Hello
I have calibrated the cameras using the new three marker wand and the first results have been very good, but in this case all the cameras where looking to the same place.
We would like to increase our working volume, then we modified the camera orientation. In that case all the cameras are not looking to the same point, but now the calibration results are catastrophic.
How can we get a bigger working volume and a good calibration simultaneously?
Thank you
Aiert
Camera calibration and working volume dimensions
Re: Camera calibration and working volume dimensions
Future updates to the 3-marker calibration will include making the situation describe work similarly to when all cameras are pointed into a central location.
That said, even the current calibration algorithm is capable of having cameras pointed away from a central point. In order to achieve this in the past, I have started with the cameras in their locations all pointing into a central point. Then perform a full calibration. Once you've got a good solution for that arrangement, you can point the cameras away from the center slightly, and then perform a 'refine' calibration. This way the solver will start with the previous solution which includes their location. You can also limit the solver scope to 'extrinsics', which means only position and orientation. If the solution is progressing towards a proper solution, you can switch the solver scope back to 'all' to allow the solution to descend to the best result.
Then iterate the 'refine' process again by pointing the cameras farther away from the center.
Again, this will become automatic in the future, but for now it is possible, albeit a more manual process.
That said, even the current calibration algorithm is capable of having cameras pointed away from a central point. In order to achieve this in the past, I have started with the cameras in their locations all pointing into a central point. Then perform a full calibration. Once you've got a good solution for that arrangement, you can point the cameras away from the center slightly, and then perform a 'refine' calibration. This way the solver will start with the previous solution which includes their location. You can also limit the solver scope to 'extrinsics', which means only position and orientation. If the solution is progressing towards a proper solution, you can switch the solver scope back to 'all' to allow the solution to descend to the best result.
Then iterate the 'refine' process again by pointing the cameras farther away from the center.
Again, this will become automatic in the future, but for now it is possible, albeit a more manual process.
Re: Camera calibration and working volume dimensions
Doug, I assume you do this with another version as the arena 1.6.3. Since we didn't find a way of 'refining calibration'. We need to track a space of 5m x 4m but this doesn't seem to be that easy even when working with 10 camera's.
It would be great if it was in a future update, it would also be nice if there is a thorough description on how to do this, if you say it is.
Best
pieter
It would be great if it was in a future update, it would also be nice if there is a thorough description on how to do this, if you say it is.
Best
pieter
Re: Camera calibration and working volume dimensions
Currently this option is only available within the Tracking Tools 3-point calibration.