The pixels in digital ortophotographs, the addresses in a directory of bicycle repair shops, and the coordinates of street centerlines are all examples of geographic data — because they all record the location and one or more features of physical objects on the surface of the Earth. Devices used to collect geographic data include GPS receivers, inertial navigation units, total stations, laser-based distance and speed measuring systems, LiDAR systems, and multi-spectral optical scanners (basically, very sensitive and sophisticated digital cameras) — carried on or mounted onto on- and off-street land vehicles, ships, aircraft, and Earth-orbiting and geostationary satellites.
On any given day in the United States, a few thousand vehicles are systematically collecting geographic data — including street centerlines, number of lanes, turn restrictions, speed limits, traffic speed, locations of television and radio broadcast antennas, and street-level images of streets, building, and signage. The two leading worldwide collectors and providers of digital map data are Tele Atlas and NAVTEQ; other companies with large mapping operations include DeLorme, Rand McNally, Yahoo, Microsoft, Google, and Skyhook Wireless (which maps the locations of wireless routers and uses this data for its proprietary location system). At the 46th annual conference of the Urban and Regional Information Systems Association (URISA), which I am attending in New Orleans, of 21 vendors present, four specialize in street-level data collection:
Additionally, state and local government agencies, also routinely collect geographic data.
I asked six companies to discuss the types and amounts of geographic data that they collect, how they prioritize which streets to drive, how often they re-drive them, what data they acquire from third parties, and some other related questions. Google told me that it does not collect any geographic data directly, but instead acquires all of it from third-party providers — in the United States, from Tele Atlas. (What about the imagery for Google’s Street View feature? Doesn’t Google use its own vans to collect it?) DeLorme declined to answer my questions, telling me that they involved “information about proprietary DeLorme methodology.” Skyhook Wireless told me essentially the same thing. Microsoft also declined to answer, but did not bother to give me a reason. By contrast, Tele Atlas’ VP of Marketing, Michael Evers, replied to my questions in writing and NAVTEQ’s VP of Technology, Aaron Crane, gave me detailed answers in a telephone interview.
Tele Atlas
Tele Atlas’ digital map database, Evers told me, contains 17 million miles of roads around the world, which can change 10-15 percent every year, or even more in busy urban areas. The information that the company collects includes millions of points of interest (POIs), 3D landmarks, traffic alert information, street and building signage, lane information, precise address locations, road slope, and curvature measurements. It gathers and validates changes and updates its maps every day. Typically, Tele Atlas partners integrate these updates within their maps four times a year or, in some instances, on demand (daily or monthly).
To update its map data, Evers says, Tele Atlas uses traditional cars and local data collectors to validate changes and mobile mapping vans to capture changes faster. “Our vans can travel quickly, taking ‘live’ pictures of things along the way, such as exit signs, street names and road curvatures, so that we can automatically see where changes have occurred and what’s on the roads themselves,” he says. Behind these drivers, he points out, is a worldwide network that includes satellite and aerial imagery, public data sources and filings, tax maps, Web crawler tools, government documents, and input from utility, fleet, and postal drivers and map users. The company encourages its partners and their end users, including fleet and enterprises with many vehicles on the road, to contribute feedback to it via its online tool, Tele Atlas Map Insight. “Additionally,” says Evers, “we work with industry partners to obtain an integrated data flow, including incident, real-time, and predictive information that we use to enhance our maps.”
Tele Atlas, according to Evers, uses data mining and integration tools and other technologies to rapidly review, validate, and implement database changes. “We use this tried and true process to check the content collected from any source, end users included, to ensure any edit to the database meets our standards.” The company’s fleets, he adds, operate around the world on a carefully prepared schedule that aims to ensure all major roadways are checked frequently and other roads are checked regularly. For example, earlier this year, its mobile mapping vans covered more than 353,000 miles of state highways in the United States and identified more than 420,000 speed restriction signs, more than a million street name signs, approximately 625,000 route numbers, and more than 530,000 information signposts. The data also identified slope of roadway, bridge heights, turn curvature, and other navigation and routing information, such as traffic lights, tolls, speed bump warnings, lane information, and information related to restrictions such as maximum dimension signage and commercial vehicle/truck attribute information.
NAVTEQ
NAVTEQ has approximately 300 vehicles out there driving now, Crane told me, distributed roughly evenly across the company’s geographic coverage areas, but it can bring on more on a temporary basis, as needed. The company drives millions of miles every year and makes millions of changes to its database every day. “We focus our collection efforts on the basis of our knowledge of the areas that we have to drive,” Crane says. “We have a field force in place throughout the United States and other countries and these people know what is happening in their area and what is changing, so that they can guide their efforts toward those changes. We know that certain metropolitan areas are expanding quickly and we schedule our drivers to cover those roads. Logically, we also focus more on where there is a greater population, but we also work to cover rural areas.”
NAVTEQ’s fleet collects up to 260 attributes and that number increases almost every quarter, Crane says, as the company adds new features to its database. These attributes include street centerlines, one-way direction of travel, turn restrictions and the times when they are in effect, number of lanes, sign text, etc., as well as attributes that describe complex intersections, overpasses, and underpasses.
“Sometimes,” Crane says, “you can get these types of things from sources, but we want to be sure that the quality and consistency of the information is up to our standards. So, if we do collect information from sources, they have to be known sources and we have to verify the data before we bring it into our database. Typically, we are collecting geometry on our own.”
NAVTEQ routinely does research on what parts of its map building process can be automated and implements additional automation whenever possible. However, Crane says, “We don’t see a world where there is a completely automated process. The local expertise is very helpful.”
The company has also investigated whether and to what extent it can use other fleets to collect data for its maps. “One of the things that we have to weigh is the accuracy of their data collection against our standards,” says Crane. “We drive a particular way, so that the geometry that we collect is consistent and matches our specifications. UPS drivers, for example, are not driving thinking about where they are in the road, so their information would be complementary to ours, but it might not replace ours, because it is not being collected with the same goal.”
From time to time, NAVTEQ has also evaluated whether to collect data that it doesn’t need for its own purposes but that it could sell to third parties, but has not yet decided to do so. “We have to evaluate it with regards to how it would be collected,” says Crane, “what the training requirements would be for our field staff, and whether it would be a distraction from what we are trying to accomplish.”
The company also currently collects GPS probe data for traffic information and evaluates it as to whether or not it would be helpful for its map-building processes. Real-time GPS data fed back to its processing centers allow it to determine traffic speeds and calculate historical traffic patterns.
NAVTEQ also uses data from local governments and businesses to help build its maps. “It depends on the geography and the type of attribute that we are trying to collect,” says Crane. “In countries where it is hard for us to deploy a mobile mapping force or in which there are good map providers, we will try to partner with them and bring their mapping data into our environment and build our map that way. Typically, though, they have not built it to our specification, and then we have to do some level of re-work on those sources in order to bring them up to our spec and our quality standards.”
“Ours is a very powerful mapping system. It allows our geographic analysts to work on our map products from different geographic areas, which is very powerful. It is centralized and maintains a level of consistency and quality at all times, as we are building the map. We have more than a thousand ‘validation rules’—automated checks that are applied to the data and get kicked off as edits are being made and as they are corrected or cleaned up by the editor or escalated to somebody in the production process who has more knowledge, to be cleaned up before the data is released to the public.”
By: Matteo Luccio, President, Pale Blue Dot Research, Writing, and Editing, LLC
