New Algorithms for Augmented Reality The Glasses know the Way

Digitale Anleitungen, die Monteure über Smart Glasses eingeblendet bekommen, könnten herkömmliche Papieranleitungen eines Tages ablösen.
Digital instructions, which are displayed to fitters via Smart Glasses, could one day replace conventional paper instructions.

In order to work reliably, AR glasses must be able to determine their exact position in the room. To facilitate the navigation of AR applications, researchers at the Ruhr University Bochum are working on algorithms for automatic calibration.

Augmented reality (AR) applications are promising for the factory of the future. Intelligent eyewear that supports AR could simplify maintenance and repair work in buildings by navigating fitters to the job site and showing them step-by-step what to do. For this to work, the devices must be able to determine their exact position in space and understand what they are seeing. The team led by Prof. Dr. Markus König from the Bochum Chair of Computer Science in Civil Engineering wants to give AR glasses a helping hand. For this reason, the team is developing algorithms that make this possible automatically without manual calibration.

The algorithm requires a digital building model for its function. It compares the image recorded by the camera in the Smart Glasses with the model. In order to provide the spectacles with as much image information of the surroundings as possible, the wearer must first take a look around the room. The algorithm rotates and shifts the digital model until the model and images of the surroundings fit on top of each other. If necessary, it proceeds pixel by pixel, using the depth information recorded by modern cameras.

Automatic instead of manual

Conventional methods rely on a manual calibration, which is carried out over at least two points, which are glued in different places in the room and also recorded in the digital model. As soon as the user enters the room with the Smart Glasses, he must first enter the information on the points into the system so that it can calculate its three-dimensional position in the room.

In contrast, the Bochum researchers can use their algorithm to automatically determine the position of the glasses not only in one room, but in an entire building. With the manual point calibration method, this would only be possible if the user recalibrated frequently or used a large number of points.

Automatic calibration currently works to an accuracy of 20 cm, which is sufficient to navigate with. For other applications, however, the Bochum researchers want to further optimize the algorithm.

In addition, it is important to the researchers that it works in real time - even if it runs on a smartphone. Smart Glasses, however, have a lower computing power, so that the application should become even more efficient in order to do its job smoothly here as well.