Library to handle 3D geometric primitives: create, intersect, display, and make basic computations
Editor's Note: This file was selected as MATLAB Central Pick of the Week
The aim of geom3d library is to handle and visualize 3D geometric primitives such as points, lines, planes, polyhedra... It provides lowlevel functions for manipulating 3D geometric primitives, making easier the development of more complex geometric algorithms.
Some features of the library are:
 creation of various shapes (3D points, 3D lines, planes, polyhedra...) through an intuitive syntax. Ex:
createPlane(p1, p2, p3) to create a plane through 3 points.
 derivation of new shapes: intersection between 2 planes, intersection between a plane and a line, between a sphere and a line...
 functions for 3D polygons and polyhedra. Polyhedra use classical vertexfaces arrays (face array contain indices of vertices), and support faces with any number of vertices. Some basic models are provided (createOctaedron, createCubeoctaedron...), as well as some computation (like faceNormal or centroid)
 manipulation of planar transformation. Ex.:
ROT = createRotationOx(THETA);
P2 = transformPoint3d(P1, ROT);
 direct drawing of shapes with specialized functions. Clipping is performed automatically for unbounded shapes such as lines or rays. Ex:
drawPoint3d([50 50 25; 20 70 10], 'ro'); % draw some points
drawLine3d([X0 Y0 Z0 DX DY DZ]); % clip and draw straight line
The library is divided into two packages:
* geom3d for general computation in 3d
* meshes3d for representation and manipulation of polyhedral meshes
Several functions require the geom2d package, available on the FEx.
Additional help is provided in geom3d/Contents.m and meshes3d/Contents.m files, as well as summary files like 'points3d.m' or 'lines3d.m'.
1.20  remove deprecated functions, update some display functions 

1.19.2  fix bug in clipConvexPolyhedronHP, fix bug in nomalizePlane.m for multiple planes (thanks to Zubiao Xiong), add orientedBox3d 

1.19.1  fix bug in polyhedronCentroid 

1.19  Several updates for processing of meshes (intersectPlaneMesh, intersectLineSphere), for display of meshes, for computation and display of inertia ellipsoids 

1.18  added trimMesh function (that removes non connected vertices from a mesh), update mergeCoplanarFaces, and update tolerance management in interesectLineMesh3d 

1.17  fix various small bugs 

1.16  fixed bug in surfToMesh 

1.15  minor bug fixes, new functions for manipulating polygons, speed improvements (thanks to Sven Holcombe!) 

1.14  several bug fixes, new functions (fitLine3d, parallelPlane, reversePlane, intersectPlaneMesh, meshCentroid...), and fonctions for creating meshes from patches or geometric primitives. 

1.13  fix bugs in demos 

1.12  bug fixes, new functions for spherical polygons, for reading meshes (off foramt), for pointedge distance, and for computing surface area of 3D polygons, meshes, or ellipsoids. 

1.10  various code cleanup and speed improvements, mainly contributed by Sven Holcombe (many thanks!) 

1.9  use degrees for drawing shapes, split lib into packages 'geom3d' and meshes3d 

1.8  Add new functions for meshes and polyhedra, and for 3D transforms (rotations, basis transform). See file changelog.txt 

1.5  fixed bugs in rotations and in drawing of some shapes, added rotation by Euler angles, various update in doc and in code 

