Hull from point cloud

Hi,
I've a problem with a hull design. I've a cloud of point describing my hull sections. I would like to model it in 3d to do some calculation on it.
If I put every point in Maxsurf as a marker and then try to get trough each point I know that it is possible but it's very long.
I also have Rhino3d. I thought about doing a curve going through a section on then doing a surface on it. But in this way the constructed curve are very unfair because of the number of points. Does any one have an idea how it could be working?
I've seen that there is a function _CurvatureGraph but I don't manage to control the error. Does any one know how this work?
Thank you
Wenc

[DaveB]

Hi,

What you're tryin' to do isn't always easy, but depends on the type of hull you're tryin' to model... Hard chined hulls tend to be easier in my experience (mind you I might not be doin' it right...)

If you post your offsets (data pts) people might be able to give you better guidance...

Best regards,

Dave

[Andrew Mason]

The Fit to Markers command in Maxsurf will fit a surface to a point cloud, however you need to have bought the Prefit module for this function to be enabled in your system.

[dereksireci]

Unigraphics will build a surface directly from a point cloud, either on screen or from a .dat file. I've seen it used once and there were some difficulties, but they finally made it go.

[Andrew Mason]

The question is not whether a particular program will create a surface from a point cloud (several programs will do this), but what the quality of the finished surface will be. Will it be good enough to build from? Will it need further fairing or modification - if so, will the control point net produced be simple enough or sufficiently well organised to allow you to do this.

If your intention is just to create a surface model so you can do hydrostatics or maybe do a pretty rendering, surface quality and control point net organisation are not important. If this is a design that is going to be built, the fitted surface needs to be of high quality.

This is why we adopted the approach we have with the Fit Surface to Markers command, even when we had a fully functional surface fitting program in Prefit. The fitting in Prefit is a lot like the skinning functions in programs such as Rhino, CATIA, ProE ect., nice for producing a reasonable looking surface, OK for doing hydrostatics, but not much use for creating production quality surfaces.

By treating the fitting problem as a constrained optimisation task, with control point fairness, regularity and orthogonality as constraints, it is possible to create a surface that is a least squares fit to the point cloud while still retaining a usable control point net.

Hi
Thank you for your replies.
I'm sorry I'm not allowed to share my offset file.

The purpose of this is to do some calculation on the hull but not only hydrostatics. My biggest problem is to be able to control the approximation error that I'm doing.
I'll try with maxsurf if I can find someone who has the module. I think it might be possible.
Andrew could you explain me what you mean with:
" By treating the fitting problem as a constrained optimisation task, with control point fairness, regularity and orthogonality as constraints, it is possible to create a surface that is a least squares fit to the point cloud while still retaining a usable control point net."
Because my English is not good enough...but it sound interesting.
Thanks
Wenc

[Andrew Mason]

Wenc

Most of the time when a surface is fitted to a set of data points, the objective of the fit is least squares error. Unfortunately this does not take into account the fairness of the surface. For example you could potentially have two surfaces that passed precisely through the hull sections you are using as original data, but one could be quite fair longitudinally and the other could have longitudinal inflections and be quite unfair. However, to the fitting algorithm they would have the same RMS error and be judged as being equally good.

Similarly, two surfaces may have the same level of error, but one may be produced by a really poorly arranged control point net, while the other may be produced by a very regular, evenly spaced control point net.

What we have done with the Fit to Markers command is used an optimisation procedure that tries to minimise the error while also trying to keep the control point net, and consequently the surface, as fair and as regularly spaced as possible. The result is a fair surface that fits well to the data, and a control point net that is good enough to have been created by a human being, as in the images below.

Andrew

[wenc]

Hi Andrew,

Thanks for the explication.
I've manage to find the prefit module. I've been trying to use it. I think I've understood how it works.
Now I want to use my offset datas. What kind of organisation should I give to the data in my txt file ?
Thanks
Wenc

[Andrew Mason]

The Prefit program requires that offset data be ordered by station position, and for each station the offsets should be in order (i.e. starting at the centerline and finishing at the sheerline). It is important that all stations have their offsets ordered the same way, for example if one station's offsets are ordered from centerline to sheerline and other station's offsets are ordered from sheerline to centerline the surface will try to twist along its length.