first time RC user

Comments

6 comments

  • Avatar
    Götz Echtenacher
    Hi James,

    you got my complete sympathies! I agree with you about the lack of RMB - I still do it sometimes and I feel like a finger is missing... :twisted:

    1. Yes, importing Ground Control Points works very well - I do that with a theodolite and get sub-mm accuracies if the alignment is good. However, that is all superfluous once you work with a laserscan - this ist the reason for RC's existence (to my knowledge). You can import laserscans with intensity, but to afaik only RGB works for aligning with images.
    Align laser scan with images

    2. Yes, it is also the Alignment button! :D The Workflow is quite different. It will optimize with repeated runs of alignment, sometimes signifficantly (trial and error). There is also no gradual selection. You need to first align with a rather high reprojection error (Alignment Settings) and then gradually lower it until your desired mean or median error is reached or the component breaks up.

    3. If you find that out, let me know!
    0
    Comment actions Permalink
  • Avatar
    James White GBG
    Hi Gotz,

    Thanks for the sympathy and information!

    At least it is refreshing to see a completely different approach, and i'm sure it will offer new opportunities.

    I shall just have to accept I need to learn this one from the beginning!

    Thanks again!

    James
    0
    Comment actions Permalink
  • Avatar
    Götz Echtenacher
    Hey James,

    NP - we were all there once! :D
    If you got the basics and get stuck, just shout.
    Although many issues are covered here already.

    About the laserscan:
    Not sure if that is actually important but it just struck me early this morning that I think RC does not use the pointcloud at all but creates 6 images from each station of the scan, arranges them in a cube (you can see that in 3D) and then treats this like an image, probably with a higher weight since the depth map is already 100% accurate. This is only speculation though and might as well be complete nonsense... 8-)
    0
    Comment actions Permalink
  • Avatar
    James White GBG
    From what i could make out from the posts i searched on here and the documentation i could find, and reading between all the lines, those 6 images per scan are used to find matches with the photographic images, and the full dense laser scan 3d data is preserved in there somewhere for reconstruction.

    I would be interested to know how it chooses between laser scan points and photogrammetric points for reconstruction when the two don't match up, i.e. scanning at different times, doors being open/closed etc, but i hope to figure that all out eventually and share on here when I can.

    Or maybe as you say, it doesn't use it at all. Which would be disappointing!

    Right now i'm up to my ears in a couple of large projects (one has been going for 12 months with a couple more months to go yet!) so can't dedicate too much time to the transition as it's already overrun somewhat! I'm planning to try using RC on smaller projects as they come in though.

    I would think the RC team would like to poach some other software users, so perhaps a specific guide to the parallels and differences would really help everyone.

    But so far I have spent a grand total of about 2 hours looking at RC (mostly the progress screen, and searching the forum while it was at it) so it's a bit early to come to any conclusions about anything yet :)

    I will need to explain to the accounts department where their 99EUR went at the end of the month though, so I'd better keep at it!
    0
    Comment actions Permalink
  • Avatar
    chris
    I'd suggest doing the laser scan alignment in something else. and importing it as exact.

    I've only tried laser scan data that had colour information, rc aligns based on colour. so not sure how well it will match photos and laser scan without that. but its probably fine with tie points.

    I was getting pretty good results matching photos to laserscans.

    but rc itself struggles with doing the laserscan by itself.
    0
    Comment actions Permalink
  • Avatar
    Götz Echtenacher
    Hey James, not to worry, you have 3 months time! Hehe...

    Well, I think it would make sense to use the laserscans in the shape of depth maps with a higher weight.
    The technology would already be there, no need to combine 2 different ones.
    And I don't see why that would be disappointing since a laserscan is 1. not always perfect (in the details) and 2. the depth maps would reflect the precision 1:1. So the spacial information is all there. In the end, it does not matter how it is transformed, just how well it is done.
    And then the answer to your question would be the same as to look at how it deals with alignments. If everything works out (which it should if the laserscan is good and also the images), then there are no issues. If not then not. And then you just need to look at how RC deals with modeling misaligned areas. Which is that it just latches onto a different alignment at some point when some threshold is reached, which results either in a step in the model (with larger misalignments) or noise (with small ones).

    Someone please correct me if my ideas are completely off...
    0
    Comment actions Permalink

Please sign in to leave a comment.