fiction 30 years ago. But as impressive as the digital cartography of Google and others can be, it’s not precise enough to feed directly to a self-driving vehicle.

So Cadillac decided to commission its own maps before it could include its highway-only Super Cruise self-driving option in the 2018 CT6. The carmaker turned to a Livonia, Mich., firm named Ushr to take navigational mapping to the next level.

“The difference about an autonomous driving map versus a navigation map, we’re concerned about the lane delineators, the slope of the road,” said Chris Thibodeau, senior vice president of Ushr. “In a navigation map, none of that information is needed.”

Plus, an autonomous driving map needs accuracy beyond what GPS can deliver—down to 10 centimeters. Ushr sent cars packed with LiDAR sensors on a tour of America’s highways—220,000 miles driven since 2013.

“It took us about a year and a half to collect and process all that data,” Thibodeau said.

Layered over original LiDAR imagery, Ushr roadway data includes details like cross-slope, lane width, lane markings, and more, all globally geo-referenced to sub-10 centimeter accuracy. Data is available every 0.5 meters along the road. (Image credit: Ushr)

In September, I had the opportunity to take a CT6 on loan from Cadillac for a test drive from Washington, D.C., to Cleveland, and the results were a kind of magic: Once the CT6 recognized it was on a highway in its database, a steering-wheel icon lit up on the dashboard to advise me that Super Cruise was available. I’d press a button to activate this mode, and the top of the steering wheel illuminated in green to show the car had taken over.

Informed by its database, the CT6 stuck to a lane as if it were a rail, slowing and accelerating as needed to compensate for traffic around me. All I had to do was keep my eyes focused on the road ahead—something the car itself watched for, using an inward-facing camera to ensure I was still paying attention.

Ushr is now looking to drive down the costs of its mapping solution, in part by applying machine learning techniques to recognize road features such as stop signs and crosswalks.

“We’re also spending a good amount of engineering resource today on basically automating those feature identification and feature extraction algorithms,” said Brian Radloff, Ushr’s vice president of business development. He added this would allow Ushr to begin mapping secondary roads.

Cadillac might not need that data—company president Johan de Nysschen told me in 2016 that bulky LiDAR sensors needed to detect pedestrians would not fit with a Cadillac’s style—but Ushr has other customers in mind.

“Some municipalities are looking at potentially using this data [in place of conducting their own surveys] if it’s accurate enough,” Radloff said. “When Amazon’s talking about things like drone delivery having a very precise HD map those drones can follow [it could] be another kind of further-out-there application of this technology.”

Bringing Eyes to the Skies—For Rent

 
Unmanned aerial vehicles, more commonly known as drones, are one of the most public symbols of GEOINT’s new era. But many companies that could benefit from the ability of drones to extend human senses to places that are difficult or dangerous for humans to reach lack the budget and expertise to buy their own systems.

That’s where D.C.-based Measure comes in, offering drones and analytical tools clients can hire for particular jobs. This business model—what it calls “Drones as a Service”—has given the firm extensive insight into what drones can and cannot do.

“Agricultural is probably one of the most overhyped applications for drones,” cautioned Abigail Lacy, Measure’s vice president of sales. “Anybody you talk to who’s been in the drone space for more than two years would probably tell you that.”

A drone’s different perspective can, however, make a difference at the margins by gathering data points about exactly where in the field a crop is flourishing or struggling.

“A lot of them really derive from just having the eye in the sky—not just the RGB, but the NDVI,” Lacy said, referring to the Red-Green-Blue of traditional imagery and the Normalized Difference Vegetation Index that a near-infrared camera can produce to indicate the presence of live vegetation.

That, in turn, can allow for a more precise, cheaper application of fertilizer.

But many farmers remain skeptical. “They just tend to be slow adopters when it comes to technology,” she continued. “They’re really hesitant to drop money on all of this different equipment.”

Measure is more bullish about the potential for drones to provide insight for industries such as construction and energy. Lacy cited solar farms as one example, touting the ability of drones to answer questions before construction, such as: “How productive will the solar farm be?” and “Am I going to have water runoff issues on the site?” Once the site is in operation, drones can help identify malfunctioning panels.

