Fitting a Sphere to Data

I recently had some data and wanted to fit a sphere through it so that I could find the radius of this sphere. As I started writing out an objective function for one of the Optimization Toolbox optimizers (yes I was taking the way too big a hammer approach), a quick query on the File Exchange brought up sphereFit, a pleasant discovery.

After downloading the files, I looked at Alan's published example, it looked like it would work!

Let's see it in action:

% Load mri of a human head
S = load('mri');
D = squeeze(S.D);

% Generate data points on the edge of the head
for ii = size(D,3):-1:1;
    % Perimeter of convex hull of each slice
    M(:,:,ii) = bwperim(bwconvhull(D(:,:,ii)>50));
end

% Find sub indices in our mask
idx = find(M); % find points in M
[rr,cc,pp] = ind2sub(size(M),idx); % sub indices
pp = pp.*floor(size(D,1)./size(D,3)); % rescale third dimension

% View data
figure;
scatter3(rr,cc,pp,10,pp);
daspect([1,1,1]);
view(-121,36);
axis tight;

% Fit the Sphere:
[cent,radius] = sphereFit([rr, cc, pp]);
fprintf(1,'\nRadius of sphere is %3.1f\nIt is centered at [%3.1f %3.1f %3.1f]\n',radius,cent);

Radius of sphere is 54.3
It is centered at [72.9 63.4 51.7]

Use Alan's example code to show the sphere through the points

scatter3(cc,rr,pp,25,pp,'*'); %points
hold on; % hold
daspect([1,1,1]); % equal axis so a sphere looks like a sphere
[Base_rr,Base_cc,Base_pp] = sphere(20);
surf(radius*Base_rr+cent(2),...
     radius*Base_cc+cent(1),...
     radius*Base_pp+cent(3),'faceAlpha',0.3,'Facecolor','m')
axis tight;
view(-121,36);

Comments

Give it a try and let us know what you think here or leave a comment for Alan.

Get the MATLAB code

Published with MATLAB® R2013b