1.4  add missing demos 

1.3  some bug fixes, demo files as published mfiles. 
Inspired: Perspective projection, intersectPlaneSurf II, STL Lattice Generator, Optimal Step Nonrigid ICP, BrillouinzoneBCCLattice(), image ellipsoid 3D, Geometric measures in 2D/3D images, GJK algorithm distance of closest points in 3D
David Legland (view profile)
@Hannah: indeed, the cylinder is considered to be "open". If you need the intersection with the extremity, one workaround is to compute intersection of the line with the 3D plane containing the end of cylinder (in practice, you would need to consider 2 planes, one for each extremity). Then you can test if the distance between the intersection point and the cylinder main axis is less than the cylinder radius. I will try adding this behaviour in a next release.
Hannah (view profile)
Hi David,
I'm trying to calculate path lengths through a cylinder using the intersectLineCylinder. It is working very well until the line goes in the side of the cylinder but is angled enough to come out at the top. The output is only one set of coordinates. Is this because the top of the cylinder is not recognised as a surface? If this is the case, do you have any suggestions how I could work around this?
Many thanks, Hannah
Patrick Bolan (view profile)
Excellent! I've always gone back to VTK for 3d computational geometry (comparing MR spectroscopy with oblique MR imaging planes) but this toolkit makes it easy. Very clear, concise.
Miten Patel (view profile)
Hi David,
I'm trying to understand how to use geom3d library to translate a series of coplanar points onto a plane defined by a different series of coplanar points.
For example, if I have a series of points that define a unit circle on the YZ plane centred on 0, 0, 0 and I had a plane at a random 3D angle and random origin then I would like the 3D global coordinates that define the same unit circle on that plane.
As a test of my understanding, I set up a matrix 'z' of following 4 points z = [ 0 1 0; 0 0 1; 0 1 0; 0 0 1]. These are on the global plane. I want to translate them to the plane defined as [ 0.0544 0.0000 0.0000 0.0000 0.9959 0.0901 0.0000 0.0901 0.9959 ]. This is simplified test where destination plane is in the YZ plane and just translated along the X axis.
I used the following statements:
z = [ 0 1 0; 0 0 1; 0 1 0; 0 0 1];
lp = [ 0.0544 0.0000 0.0000 0.0000 0.9959 0.0901 0.0000 0.0901 0.9959 ];
t = createBasisTransform3d('global', lp)
transformPoint3d(z, t) gives
0.9959 0.0901 0.0544
0.0901 0.9959 0.0544
0.9959 0.0901 0.0544
0.0901 0.9959 0.0544
This is not what I expected to get. Firstly the destination plane is translated by 0.0544 on X axis whereas the new points have this translation reflected on the Z' coordinate.
Also, the resulting X', Y' coordinates are what I thought should be in the Y', Z' coordinates. Also, the rotation around the four points is counterclockwise (if X' > Y and Y' > Z) whereas the original points were in a clockwise rotation.
Can you advise where I am going wrong?
Many Thanks, Miten
Halil Ibrahim Kayim (view profile)
Hi David,
Thanks for this library, it's helping me a lot in my master's study. The only trouble I had with this awesome toolbox was with the mergeCoplanarFaces method. It seems you added the threshold variable in a later version, however I don't think it works as it should. After investigating how this threshold was used, I saw that it is used for two purposes. First one is to test the value of a cross product of two normal vectors. The second one is later used inside mergeCoplanarFaces when the candidate nodes are tested for coplanarity in the isCoplanar method. The trouble with this is that while the first threshold is only meaningful for values between 0 and 1, the second threshold can be any value greater than 0, and does not behave the same for geometries of different scale.
Now, I don't know how to go about solving this behavior, but it doesn't seem to break anything right now to just comment out the isCoplanar check inside mergeCoplanarFaces. This may be a horrible workaround and may ruin stuff in the future, so can you please have a look at this issue?
Thanks for this toolbox and for your troubles.
Best,
Halil
David Legland (view profile)
@ThT: I do not know why this behaviour is happening. Please send me a sample data set by email, and I will try to have a look.
