6

I am wondering if there is any way to create a model that can be used in a RANSAC scheme where a spline or polyline could be determined from a noisy 3D point cloud.

What I have is a volume containing a set of points in each XY-plane, having say 400 of these in the Z-direction and a spline shape emerges visually while simply looking at the plotted data in 3D space. I already have a line fitting RANSAC scheme where the error is determined by a simple distance measure but I can't get my head around how to determine a model for a spline/polyline.

Milan
  • 1,743
  • 2
  • 13
  • 36
tompish
  • 79
  • 3
  • So the points are an contour of a 3D surface ??? sorry but my English isn't worth much and that is the only case I can think of from your description. Image would be best I think. – Spektre Jan 08 '14 at 10:41

1 Answers1

0

Here is a link to a quite interesting optimization method which, given a point cloud, computes a 3D spline surface which covers it. First, check out the videos:

http://users.acin.tuwien.ac.at/tmoerwald/?site=5

Here is the homepage of the project (I think):

http://www.acin.tuwien.ac.at/?id=316

The source code is also available within the PCL (PointClouds Library) project.

I have been reading up on this topic myself a bit, and it seems that the general problem of nurbs/spline fitting gravitates around finding out what is the best knot vector. Most approaches alternate between optimizing the knot vector (assuming the control points are fixed) and optimizing the control points (assuming the knot vector is fixed) and using the summed distance of the points from the cloud to the approximated spline as error metric. I think RANSAC would not fit well to these methods, because it would "drag the spline" towards the outlier points, but this is only my wild guess.

Violin Yanev
  • 1,507
  • 2
  • 16
  • 23