The code implemented a special case of the ellipsoid fitting technique proposed in the paper Least Squares Ellipsoid Specific Fitting, corresponding to case k= 4. Just put the two files in the same directory and run "testFit4". You can change the noise level to observe the fitting accuracy and efficiency of the algorithm. It works well when the underlying shape characterized by the data points is quite spherical. However, if this is not the case, some modification is required. For the technical detail of the fitting method, please refer to the above mentioned paper, which can be found from the following webpage:
Q LI (2020). Ellipsoid Fitting (https://www.mathworks.com/matlabcentral/fileexchange/23377-ellipsoid-fitting), MATLAB Central File Exchange. Retrieved .
How can I get the semiaxis from matlab?
The above link to the paper is broken. The paper for the algorithm can be found from ResearchGate at
or from IEEE Xplore at https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=1290055
Compared with the algorithm presented by Yury at https://uk.mathworks.com/matlabcentral/fileexchange/24693-ellipsoid-fit?s_tid=FX_rc1_behav, it is all about fitting accuracy. To compare the two algorithms, you have to use the same set of datasets.
If you find the fitted ellipsoid cannot be displayed properly, please change the line of code for creating the mesh grid in testFit4.m from
[x, y, z]=meshgrid(-a:stepA:a, -b:stepB:b, -c:stepC:c);
[x, y, z]=meshgrid(minX-stepA:stepA:maxX+stepA, minY-stepB:stepB:maxY+stepB, minZ-stepC:stepC:maxZ+stepC);
The code cannot be used in a general manner.
Why is the size of the points in the testfit (11,6)?
Should it not be 11,3 for xyz?
what is the meaning here?
where I check your test, I find the values of the major, medium and short radius from the equation u^3 - I*u^2 + J * u - K = 0 have some problems...
Here I = a + b + c
J = ab+bc+ac-f^2 - g^2 - h^2
K = |a h g; h b f; g f c|;
from "least squares ellipsoid specific fitting" GMP-04