Laser scaning to 3D model

Discussion in 'Software' started by kaptcatb, Aug 3, 2016.

  1. kaptcatb
    Joined: Jan 2013
    Posts: 39
    Likes: 1, Points: 8, Legacy Rep: 10
    Location: Bulgaria , Varna

    kaptcatb Junior Member

    Hello all,

    I work for company based on Europe. We do laser scanning on different civil buildings, temples, factories and ships.

    I'm looking for advice how do you proceed with point cloud, created from laser scans? Software should be capable to auto create model from point cloud data.

    Example: our last task was to scan ~150m ship, create model and do hydrostatic analys. What would be easier way to create model from point cloud?

    Best regards
    Todor
     
  2. TANSL
    Joined: Sep 2011
    Posts: 7,369
    Likes: 699, Points: 123, Legacy Rep: 300
    Location: Spain

    TANSL Senior Member

    Is it possible to obtain the scanned points in an orderly fashion ?. That is, can you get the points, for example, for each cross section, separated from the other sections ?.
    There are CAD naval programs which approximate NURBS surfaces to a given collection of points in space, but you must to make entries in an orderly manner. Still, the surface is very close to the real, but you must work hard to get exactly the actual surface. Which involves manual work carried out by experts. Naturally it depends on the accuracy you want to achieve.
    Reverse engineering is practiced for some time in shipbuilding but you´d tell us what you can get to give you a more specific answer.
     
  3. kaptcatb
    Joined: Jan 2013
    Posts: 39
    Likes: 1, Points: 8, Legacy Rep: 10
    Location: Bulgaria , Varna

    kaptcatb Junior Member

    Hi Ignacio and thanks for reply,

    We're looking for some automative process/plug-in/software, that would enable us to create surfaces ( ship shell ) from scans in easier way. Next step would be calculations and analys.

    We have specialized in larger vessels, but ship shell was not major part of our task in projects. Thats why I'm looking for some way to avoid manual work on surfaces.

    Hope now it's more clear,

    Best regards
    Todor Penev

    PS: we have done some research: following products can be used in order to modify data from scans to 3d model --> Rhino ( plug-ins: PointCloud;Mesh2Surface;Mesh from points ) ; VRMesh ; Meshlab ; PointCloudGenerator ; Geomagic ; Rapidform. Have someone used them?
     
  4. b1ck0
    Joined: Mar 2010
    Posts: 128
    Likes: 3, Points: 18, Legacy Rep: 40
    Location: Varna/Hamburg

    b1ck0 Senior Member

    Hi Todor,

    I believe that trying to reverse engineer the hull surface in NURBS or whatever other surface format is complete waste of time. Instead what you should be looking for is to create a mesh from the point cloud (for example in stl format) and then make your calculations based on the mesh. This is what most of the 'certified' stability programs are doing anyway (they create a computational mesh from your surface).

    Best Regards,
    Vasil Yordanov
     
  5. TANSL
    Joined: Sep 2011
    Posts: 7,369
    Likes: 699, Points: 123, Legacy Rep: 300
    Location: Spain

    TANSL Senior Member

    Hi Todor,
    I have done that reverse engineering on occasion and have outsourced the same work to specialized companies and, in my opinion, it is impossible to make a sufficiently accurate work without manual work. That work should also be done by skilled personnel in the "smoothing" of forms, if you want the results of the hydrostatic calculations are correct.
    The problem is not to create a nurbs surface or a mesh, the problem is that "it" pass exactly where it should go.
    As an example I will say that MaxSurf can create surfaces from a few points, "markers", properly distributed in space.
    I needed to develop my own software, based on AutoCAD, to create 3D models because there was no software on the market that perform that work with the accuracy I needed.
    BR.
    Ignacio
     
  6. SukiSolo
    Joined: Dec 2012
    Posts: 1,269
    Likes: 27, Points: 0, Legacy Rep: 271
    Location: Hampshire UK

    SukiSolo Senior Member

    I have reverse engineered IGES data, surfaces and point cloud data, but nothing on the scale you mention. If you want a clean NURBS model, you'll probably need to manually edit it and recreate the geometry. Some programs will show the differences in dimensions between the new model and the original.
    If you were to recreate a model, it would be almost mandatory to obtain a baeline number of true measurements as a basis. This would also rule out any floating point errors or similar that might happen in computing it.

    Half the trick with point clouds is to remove the thousands of non critical points....been there done that. Straight NURB from mesh and vice versa is done by programs like Rhino but be aware it might take a long time and result in ultra complex surfaces which should be more straightforward geometry.

    Maybe some software specialists in reverse engineering have a more capable solution for your needs?
     
  7. ludesign
    Joined: Aug 2003
    Posts: 213
    Likes: 7, Points: 18, Legacy Rep: 60
    Location: Sweden

    ludesign Senior Member

    The key question is whether you scanner software is able to extract a few relevant point instead of the millions of irrelevant points you get from a point cloud. These millions are required to keep you from missing relevant information, such as edges, whereas most hulls have a fairly smooth and predictable shape in between. Once you have such features in the scanner software, it's usually fairly easy to extract good and usable NURBS data, at least in my experience.
     