@Leonardo: If I understand well you would like to compute intersection between a 3D triangle and a 3D polyline? Unfortunately there is no such function for this at the moment. I think compmuting intersection with line, and postprocessing the result could give some results, but this would require some coding efforts.
Regards,
David
Leonardo Colavitti (view profile)
Hi David,
thanks for the library and your efforts. I'm using the function TriangleRayIntersection of the package geom3D and I was wondering if in the input, instead of giving the direction of a straight line, it exists an implementation if my line is a slightly curve line and I know it in every point (basically, I have a vector in the 3 dimensions). I need this for a raytracing problem...
Thanks for your kindly help,
Best regards,
Leonardo
ThT (view profile)
Hi David,
Thanks for the library and your efforts. I have a question though regarding an issue that I am facing. I am loading a CAD model in matlab from an .stl file. This gives two matrices with the vertices and the faces of the model, already in triangles. However, the existing trianglulation is a kind of weird and for that reason I am trying to reformat it in a better way. Therefore, I am using the "mergeCoplanarFaces()" in order to create faces with more that 3 vertices and then retriangulate the new mesh in a more proper way. However, the output of the "mergeCoplanarFaces()" function gives me some multivertex faces with "NaN" points included and if I try to plot the new faces and vertices there are some gaps within the mesh, which were not existing in first place. Do you know what it might be happening, and if there is any workaround in order to avoid this behaviour.
Thanks.
Theo
David Legland (view profile)
Hi Lina,
The (dx,dy,dz) components of the line representation simply correspond to the differences in coordinates of the two 3D points.You can check the utility function "createLine3d", that encapsulates this process.
You can try the following example:
sphere = [40 50 60 30];
figure; drawSphere(sphere);
hold on; axis equal;axis([0 100 0 100 0 100]);
p1 = [50 50 50]; p2 = [51 52 53];
line = createLine3d(p1, p2); drawLine3d(line, 'b');
pts = intersectLineSphere(line, sphere);
drawPoint3d(pts, 'k*')
Otherwise, you can send me your data by email, I,can try to have a look;
Regards,
David
Tsvetelina Germanova (view profile)
Hi David,
I'm using the Intersect line sphere function but I seem to not obtain the proper angle of the intersection. I think I may be giving a wrong input for the dx dy dz component of the line (or at least I cannot find other explanation). Can you please give me an example of what would be the the dx dy and dz if I have a line defined by two points?
Thanks,
Lina
David Legland (view profile)
HI Annie,
I would suggest trying some functions from the geom2d package (also available on FEx):
* the "removeMultipleVertices" function merge adjacent vertices whose distance is small enough
* the "mergeClosePoints" merge points with distance small enough, without taking into consideration point order (ie, it is possible to merge point with not adjacent indices)
These two functions were designed for planar geometry, but should for for 3D as well.
Regards,
David
Annie Tran (view profile)
Hi David,
I used the intersectPlaneMesh and when calculating the intersection between a plane that coplanar with faces of the mesh, the output polygon contained repeated vertices. Could you suggest idea to eliminate the repeated vertices since it is helpful for me in further process.
HaT
Olabanji Shonibare (view profile)
Rose Ann Haft (view profile)
The lines, sphere intersect isn't very useful. The example is not easy to follow.
Li Zheng (view profile)
David Legland (view profile)
HI Alec,
you can check the sphericalAngle function, that computes angle between three points on a sphere. Maybe you will need to convert spherical coordinates.
If have used the principle to compute spherical angle of Voronoi domains on the sphere for another contribution (estimation of surface area in 3D images). I think you can grab some pieces of code from the source:
https://github.com/dlegland/matImage/blob/master/matImage/imMinkowski/private/sphericalVoronoiDomainArea.m
Regards!
Alec Day (view profile)
Hello, I'm trying to use this software (its great by the way) to calculate the area of a polygon stretched over a sphere. If I have a range of azimuth and elevation values that connect to form a polygon over a sphere, which function could i use to calculate its surface area (on say a unit sphere, or better yet, a sphere of radius r)?
Thanks!
Alec.
David Legland (view profile)
@Charles,