Measure doesn’t disclose its rates, but Lacy cited internal research that the company’s service can yield $7,200 in annual savings on a 10-megawatt solar facility compared to traditional inspections.

She noted drones don’t just operate at a lower cost than manned aircraft, they can also get lower to the ground. The firm relies mostly on visual and thermal cameras.

“We are keeping a close eye on how LiDAR is evolving,” she said, but added that so far costs are too high and quality is too low.

The firm also often has to deal with a lesser GEOINT hindrance—every company seems to have its own proprietary software. “You’ll get 15 different software providers that all have a unique system,” Lacy said.

Measure hopes automated data processing will cut down on its own overhead, but the real “game changer” would be automation of a drone’s flight—which, in turn, will require a loosening of regulations that today ban drone flights beyond a human operator’s visual line of sight.

Fusing Maps and Live Data

 
Now that so many mobile devices come equipped with GPS receivers—meaning the apps on those devices can also geo-tag user activities—coping with the massive scale of the resulting data becomes a challenge.

“As the variety of channels and devices that connect customers, companies, and physical assets increases, so too do the ways to measure and analyze spatial information,” a 2016 Forrester report observed. “One of the great challenges for effectively making use of location data has been integrating it with other data sets and analysis to provide deeper context and insight.”

That’s a big theme in the work of MapD, a D.C.-based firm that’s made a specialty out of integrating live data with maps.

One of its most fascinating demos tracks the last several weeks’ worth of geo-tagged tweets around the world, placing them on the map and color-coding them by language. Users can search for keywords and hashtags or just float the cursor across countries to see what is trending. For example, the large rectangle hovering over Finland turns out to be @EveryFinnishNo, a bot that tweets out the Finnish word for a new number every minute.

Another MapD demo offers a similarly granular look at ship movements around the U.S. from 2009 to 2015, both offshore and in lakes and rivers. “Tug” is overwhelmingly the most popular type of vessel, with more than five billion records.

MapD’s New York City taxi ride data set currently totals approximately 1.2 billion records. (Image credit: MapD)

A third demo provides a look at nearly seven years’ worth of taxi rides across New York City, from 2009 to 2015. During that time, cash transactions outnumbered credit, at more than 632.1 million cash transactions versus more than 510.8 million credit—while more than 2.2 million rides were recorded as going uncharged.

The massive computational power provided by GPUs is critical to these efforts.

“GPU computing is really going to take data to the next level and analytics to the next level,” said Monica McEwen, MapD’s vice president for U.S. federal customers. She pointed to how this revolution in processing power allowed Verizon Wireless to accelerate its analysis of network problems.

“Historically, they had to do that in batch mode,” she said. “Today, they’re looking at that in real-time.”

Also important: Ensuring interfaces scale up to meet a density of data she predicted will mean “being able to display literally billions of records and have a response time in the milliseconds.”

“The pure volume of [data] makes it nearly impossible to present it in a fashion in which people can make meaningful sense of it,” McEwen said. As a result, MapD’s interfaces let users easily add or remove layers of data so they can focus on particular variables.

Crowdsourcing Exercise Intelligence

 
Strava, a workout-tracking app popular with many cyclists and runners, has a different challenge to address. Mashing up the location reports it gets from users can inform individual Strava athletes looking to find popular routes on its heatmap.

That trove of data soon caught the attention of urban planners, and that led to a complementary product: a database of cycling and pedestrian activity over time called Strava Metro.

“We started hearing from departments of transportation who said, ‘This is cool, but we can’t see the temporal details,’” said Brian Devaney, sales and marketing lead for Strava Metro. “We had to figure out a way to get all those GPS pulse points and aggregate them and anonymize them.”

Strava’s Global Heatmap of New York City reveals popular routes and activities. (Image credit: Strava)

Combining the heatmap with Metro required the firm to address privacy risks.

Part of its answer is privacy options. Users can choose to place a geo-fence around a home, office, or other location, hiding it and the last 500 meters of a route from the view of others. A more comprehensive enhanced privacy option suppresses even more data from the feeds of other Strava users, down to your last name.

