Plane fitting using 3D points

I have my 3D points(lidar data) as a text file. can anyone suggest me a method for fitting the plane.

[KSSV]

Read this: https://in.mathworks.com/help/curvefit/fit.html

Answer 1

[Star Strider]

Try this:

P = bsxfun(@times, rand(49, 3), [1 10 100]);                            % Create Matrix [x, y, z]

B = [P(:,1), P(:,2), ones(size(P,1),1)] \ P(:,3);                       % Linear Regression
xv = [min(P(:,1)) max(P(:,1))];
yv = [min(P(:,2)) max(P(:,2))];
zv = [xv(:), yv(:), ones(2,1)] * B;                                     % Calculate Regression Plane

figure
stem3(P(:,1), P(:,2), P(:,3), '.')
hold on
patch([min(xv) min(xv) max(xv) max(xv)], [min(yv) max(yv) max(yv) min(yv)], [min(zv) min(zv) max(zv) max(zv)], 'r', 'FaceAlpha',0.5)
hold off
grid on
xlabel('X')
ylabel('Y')

producing (with this set of random data):

[Ashicka Banu Mubarak]

Hello star strider,

I already have 3D points and I'm need of a multiple planes for those points and not a single plane considering all the points. Can you suggest me any codes sir?

[Star Strider]

I have no idea what you are asking.

If you have multiple sets of points ((Nx3) matrices), use my code for each set.

[Ashicka Banu Mubarak]

I'm doing research on 'Automated Building Extraction'. I have 3D point data for an urban region with no vegetation and ground points. Now the points I have is mostly on building roof. To extract each building roof separately, I need to fit separate planes on each building roof with the points that lie on the respective roofs. Since the points on each roof will be planar. I hope you understood my query. Can you help me to find a solution sir? Looking forward for your help sir.

[Star Strider]

Just use my code with the roof coordinates instead of the random data.

It will be best to do each plane of each roof separately.

[Ashicka Banu Mubarak]

lidarFile = '/Users/Ashi/Downloads/las_data/Subset/LIDAR_subset.txt';

fp = fopen(lidarFile, 'r');

ALS = fscanf(fp, '%f %f %f', [3 inf]);

fclose(fp);

ALS = ALS';

x=ALS(:,1);

y=ALS(:,2);

z =ALS(:,3);

% Create Matrix [x, y, z]

B = [ALS(:,1), ALS(:,2), ones(size(ALS,1),1)] \ ALS(:,3); % Linear Regression

xv = [min(ALS(:,1)) max(ALS(:,1))];

yv = [min(ALS(:,2)) max(ALS(:,2))];

zv = [xv(:), yv(:), ones(2,1)] * B; % Calculate Regression Plane

figure

stem3(ALS(:,1), ALS(:,2), ALS(:,3), '.')

hold on

patch([min(xv) min(xv) max(xv) max(xv)], [min(yv) max(yv) max(yv) min(yv)], [min(zv) min(zv) max(zv) max(zv)], 'r', 'FaceAlpha',0.5)

hold off

grid on

xlabel('X')

ylabel('Y')

This is the code I'm using for my points. I have included my result image and the points I'm using and the original orthoimage.

[Ashicka Banu Mubarak]

Orthoimage

[VIGNESH BALAJI]

@Star Strider When I use your method, the plane fit for a 3D dataset is not a good fit, maybe because instead of regression you can calculate mean of the points and find a basis of the plane and fit the plane to the data.

When I used Plane Fit (Affine fit function to fit the data) https://nl.mathworks.com/matlabcentral/fileexchange/43305-plane-fit . I got a better fit