Determine outlne of dataset points

Hi, I'm trying to get a curve that best traces the outer layer of the points in the attached file orproduces a close fit of the outline. I've tried polyfit and isqcurvefit with the former being the best but this barely works as my data set is not a true second order system. I've tried interpolating polynomials but this hasn't worked or I might be implementing it wrong. Any guidance would be appreciated. Thanks

[Siyu Guo]

I'm not very sure about what you are willing to mean by the "outer layer". If you mean a rough shape containing most of the data points and meanwhile its boundary does not deviate from the data points "too greatly", I guess a least squares fitting should be able to provide a good start point. If you mean a polygon containing "all" of the points while the deviations of the points from the polygon are small enough, then maybe you can first find the convex hull of the points, then use polygon edge splitting to approach a polygon that fits the data better.

[Walter Roberson]

Perhaps boundary()

Answer 1

[Steven Lord]

As Walter suggested, boundary is one option. Depending how tightly you want the boundary to surround the points, alphaShape or convhull are other options.

If you want a mathematical equation instead of a computational geometry representation, assuming you know the general form of the equation you could try regression.