Learning OpenCV 4 Computer Vision with Python 3 - Third Edition

Performing 2D-to-3D spatial conversions

Remember that we have a reference image, reference_image.png, and we want our AR application to track a print copy of this image. For the purpose of 3D tracking, we can represent this printed image as a plane in 3D space. Let's define the local coordinate system by saying that, normally (when the elements of the 6DOF pose are all 0), this planar object stands upright like a picture hanging on a wall; its front is the side with the image on it, and its origin is the center of the image.

Now, let's suppose that we want to map a given pixel from the reference image onto this 3D plane. Given the 2D pixel coordinates, the image's pixel dimensions, and a scaling factor to convert from pixels to some unit ...