How to fit a 3D circular surface?

[Thang Nguyen]

Hi,

 

I have data from laser scan of a circular surface. I converted data to image and using circular edge detection to mask out the background. What I like to do next is to fit the surface then subtract my original data from it. I am not sure if there is any function I can use to fit specifically in the boundary of the object only. Please let me know what function that I should use.

Thanks you.

Edited June 1, 2015 by Thang Nguyen

[infinitenothing]

Something like this?: http://zone.ni.com/reference/en-XX/help/370281P-01/imaqvision/imaq_fit_circle_2/

[Thang Nguyen]

I am sorry, I was not clear but I mean a 3D scan of circular surface, not a circular fit.

[infinitenothing]

You might have to show us some pictures. It sounded like you had some edge detection which should give you points that make a circle on the edge of the cylinder. You can use a circle fit to get the radius which will give you an area. If you need, you can take an average inside the circle to get the height of the cylinder.

Edited June 1, 2015 by infinitenothing

[Thang Nguyen]

@infinitenothing: Thanks for your reply.

One of the issue is the surface can be tilt, so I cannot use the average value.

I also have the case the cylinder diameter is greater than the range of the laser scan which I need to do 2 scan. In this case, I have to fit two half circle before I can stitch them together. To generalize my question, I am looking for a method to fit a 3D surface with some certain shape.

[infinitenothing]

A tilted circle projects to an ellipse right?

http://zone.ni.com/reference/en-XX/help/370281M-01/imaqvision/imaq_fit_ellipse_2/

 

Not sure about the general solution but maybe something involving this solver

http://zone.ni.com/reference/en-XX/help/371361J-01/gmath/find_all_minima_nd/

[Thang Nguyen]

@ininitenothing: I would like to fit the plane not the boundary. Thanks.

[infinitenothing]

That makes sense. Let me throw out an idea:

 

So you're trying to find the coefficients in this formula: Z=a₁x+a₂y+C. You might be able to use this: http://zone.ni.com/reference/en-XX/help/371361H-01/gmath/general_ls_linear_fit/

 

Where Y is an array of all the pixel intensities and H is {x1,y1,c}, ...

 

You might want to use real world coordinates instead of pixel coordinates.

Edited June 2, 2015 by infinitenothing