The company won’t say how many users have exercised either option, but many customers never touch the default settings. To keep their information safe as well, Strava aggregates individual GPS measurements without reference to where users started and ended their workouts. The result is an alternative map in which a major highway like Interstate 66 in northern Virginia vanishes from view, while the bike trail next to it glows yellow but leaves no hint of where along the path one person started or ended their ride or run. Strava’s underlying maps, developed by D.C.-based Mapbox on top of cartography from OpenStreetMap, automatically show bike-and pedestrian-hostile roads in gray. Even in small towns like rural Lincoln, Va., enough users walk, run, or bike to leave a dense web of trails on Strava’s heatmap. To use this to get a sense of an individual’s whereabouts, you’d need to know where they live first.

Strava Metro, launched in 2014, offers customers not just the heatmap’s static view (updated once a quarter) of overall movements but also minute-by-minute data about how many people went in one direction on one street. Again, the company boils its data down to GPS points, this time showing direction and time, while removing everything else.

“We do buffer for privacy,” Devaney said of the company’s decision to obscure the start and end of a workout route.

And users can opt out of having their data used in Metro at all, although only “a fraction of a percent” have done so.

Customers such as city and state transportation departments as well as cyclist advocacy organizations use this data to plan or push for improvements such as bike lanes and wider sidewalks, then audit how the new infrastructure performs.

They can learn what corridors are most busy during peak commute times versus on weekends,” Devaney said. “A lot of groups are using the data to understand how behavior changes after they put in infrastructure.”

You can imagine that this data would also be enormously attractive to such businesses as athletic-wear manufacturers, but Strava has chosen to limit its sales of Metro data to organizations “working to influence policy and infrastructure.”

Eschewing commercial use of data gathered from workout-tracking apps happens to line up with one of the core privacy principles put forth last year by the Future of Privacy Forum.

What’s Next? Megacities, Drones, and Small Sats

In terms of its commercial evolution, GEOINT is barely old enough to run for office. What could it look like by the time this roughly 25-year-old discipline is old enough to run for president?

USC’s Fleming said a larger trend—humanity’s move to large cities, in which the height and volume of buildings make the traditional references of GPS unusable or suspect—will force a switch to more resilient location technologies that work better inside and next to large structures.

“We’re piling up people along the coastlines of the world, so we have to deal with megacities better.”

Many smartphone users have already seen this problem when location-based apps lose a GPS signal, decide a nearby WiFi router that happens to have been moved from one venue to another is just as valid, and vault the user to a spot miles away.

Fleming also expects drones to become even more on-demand, “where everyone expects them to be around and they’re providing things like public safety services.”

But a world in which the whine of quadcopter rotors is a normal part of the background din may take some persuasion by drone vendors.

Fleming’s colleague Andrew Marx pointed to a different form of GEOINT system: small sats.

“The advantage of a small sat is you can build up an activity of an object,” Marx said. “You can have so many repeat observations.”

But adding this temporal dimension to GEOINT will require further innovation to display it. The tasks MapD and Strava already face will only grow more arduous.

“It’s a struggle, because we’re trying to depict things in four dimensions,” Marx said.

Penn State’s Bacastow, meanwhile, warned about two trends people might not appreciate as much.

One is which countries are focusing their efforts in this area. “Many of our students in AI and deep learning are not from the U.S.,” he said, referring to a recent presentation by the University of Missouri’s Dr. Curt Davis. “Counting publications, you’d find that scholars from other countries, such as China, have a significantly larger number of publications than scholars from the U.S.”

Another is how different generations view the privacy implications of having their geo-location harvested by smartphone apps. Bacastow recounted a freshman seminar he taught two years ago about geospatial privacy.

“I thought students would be concerned and engaged in a seminar about their loss of privacy,” he said. “Quite honestly, they didn’t care. For them, while they understand the loss of their location privacy, as one student put it, ‘I want my pizza delivered to the right place with the push of a button.’”

Related


Posted by Rob Pegoraro