sorry for delay... Thanks for your suggestion. I do not use Doxygen, so maybe this will be difficult for me to apply this solution. But this should open ideas for scripting or automatising. Thanks!
David Legland (view profile)
@Chirag,
I do not know exactly where your problem can come from. Could you identify some specificities of the meshes causing troubles? Can you sent me by email a problematic mesh?
Some ways of investigation:
* maybe use the trimMesh function before computing may help.
* appy the triangulateFaces before computing smoothing
Regards,
David
David Legland (view profile)
Hi S SS, and sorry for the delay...
The intersectPlaneMesh function works for the moment only with infinite planes. One solution would be to project the intersection on the plane, to apply 2D intersection algorithms on the projection, and to project back to 3D. I am not sure this is quite simple however...
David Legland (view profile)
Hi Pauline,
the intersectPlaneMesh function requires for the moment a "closed" mesh, that is a mesh without free boudary. Torus and cube mesh fullfil the requirement, but if you apply on open meshes the error you obtained is expected. I will try to work on an enhanced version that manages also the case of open meshes.
Chirag Bhuva (view profile)
Hi David ,
I am using smoothMesh function,it is working good but sometimes it is not worked,the error is Inner matrix dimensions must agree.
Error in smoothMesh (line 55)
v2 = adj * v2;
Error in DemoMain (line 31)
[FV2.vertices, FV2.faces] = smoothMesh(FV.vertices, double(FV.faces));
because of adjacency matrix and vertices's size is not same .so could you please guide me.
Pauline Huet (view profile)
Thank you David for the code !
I am quite new to matlab/geom3D and I have some trouble with the intersectPlaneMesh function. I am trying to get the intersection of a mesh read from an stl file as faces and vertices (successfully displayed with drawMesh) with a plane. Unfortunately I always get the same error 'crossing edge 1 (4,5) is associated to 1 faces' whereas with standard privitives like cubes or torus all goes right. Any idea of what could have caused this error ? Thanks in advance
S SS (view profile)
Dear David, thank you very much for your submission. It is very useful for me.
I am using the intersect plane mesh function to have the intersections between planes and a mesh. Actually I have several intersections because the plane is infinite. I want to have only one intersection. For this I want to have a limited plane. Have you a solution for me? Thank you very much
Charles Brillon (view profile)
Dear David, I sent you a personal email (in french). Have you considered using this : https://www.mathworks.com/matlabcentral/fileexchange/25925usingdoxygenwithmatlab to generate documentation?
That might prove helpful.
Regards,
Charles
David Legland (view profile)
Hi Gary,
yes, keeping consistent numbering of version and doc is complicated... But I agree, this kinf of information is important. I'll try to enhance this, maybe this can be automated in some way.
regards,
David
Gary Kenward (view profile)
David:
NP. Glad I could help.
One additional note: you might want to update the version and date in the Contents file. I understand what a pain maintaining documentation can be, however, I just spent a half hour trying to figure out why I had a 2011 version of GEOM3D loaded into Matlab. It can be troublesome to suspect that you are using a very old version of a toolbox.
Also, you might want to update the version to 1.21 or such.
Anyway, great toolbox. Thanks for sharing it with us.
Gary
David Legland (view profile)
Hi Gary,
thank you for reporting the problem, and quickly providing the solution! I have updated a new version that uses your fix (also in other similar functions).
regards,
David
Gary Kenward (view profile)
I fixed the problem I encountered with fillPolygon3d with two changes, marked with the characters " %<======= " below. I may have missed something, but the changes appear to work.
function varargout = fillPolygon3d(varargin)
%FILLPOLYGON3D Fill a 3D polygon specified by a list of points
%
% fillPolygon3d(COORD, COLOR)
% packs coordinates in a single [N*3] array.
% COORD can also be a cell array of polygon, in this case each polygon is
% drawn using the same color.
%
% fillPolygon3d(PX, PY, PZ, COLOR)
% specify coordinates in separate arrays.
%
% fillPolygon3d(..., PARAM, VALUE)
% allow to specify some drawing parameter/value pairs as for the plot
% function.
%
