Object recognition

Augmented reality lets you place virtual 3D objects in your physical environment. In some cases you can also create relationships or interactions with the physical object. To make this possible, the AR software has to be able to recognize the object within your environment. The first step in achieving this is to generate an image of the object using the camera from a phone or tablet.

To identify this image, the software assigns a series of feature points on the object that’s been captured on camera. It then looks closely at these points’ distinctive values, which will be found in places such as an object’s contours and the dividing lines between different planes. Together, these combine to form an identifiable constellation of points.

In this way, the software generates a kind of ‘digital fingerprint’, with a series of more or less unique features. It can then compare this constellation of points against a database of previously scanned objects. Just like when matching fingerprints, an object is recognized when enough distinctive points match up.

Image recognition and tracking
Visual serach in the IKEA Place allows users to search for furniture based on pictures they’ve taken. A world first! See the video above for how it works.