Best fit cylinder to survey points

[Anonymous]

Exciting times, first ever post on this forum.

Was hoping someone with more ACAD knowledge could help me out...

Problem: I'm a surveyor and majority of jobs I do is at a refinery where we survey pipes using a total station. With some of these pipes being out of reach, I'm wanting to survey random points anywhere on the surface of this pipe and then do a best-fit cylinder in civil 3d to these surveyed points (in doing so I'll have a 3d pipe to work with in ACAD). 

I would be very appreciative of any help, please let me know if you need me to go into more detail. 

[Sinc]

I created something like that a few years back, but it was done as a special commission for a client in Australia.  Feel free to contact me if you want at "sinc at quuxsoft dot com", and maybe we can work something out, but I have to get permission from the client who commissioned the original creation of the code before I can share it.

[Buzz12345]

I don't think you'll find anything out of the box in Civil3d, but you might want to check out the Point Cloud feature extraction tool which has recently graduated from Labs to a subscription Productivity Tool For C3D 2012 (see http://subscription.autodesk.com/sp/servlet/download/item?siteID=11564774&id=19407011).  I suspect that it will work best with lots of points (it's expecting data from a scanner).  I haven't got it installed, so I'm not sure if it does cylinder modelling.

I have trialled this when it was a labs product using data from a scanner, and I found it patchy at best - maybe it has improved. I had considerably more success with Trimble Realworks (But once again I was using data from a scanner which means a large set of points).

We have developed and used a methodology for looking at large tanks in Civil3D, but it requires the object to be near vertical.

Good luck

[AllenJessup]

The problem I see with using any kind of "Best Fit" method is that there will be areas where most of the random points will be along one exposed surface of the pipe and then in other areas most of the points will be on top. This type of unintentional weighting would bias a best fit in those directions. (I'm sure Sinc has figured out how to correct for this).

My first inclination would be to handle it in the field with distance and angle offsets to try and collect the CL of pipe as closely as possible. Then use a best fit on the CL points and use that for the CL of the cylinder.

Allen

[antoniovinci]

Well, in order to define a cylinder, you only need to survey 4 points: 3 on the beginning flange perimeter, and 1 on the ending one (to set the closing plane).

If you measure random points on the pipe surface, you should join them manually to get a watertight surface, then you could solidify all, but the resulting object will be far from a cylinder, I'd bet closer to a carrot...

[Anonymous]

If helpful, our team does offer a total station and/or point cloud based solution for AutoCAD to deal with cylinders/pipes in the field. I am not one for selling in forums but it will provide an option for you to get the job done.

Total Station: Our software connects AutoCAD directly to the total station in the field so your station acts as a mouse for AutoCAD. You draw to CAD with the total station. Our cylinder command allows you to aim and shoot points to derive pipes directly in CAD. The points are not exactly random as you need 3 points on either side of the cylinder. One point defines the starting point of the pipe and the other two help develop the diameter and orientation of the circle for one side. After doing this for both sides (six shots) a best fit axis and diameter are determined to form a 3D cylinder in AutoCAD.

Point Cloud: If you really want to shoot a series of random points along the cylinder than you will need a number shots per pipe. You can save your total station file as ASCII and use it with POINTCLOUDINDEX in AutoCAD to convert from ASCII to PCG (AutoCAD's point cloud format). We offer cylinder and catalog piping extraction from cloud data. Click two points on the pipe and a 3D solid will extract. The data will need enough points to resemble a cylinder of points in CAD at least from one side. Otherwise, there is not enough data for the algorithm to perform an extraction.

Scott, kubit USA

[Anonymous]

---

@Buzz12345 wrote:

We have developed and used a methodology for looking at large tanks in Civil3D, but it requires the object to be near vertical.

Good luck

---

Out of curiousity, what is your methodology for looking at large tanks? We are currently going through surveying deformation in large tanks, reporting on the deformation by obtaining cross sections at selected intervals up the tank.

[HansSMS]

We successfully use Trimble Realworks which has functions to fit cylinders and other geometric solids to point clouds. If the surveyor is using a Trimble VX  Robotic Total Station, this is the way to go. Also drapes images over the point clouds taken with the TSC3 controllers.

For simple operations like checking verticality of a vessel, we measure a bunch of points (30-50) around the bottom and again around the top, fit an arc to the points in Civil3D, and then compare the arc centres. The more points the more accurate. as the reflectorless shots are often only good to 5mm. With 50 points you get the accuracy down to about 1-2mm, depending on the geometric integrity of the vessel.

[Buzz12345]

strsi681 

Here's our methodology - it's been a while since I used it, so I'm working from a creaky memory here.

We essentially rolled the tank out on the ground and then created a dtm.

First of all get a best fit circle from all the points. Create an alignment from this circle. Create a station offset report for all the points relative to the bit fit circle alignment. Do some manipulation in excel for each point where you make x = station, y = elevation, z = offset. Reimport the points and create a dtm. Show it off with an elevation analysis and style.

It gives you a great visualisation of the tank compared to a vertical cyinder.

Please note we weren't scanning - we measured discrete points, so we didn't have a huge mass of point data. And I think where I say circle I mean an almost complete arc - because C3D does best fit arcs, not circles.

[Anonymous]

Thanks Buzz123 that solution was exactly what I was looking for and it definitely has us heading in the right direction.

The station offset report is only limited by the points needing to be cogo points. This means I have to downgrade a point cloud. 20,000 odd cogo points does a good job at slowing the computer down.

I'm going to do a bit of research to see if a point offset report can be manipulated to include autocad/pointcloud points. Here's hoping.

Thanks once again.