virtual interactive whiteboard
HOLOS stands for High speed Object Location and Orientation Sensing. HOLOS tracking software detects both the exact absolute 3D position and 3D orientation of a patterned sphere in space just from its appearance in a single 2D camera frame alone. Getting this full 3D information from only 2D is possible because the third dimension can be inferred from the change in size of the sphere in the image, depending on its distance from the camera. Different orientations in 3D space result in different curvatures of the bands in the pattern on the camera chip, as indicated in Figure 1.
When the user presses a button or activates the pen tip by pressing it against a hard surface, infrared leds inside the pen light up the sphere. The pattern on the sphere is made in such a way that it is only visible to the camera when these leds are on. Mouse left and right clicks, writing or drawing action is distinguished by switching the pattern on at different levels of brightness. Figure 2 shows how simple and cost effective the electronics can be: The pen is essentially nothing more than a small infrared led flashlight.
The core advantage of our proprietary detection algorithm derives from the exceptionally high precision at which it can distinguish very slight subpixel changes of the pattern’s position, diameter and band curvature in the camera image even if the sphere is far away and the pattern appears very small and pixelated, blurred or out of focus. Further the algorithm is very efficient and only needs a few milliseconds to do the calculation running on a single core of a typical processor. Subpixel changes smaller than 1/50th of a pixel can be detected even if the overall pattern’s appearance is not bigger than just 20 pixels in diameter (about the size of a single character in this text).
The software is further able to extract the angles of 3D orientation from the band curvature in such tiny patterns at an accuracy as good as 0.07 of an angular degree. Note that the ability to extract the angular orientation at this high precision is what enables our unique method of inferring the exact tip position from looking at the back of the pen alone. Any less accurate detection would magnify the error too much along the length of the pen, see Figure 4.1 and 4.2.