  8. wet feet
    Joined: Nov 2004
    Posts: 1,391
    Likes: 434, Points: 83, Legacy Rep: 124
    Location: East Anglia,England

    wet feet Senior Member

    Around fifteen years ago I attended a course in using a laser scanner on a Faro arm and it was quite interesting.Unless things have changed a lot,I'm not sure it is the easiest way to document a ship surface.My concerns would be the need to stitch together data from a number of scans and the huge file size from the original scan data.I suspect the mid-body could still be modelled more rapidly by taking physical measurements and I can imagine the benefit of scanning local features as part of the process.Either way you are in for a lot of work.
     
  9. TANSL
    Joined: Sep 2011
    Posts: 7,369
    Likes: 699, Points: 123, Legacy Rep: 300
    Location: Spain

    TANSL Senior Member

    Probably the scanning technique has changed a lot.
    Points can be obtained in an indiscriminate way or can be obtained, for example, points on the same transverse plane, obtaining points coordinates of the various frames separately.
    On the other hand the memory has increased dramatically in computers and they are able to handle huge amounts of data with much more ease and agility.
    But the real problem that arises in the ships is to achieve surfaces exactly like the boat's surfaces. Most procedures get very similar surfaces, but not identical, to the boat. I mean that obtaining the coordinates of the points is extremely accurate but the algorithms that operate not creates as accurate surfaces. Therefore it is necessary always a "manual" work to get the correct shapes.
     
  10. kaptcatb
    Joined: Jan 2013
    Posts: 39
    Likes: 1, Points: 8, Legacy Rep: 10
    Location: Bulgaria , Varna

    kaptcatb Junior Member

    After we assembly all scans together in Scene, point cloud size is 210GB :p. With removing app. every 10 point in column/row of cloud, we succeed to reduce size to 1GB.
    The main problem is that we don't have much expirience with meshes/surfaces. We're looking for some advice what would be the easier way to convert point cloud to "something" that can be used in software for hydrostatic calc ( in this case we scan ship shell ). Other projects may include ER scanning and remodeling pipes, steel structures, equipments and etc.

    We also have Faro arm ( Freestyle ) but it suitable for scans in areas, where you can't do scan with bigger scanner - let say you want to scan pipes below raised floor in engine room.

    Best regards
    Todor
     
  11. TANSL
    Joined: Sep 2011
    Posts: 7,369
    Likes: 699, Points: 123, Legacy Rep: 300
    Location: Spain

    TANSL Senior Member

    Hi Todor,
    Can you send me a sample file, not too big, to see what I can do with my software?
     
  12. Ad Hoc
    Joined: Oct 2008
    Posts: 7,774
    Likes: 1,679, Points: 113, Legacy Rep: 2488
    Location: Japan

    Ad Hoc Naval Architect

    Ask these guys, they will most probably be able to help you.
     
  13. MikeDrummond
    Joined: Nov 2006
    Posts: 11
    Likes: 4, Points: 3, Legacy Rep: 10
    Location: New Zealand

    MikeDrummond Junior Member

    Todor,

    I have modelled hulls by fitting NURBS surfaces to arbitrary 3D points by a least squares method. The surface does not pass exactly through the scanned points but the result is easily editable.

    In theory it is an automatic process, but with a large number of points and high accuracy it is painfully slow. Manual intervention can speed up the process dramatically.
    Eg delete points within a specified radius of others; review the hull shape for parallel sections, use weighting with important points.

    If the hull is a fair shape, very few points with a high order surface might be best; but if there are many hull features, you will need many points and use a low order surface.

    How many points are in the original and culled files?
    What is the allowable maximum error between a point and the surface?
    What is the allowable average error?

    Regards
    Mike
     
  14. TANSL
    Joined: Sep 2011
    Posts: 7,369
    Likes: 699, Points: 123, Legacy Rep: 300
    Location: Spain

    TANSL Senior Member

    What I gather, from all that the OP has told us, is that he does not need anyone to teach him or help him to get points in space using a 3D scanner. He knows how to do this perfectly and even knows how to apply filters to eliminate superfluous or redundant points. What he now needs is a real "loftman" that, with appropriate modern tools, generates a surface of the hull of sufficient accuracy to make reliable calculations of naval architecture.
    Moreover, I have to say I quite agree with the views of Mikedrummond.
     

  15. kaptcatb
    Joined: Jan 2013
    Posts: 39
    Likes: 1, Points: 8, Legacy Rep: 10
    Location: Bulgaria , Varna

    kaptcatb Junior Member

    Exactly

    Yes, I was thinking as laser scan colects data, it would be possible automaticaly to generate mesh ( with very good quiality ) for different engineer purposes. We tested trial versions of 2-3 software capable to convert point cloud to mesh, but they still require manual work on mesh or surfaces.

    I don't know what is amount of scan points, but mean error of scans points ( compared to reality ) is arround 0,5mm, which compared to ship length of 150m is nothing. I not sure what is allowed difference between point and surface.
     
Forum posts represent the experience, opinion, and view of individual users. Boat Design Net does not necessarily endorse nor share the view of each individual post.
When making potentially dangerous or financial decisions, always employ and consult appropriate professionals. Your circumstances or experience may be different.