SteveBury wrote:I'm less than convinced by the methodology that has been banded around for carrying out georeferencing using spheres. My advice would to not bother using them at all. Sphere recognition and noise just don't work well together!
In the trials we carried out using spheres seemed to create more problems than it solved - particularly in level alignment. Even over quite a large area the levels in the Revo point cloud are remarkably good, so I'd just worry about getting the rotation and the XY coordinates correct and trust the level results the Revo's IMU is producing.
I'm finding that surveying-in easily identifiable features such as wall corners then moving/rotating a Pointcab or CloudCompare ortho images created from Revo data onto these in CAD is simple and works well. And you have complete control over what's happening - unlike using a transformation that's easily thrown out because it's relying on spheres that could easily be 20 or 30mm out of their true position because of being selected from noisy data. In addition we use CloudCompare to set the point cloud at the correct level against a known level on a good flat surface.
We're still in the early days of working out the best methods to utilise the Revo. I'm sure we'll develop better systems over time, but at the moment we're finding that, although a bit more time consuming, controlling everything manually gives by far the best results.
Thanks for your reply.
Unfortunately we was told by our supplier this is the way, and this is what we have done! And it's a little late now to revisit site typically.
The problem I have is I can see the spheres when I reduce the scan down by using the viewing tools, but when I try and reference, the cloud goes back to it's entirety meaning I have a ceiling in the way of being able to select the spheres.