In Clovis, N.M., the painstaking, manual task of geocoding every street sign and roadway object has been automated using an imagery platform from Mapillary that integrates cameras, computer vision and algorithms.
Even in a town the size of Clovis, N.M., no one has ever taken on the job of counting the several thousand highway and street signs.
“Nobody had done a sign inventory before,” said Steve Hewett, GIS specialist for Clovis, a city of about 40,000 near the Texas border. “This was basically our first sign inventory that the city has done.”
Last year, using software developed by Mapillary, a street-level imagery firm, Hewett catalogued and mapped some 4,000 to 5,000 traffic signs. He’s able to accomplish this daunting task by quickly shooting photos of the signs and letting the Mapillary technology determine the type of sign, its location and other data. If done manually, this work would require a team of professionals and can cost $4 to $6 per sign, according to Mapillary officials.
“It was done as a way to relatively inexpensively collect [information about] how many signs we have in the city without paying somebody else to do it. That would be incredibly cost-prohibitive for us,” said Hewett.
Mapillary uses computer vision and algorithms to scan and detect more than 40 different objects in a scene and catalogue those onto a map. These range from fire hydrants to utility poles to mailboxes and more than 1,500 different sorts of traffic signs.
The company, based in Sweden, recently introduced its Capture Projects feature, which allows anyone to gather the imagery and mapping data.
“It could be a road authority that’s interested in just capturing an intersection that’s being developed, or a construction site," explained Janine Yoong, vice president of business development at Mapillary. "Basically, the product allows you to define the area you’re interested in diving into and it gives you the ability to manage an unlimited number of drivers to capture imagery at any given point. Once that’s done, it goes through our platform, which automatically processes the imagery in computer vision, and extracts the object classes like utility poles and fire hydrants and traffic signs. And again, it depends on what the city is looking for and needs.”
Capture Project is the latest push by Mapillary to aid cities that want to have “more control over map data collection for building better maps,” said Yoong.
Other technology companies like Coord have also ventured into mapping technology with its Surveyor iPhone app, which allows users to mark the beginning of a street curb and then photograph objects in the right-of-way.
In Clovis, Hewett has covered some 1,300 miles of streets and gathered nearly 277,000 images across the city’s 23 square miles. He plans to branch out to mapping other data sets like fire hydrant locations and streetlights. The scenes were shot with three cameras he controlled with the Mapillary app on his iPad.
“The cameras were set to take a photo every second, which was the fastest time I could get and have a fairly accurate GPS point,” said Hewett, explaining how he gathered so many images. “I tried a half-second rate but found that there was more positional error, and points would jump forward and backward with the error, and I was not okay with that. I also tried using my phone mounted to my windshield and that did okay, but I would sometimes get out of focus photos and the camera chose to focus on a smashed bug or dirt instead of out at infinity.”
“This [Mapillary] has been a life-saver,” he added. “Because I could have spent my entire career just collecting traffic signs and I’d probably have retired before I got them all."