% H = fillPolygon3d(...) also return a handle to the list of line objects.
%
% See Also:
% polygons3d, drawPolygon, drawPolyline3d
%
% 
% Author: David Legland
% email: david.legland@nantes.inra.fr
% Created: 20070105
% Copyright 2007 INRA  BIA PV Nantes  MIAJ JouyenJosas.
% check case we want to draw several curves, stored in a cell array
var = varargin{1};
if iscell(var)
hold on;
h = [];
for i = 1:length(var(:))
h = [h; fillPolygon3d(var{i}, varargin{2:end})]; %#ok<AGROW>
end
if nargout>0
varargout{1}=h;
end
return;
end
% extract curve coordinate
if size(var, 1)==1 %<======= check if var is 1 x N
% first argument contains x coord, second argument contains y coord
% and third one the z coord
px = var;
if length(varargin)<3
error('Wrong number of arguments in fillPolygon3d');
end
py = varargin{2};
pz = varargin{3};
varargin = varargin(4:end);
else
% first argument contains both coordinate
px = var(:, 1);
py = var(:, 2);
pz = var(:, 3);
varargin = varargin(2:end);
end
% extract color information
if isempty(varargin)
color = 'c';
else
color = varargin{1};
varargin = varargin(2:end);
end
% ensure polygon is closed
%<======= fill3 closes polygons if needed; code below generates concatenation dimension inconsistency error
% px = [px; px(1)];
% py = [py; py(1)];
% pz = [pz; pz(1)];
% fill the polygon
h = fill3(px, py, pz, color, varargin{:});
if nargout>0
varargout{1}=h;
end
Gary Kenward (view profile)
I'm having a problem with fillPolygon3d(PX, PY, PZ, COLOR), where PZ, PY, and PZ specify coordinates in separate arrays.
fillPolygon3d([0.4 0.6 0.5], [0.5 0.5 0.5], [0.2 0.2 0.4], 'blue');
Error using fill3
String argument is an unknown option.
Error in fillPolygon3d (line 76)
h = fill3(px, py, pz, color, varargin{:});
Any suggestions?
Binu (view profile)
David Legland (view profile)
Hi Dimitris,
unfortunatley there is for the moment no function for computing offset of 3D polygons or meshes. Depending on what you want to do, you can (1) project the 3D polygon on a plane, compute offset in 2D, then project the result back in 3D, or (2) compute an extrusion of the polygon, by duplicating the polygon, translating one of the two polygons by a given distance in the direction perpendicular to the supporting plane of the polygon, then associate the vertices to form a 3D mesh.
Dimitris Mylonas (view profile)
Hello how to offset polygons in 3d?
Manar Al Asad (view profile)
Hello,
I'm experiencing problems with the centroidrelated functions (calling nonexisting functions). Has anyone else encountered problems with them. If so, does anyone have a fix?
anusha gorrila (view profile)
Caleb Brown (view profile)
MANIK BANSAL (view profile)
Collin Stillman (view profile)
Wei Zhao (view profile)
David Legland (view profile)
Hi Anusha,
it is difficult to figure out what could be wrong. Usually, for 'closed' meshes (such as boundary representations), intersesctions with lines produce even number of intersections. The "pos" output corresponds to the relative position on the line. Negative value means the position is before the line origin. You can also investigate the "ind" output, that corresponds to index of intersecting face.
It is still possible there are some numerical computation errors, that cause apparition of unexpected points, but I did not encounter yet...
anusha gorrila (view profile)
Hello David, Thanks for such a wonderful code. I've been using the intersectLineMesh3d function for finding the intersection point on the mesh from a line and i get 2 intersection points and the position of the 1st one is a negative value and the position of 2nd one is a positive value. Could you explain the output as my line intersects the mesh only at one point. The output is something as below:
points =
536.1685 361.9645 40.1591
573.2433 399.9937 77.9874
pos =
28.0442
10.0410
ind =
19
201
Thanks,
Anusha.
David Legland (view profile)
Hi Hannah,
hmm, strange behaviour I agree... In the current version, the algorithm starts by triangulating the mesh. By consequent, each square face of the cube is transformed into two triangular faces. It the origin of the line is in the middle of the cube, it will intersect the two triangles corresponding to a single square.
By the way, if you use additional ouptput arguments, they will refer to the triangulated mesh! (I will add warning or checking in future version).
regards,
David
Hannah (view profile)
Thanks for this code, I'm finding it very useful. I just have one question, when I'm using intersectLineMesh3d, why do the intersection points repeat themselves, for example for a cube with a straight line through the middle you get 4 intersection points?
