The momentum building behind AR has been fuelled by the growing sophistication of cellphones. With the right software, devices like the iPhone can now overlay reviews of local services or navigation information onto scenes from the phone's camera. A good example of what is possible today are the AR features in an app released by Yelp, a website that collects shop and restaurant reviews for cities in North America, Ireland and the UK. The company's app, for the iPhone 3GS released last month, uses the phone's camera to display images of nearby businesses, and then lists links to relevant reviews. Yelp's app relies on the GPS receiver and compass that are built into the handset. Together, these sensors can identify the phone's location and orientation, allowing such apps to call up corresponding information. Other AR apps include virtual Post-it notes that you can leave in specific places, and a navigation system that marks out a route to your destination. Meanwhile, companies are working on games in which characters will appear to move around real environments. However, when iPhone was tested in downtown San Francisco, the error reported by the GPS sensor was as great as 70 metres, and the compass leapt through 180 degrees as the phone moved past a metal sculpture. Yelp says the app's AR features are a ‘very early iteration’ that the company will improve as it gets feedback.
Some researchers doubt whether high-accuracy GPS systems will ever be small or efficient enough to incorporate into mobile phones. Others suggest achieving the sub-metre positioning accuracy that really good AR demands, mobile devices will have to analyse scenes, not just record images. One way to achieve this is to combine laser scans of a city with conventional images to create a three-dimensional computer map. Each building in the map would be represented by a block of the right size and shape, with the camera image of the real building mapped onto it. The phone could use GPS and compass data to get a rough location and orientation reading, then compare the camera image with the model to get a more precise location fix. Various interested companies are building 3D city maps, including Google and Microsoft, but it is doubtful that such maps will achieve truly global coverage. The models will also inevitably lag behind reality, as buildings are knocked down or new ones appear. So such shortcomings have inspired other researchers to consider a ‘crowd-sourced’ solution to speed up data collection. In this approach, software would pull photographs of a location from the internet and stitch the pictures together to create a 3D image of that place. Such images could also have GPS information attached, and even though the coordinates might be slightly inaccurate, combining many photographs of the same place would fine-tune the location information embedded in the resulting composite image.
More information:
http://www.newscientist.com/article/mg20327267.700-augmented-reality-gets-off-to-a-wobbly-start.html
More information:
http://www.newscientist.com/article/mg20327267.700-augmented-reality-gets-off-to-a-wobbly-start.html