Not a member yet? Register for full benefits!

AR: Object Recognition in Real-Time

It has always been a staple of augmented reality that the technology integrates physical and virtual data together as one. That said then, the work of a company known as iVisit should come as no surprise. Using a process of machine vision, not altogether different from OCR, the Santa Monica, California, based company has created an application which identifies physical world objects held in front of a camera-phone, regardless of orientation.

The technology, designed to enable partially sighted individuals to have confirmation on what an object is, works by matching a single object on a plain backdrop, to an internal database. It helps if the object contains writing, such as a bank note, or chocolate bar, as this can be used to translate the image to the correct orientation. Facial pictures such as those found on banknotes, work too.

So long as enough of the object is visible, to make a unique identification, then the phone reads off what it is. This means that a folded, crumpled banknote that's been in someone's pocket, can be identified.

Whilst a long way from perfect, and limited to objects that the application has already come across, besides getting hopelessly lost if the camera is pointed at two or more objects, or a single object on a busy background, this attempt at augmenting sight for those who have limited use of the sense, is a promising step forwards.


SeeStar Product Brochure (PDF)

Camera Phones to Interpret Visible World for Blind


Staff Comments


Untitled Document .