Remember meForgot password?
    Log in with Twitter

article imageBlippar builds computer vision solution to improve AR in cities

By James Walker     Aug 5, 2017 in Technology
Augmented reality startup Blippar has announced a prototype app that makes augmented reality in cities faster and more precise. The company intends to make future augmented reality apps, like the next Pokémon Go, more responsive and visually accurate.
Pokémon Go singlehandedly introduced smartphone augmented reality to millions of people, capturing new fans by overlaying gameplay on the real world. Blippar has been working with the technology for years though, developing AR apps that display information about the places around you.
AR on smartphones faces two related challenges: it's often slightly sluggish and it tends to look slightly out of place. Because AR apps have to continuously work out your location using a combination of GPS, accelerometer and compass data, performance can be lacking.
When the GPS signal is weak, such as in a city, responsiveness can be further degraded and measurement inaccuracies occur. The result is the virtual objects don't quite move with the phone and can seem to be misaligned with the edges of the real world.
READ NEXT: Virtual and augmented reality coming to the fitness world
Blippar's now working to solve these problems using a new approach. The MIT Technology Review reports the company is licensing street-level imagery from services such as Google Street View. It's then indexing the images and matching them with their real positions. When a smartphone user points their device at the location, the image in the camera viewfinder should match one of the indexed reference shots.
By using computer vision technology, Blippar can determine the phone's position using the feed from the camera alone. The app continually matches the image from the camera with its set of indexed photos. The dataset contains photos of buildings and public places from different angles, directions and distances, accounting for every possible position of the phone.
Blippar claims the technology is accurate to within eight meters. In a best case scenario, it could be accurate to three meters. This compares favourably with the typical five metre range of GPS, especially as GPS becomes notoriously less accurate in built-up areas.
READ NEXT: Report: AR app Layar's creators want to buy it back from Blippar
A prototype video created by Blippar demonstrates the technology in action. Although the graphics are obviously basic, they do seem to respond quickly and move smoothly through the scene. Blippar intends to rollout iOS and Android apps later this year that will offer public demonstrations.
Blippar's approach to AR positioning is innovative but not without its own fault. Because it relies wholly on previously captured imaging data, it only works in locations that have been indexed by the app. So far, it's limited to select areas of London in the UK and Mountain View and San Francisco in California. Although the wealth of Google Street View data available makes expansion relatively simple, it's not going to offer the planetary scale of GPS anytime soon.
More about blippar, computer vision, augmented reality, future tech, Smartphones
Latest News
Top News