Stereo Imaging
Calibrated cameras can be used to measure accurate 3D locations of objects in world coordinates. As already noted in Summary, one camera can only find a line in a 3D space on which the measurement point must lie. Another camera is needed to locate the exact position of the point on that line. This calculation is often called stereo triangulation.
In fact, it is sometimes possible to measure world coordinates from a single view. But this requires prior knowledge of the scene. For example, when controlling a robot that picks items from a planar surface one can fix the z coordinate to zero. There are also techniques that make use of visual cues such as texture gradient and line orientation to estimate 3D structure.
There are a couple of difficulties in finding a unique solution even with two cameras. Obviously, both cameras must see the point to be measured. But they must also know that they are looking at the same object. Even if these prerequisites hold, it is highly unlikely that the two lines the cameras get for the point's location actually intersect in the 3D space. They are almost certainly slightly apart, and the location of the point needs to be estimated from inconsistent information. Usually, the estimate is placed halfway between the 3D lines where they are closest to each other. An even better approximation can be obtained if many cameras are used. PiiStereoTriangulator does just that: it considers all possible pairings of cameras and uses the stereo triangulation technique to find the most likely 3D location by averaging over all stereo pairs.
Add a note
Not a single note added yet. Be the first, add yours.