I have a $2D$ point cloud that I am trying to find the border of. I have created the convex hull of the points,
But I would like to find a better enclosing line segments or possibly a curve that defines the shape of this blob. I could convert these points to an image grid and use an image segmentation algorithm but I would lose accuracy and performance would be slow.
I worked with optimization before, I know machine learning algorithms, which approach could I use to define a better contour for this shape? One idea I had was iterating through all convex hull segments, from midpoint scan inward, if there are empty regions between segment and points, break segment into two create two segments each connecting to outermost point of the point cloud. I am basically trying to reach something like this (roughly jotted the left side only);
