Trimble Business Center

 View Only
Expand all | Collapse all

Registration on regions instead of complete clouds

  • 1.  Registration on regions instead of complete clouds

    Posted 06-15-2020 00:47

    Is it possible to only use cloud regions when registering point clouds pairwise? That way, you can clean up the point clouds before registering others. Handy for removing moving objects (cranes, trucks, cars, people), or even when monitoring to delete parts that have "evolved".

    Or simply only register on a part of a building. Registration on a complete building (leaving all the PC's on) gives (numericaly) less overlap than on the only part that matters (turning some PC's off).

    It would give a better view on how accurate the registration is. 



  • 2.  Re: Registration on regions instead of complete clouds

    Posted 06-15-2020 17:34

    Hi Nico, this is possible in TBC, but I'm not sure if that's the best way to process the data. Imagine this scenario, you have several scan stations. The point clouds will be imported and sit on top of each other. When they enter TBC, they are all in the Default region. If you classify them altogether, the result would be very bad because building walls can all be tilted or misaligned. If you classify them one by one, that takes quite some time and the result can be bad as well, because one scan may only captures a small portion of a wall, and TBC doesn't know if it should be classified as a building or a tree trunk. Under this circumstance, it makes more sense to classify the point cloud after you align/register all of them.

     

    As for the registration accuracy, TBC doesn't get much influence from noise or ghosting points. It tries to use large chunk of planes to do the registration. Therefore, even if it's a busy road with cars in it, it doesn't impact the result. As for the overlap percentage, it's not very accurate in many cases. I would absolutely trust the visual inspection more than that value.  

    Let me know if this makes sense. Thank you.



  • 3.  Re: Registration on regions instead of complete clouds

    Posted 06-29-2020 03:05

    Thanks for the response, Rui.

     

    For the moment I'm trying to register some scans. In this case it's from a building "in progress". Some time ago I've made a scan of the building with the walls only partially covered in isulation. The second time around the walls were more covered and there was an extra floor on top. Despite that I have several good overlaps, TBC does some very strange things (OK, things I find strange). Clearly, in the registration it is hard not to mix the concrete wall with the insulation (not for me, but for TBC). I'll keep on looking.



  • 4.  Re: Registration on regions instead of complete clouds

    Posted 06-29-2020 09:54

    Nico,

    Try exporting your region to LAS/LAZ format, the re-import it into TBC. This would effectively import as its own scan and allow you to use the registration tool on it.



  • 5.  RE: Re: Registration on regions instead of complete clouds

    Posted 02-16-2024 13:25

    This should work. I had a situation where there were 3 houses surrounding by large mature trees. Did a number of scan stations, but TBC couldn't register them correctly. It was painfully obvious how the houses overlapped (at least the upper half of each house was in each scan) and I picked obvious registration pairs, but I think it was looking at the millions of leaves and pairing them instead, overriding my points. But I didn't think of classifying the homes alone, exporting then importing them, then registering the original scans on the imported house scans. Great idea - thanks.



    ------------------------------
    david silverman
    ------------------------------



  • 6.  Re: Registration on regions instead of complete clouds

    Posted 06-29-2020 17:11

    Hi Nico, since the condition has changed from your initial scan to the current scan, it could be challenging for TBC to see the "common" plane. You can assist TBC by manually picking point pairs from the true common planes, i.e., the partial wall covered in insulation in both scans. The point pairs don't need to be at corners or on edges, they just need to be roughly at the same location in the plane. 

    Below is a typical scenario. Usually we would think the pier wall would be a good choice. But they are not the same actually. One is scanned from the left side, one is from the right side. In your case, it would be one with insulation and one without. So I would pick a point from a common plane, even though it's smaller than the ideal wall.