Laser Import missing detail

Development Asset Creation
,
1

I have a laser scan that I am importing as e57. It lets me import it, align it and create a mesh but it seems to miss a lot of detail.

I import it as, exact registration. Intensity, noise free. I’ve changed all of these to alternative options without any success.

Following a search on these forums I’ve tried,

  • Adjusting the reconstruction region. (going to a small region doesn’t help).
  • Changing the min distane between two verts.
  • Changing Min distance between two points.

All without any variation of what comes through in the model.

In cloud compare the scan looks complete.

 

In RC most of the trees disappear, cars aren’t visible.

Any suggestions?

 

 

2

Hey Nick,

just to be sure - you did import it as ordered (one file for each station)?

Did you simplify and/or look at the model in a diffeent software?

3

Hi Götz,

Unstructured point cloud from a Yellowscan laser attached to a UAV. I collected the 6 scan files from this and combined them in Recap. I then exported it as a single structured file (e57).

Then view this e57 in cloud compare (first image). Or import it to RC (second image). 

 

 

4

Hi Nick,

so there we have it - you need to ex-/import the scans individually.

RC converts them into 6 images arranged in a cube at the position of the scan. So if you do only one, there will be only one cube and hence a lot will be omitted.

5

Hi Götz,

 

Thank you for the response. I don’t think that’s the problem though.

 

I followed a workflow that had me saving each laser scan as an e57 individually. Then bringing those 6 .e57 files into RC.

 

6x e57 files imported into RC result (looking at the whole site this time, my previous post was just an area of interest):

Compare this to the bulk e57 which has all the scans put together in Recap.

 

It gives the same result. After alignment is complete using the bulk e57 RC shows 6 camera locations. So its not trying limiting it to a single cube.

 

Compared to what the e.57 looks like in cloud compare. 

 

Any other Ideas?

6

Hi Nick,

weird, I got your response by mail but can’t see it here!

Ok, so I misunderstood what you said. The main thing is that they are individual scans and ordered.

Sure you can send me a scan (preferably the RC converted files since they are much smaller usually). You can find my mail address by searching my name - it’s pretty unique. :slight_smile:

But I’m not sure there is too much I can do though. Have you tried to render the pointcloud to make sure it’s not just a display issue?

Another thought - the missing parts seem to be the darker ones in the cloud rendered in CC, which I presume is the intensity. Are there any RGB values in the files at all? because RC needs those to reconstruct the pointcloud. It doesn’t use the points directly but goes through the images. The mesh is then reconstructed in the usual way by projection.

7

Hi Götz,

Seems my workflow of going from unstructured point cloud to structured .e57 using Recap isn’t a clean method. 

Post from a year ago with someone else trying the same thing.

 https://support.capturingreality.com/hc/en-us/community/posts/115000784951-E57-PTS-formats

So looks like i’m out of luck. Ground based LIDAR only, rules out any UAV laser work. I guess one day RC will implement the ability to use unstructured point clouds. Fow now I’m stuck with an alternative program Z that is slower but does support it. :frowning:

8

Hi Nick.

aaah, I missed UAV!

If it is possible to set an artificial center of the pointcloud, it might be worth trying.

If you could do that repeatedly, chosing the positions like you would with a ground based laserscan, I think it might work.

Can I contact you about the program Z? Or is it possible to circumscribe it a bit more?  :wink:

 

9

Hi Nick, Gotz,

I’m running into the same problem 3 years later. Were you able to find a good solution? May I ask what program Z is?

Thanks.