360VR Camera Systems

< Previous | Contents | Manuals Home | Boris FX | Next >

‌360VR Camera Systems

You might observe that the random dot calibration method provides a very rich data source that is well suited for calibrating multi-camera 360VR camera rigs: there are many dots and they are imaged by all cameras, just not at the same time.

Conceptually, the main task in 360VR calibration is to calibrate the lenses of each camera, and to determine the relative orientation of each camera.

NOTE : 360VR camera rigs are by definition nodal, ie all cameras are located at the same point in 3D. Since this is rarely true for real rigs, but is nonetheless required of the result, the calibration environment must be large compared to the actual distance between cameras. If the cameras are 10 cm apart, and you moved one camera 10 cm, how many pixels would the location of an unmoving spot on the wall move? That will be the error it's a back-of-the-envelope calculation how big the room must be based on your acceptable error.

Determining the relative orientation between cameras is pretty similar to the process we use to locate the 360° loopback point in random dot calibration, which includes not only finding the closest matching frame, but determining the fractional portion and any inconsistency in the tripod sweep.

However, at this point in time we do not have the superstructure in place to do the entire overall process. Obviously there are already commercially available software packages that do 360VR calibration to a usable level of accuracy, likely based on simpler setups. If you feel that adding this capability based on the random dot experimental setup to SynthEyes would offer significant advantages over existing approaches, let us know.

< Previous | Contents | Manuals Home | Boris FX | Next >