What would provide the best tracking?
1) using a cylindrical target
2) using a faceted 8-sided cylindrical approximation using a mutli-target with images.
I presume that the multi-target uses the same single image recognition underneath as single image targets and thus there will be a frame around each image of the multi target that will not contain features that are tracked. Thus the transition of the camera's field of view from one plane to the next one in the 8-sided scenario will contain some time where there are very few features visible to the camera.
But how accurate is the feature-detection of cylindrical features that need to go through another projection based off the cylinder's diameter with respect to the flat image target that is uploaded for its side image. I guess the features here are at least more 'continuous' when the camera's field of view moves sideways or when the cylinder is rotated, so there will be more features visible during this....?
If the image targets in the multi-target scenario provide more continuous features between the boundaries of the images that make up the multi-target, does scenario 2 provide better tracking?
I'd appreciate if someone can provide more insight into this.