Many thanks, Hannah
David Legland (view profile)
Hi Sandy,
I'm glad you could find the way of using isPointInPolygon!
For other kind of intersections, you can consider 'intersectLineMesh3d', as well as 'intersectPlaneMesh' functions.
regards!
Grant Sandy (view profile)
Please disregard comment on August 5. I see now that the geom2D package is required. So far everything is working perfectly with the addition of geom2D functions.
Grant Sandy (view profile)
There seems to be a missing function called isPointInPolygon which is called by intersectLinePolygon3d. Is this something that was overlooked?
Grant Sandy (view profile)
Quite a treasure! I am using the line/plane intersection function now. I will also use the other line/surface intersection functions.
David Legland (view profile)
Hi Tariq,
yes, the inertiaEllipsoid function computes an equivalent ellipsoid from data, based on inertia of the input points.
You can get the centroids either from the first three columns of the result of inertiaEllipsoid, or by using the 'centroid' function from the geom2d package.
For computing surface area of a bounding box, I would suggest summing the area of each face. The area of each face can be obtained by multiplying the corresponding side lengths.
regards
tariq bdair (view profile)
Hi David,
Is the inertiaEllipsoid function calculte the best fit ellipse like in 2d data?
Also, how can i get the centroids for double data?
How can i calculate the surface area for bounding box?
Thank you
David Legland (view profile)
Hi Md Shadir,
You will need a function to parse the PLY file, and extract the vertex coordinates and vertex indices of each face. There are several tools in other libs, for example the graph toolbox of Gabriel Peyre (function 'read_ply'), or the computer vision toolbox (function 'pcread').
Note however that to my knowledge, PLY files provide only vertex coordinates. So you will need to compute the faces of your mesh. In this case, I would recommend using delaunay triangulation, and then try to remove non necessary tetraedra (this can be a complicated task however...), and extract their outer face.
regards,
Md Shakir Khan (view profile)
Hi David,
great submission. I have been looking into the "geom3d" and trying to implement the "intersectLineMesh3d" for my project. But I need to work with 3D mesh data (.ply file) and check for the intersection. The .ply (3D mesh) file contains element vertex as property floats x,y,z and element faces. Is there any way to feed the .ply file into your code and run the "intersectLineMesh3d" algorithm? In fact, I tried by directly copying the vertices and faces from the .ply file (as below) but got the error: Subscript indices must either be real positive integers or logicals.
vertices = [0.289329 0.230469 0.550781
0.285156 0.230469 0.543497
0.285156 0.229169 0.550781
0.277344 0.230469 0.549533
0.285156 0.229169 0.550781
0.285156 0.230469 0.543497
0.277344 0.230214 0.550781
0.285156 0.229169 0.550781
0.277344 0.230469 0.549533
0.276716 0.230469 0.550781]
However, it works well with the positive data.
Can you please help me out? I would really be very grateful to you. Thanks.
An example (part) of my data:
ply
format ascii 1.0
comment file created by Microsoft Kinect Fusion
element vertex 2959173
property float x
property float y
property float z
element face 986391
property list uchar int vertex_index
end_header
0.300781 0.625023 1.113281
0.296898 0.628906 1.113281
0.300781 0.628906 1.117165
0.695368 0.660156 1.496094
0.691406 0.659174 1.488281
0.691406 0.660156 1.491550
0.695368 0.660156 1.496094
0.699219 0.658886 1.496094
0.691406 0.659174 1.488281
..........................................................
..........................................................
..........................................................
3 2959146 2959147 2959148
3 2959149 2959150 2959151
3 2959152 2959153 2959154
3 2959155 2959156 2959157
3 2959158 2959159 2959160
3 2959161 2959162 2959163
3 2959164 2959165 2959166
3 2959167 2959168 2959169
3 2959170 2959171 2959172
Md Shakir Khan (view profile)
But I need to work with 3D mesh data (.ply file) and check for the intersection. The .ply (3D mesh) file contains element vertex as property floats x,y,z and element faces. Is there any way to feed the .ply file into your code and run the "intersectLineMesh3d" algorithm? In fact, I tried by directly copying the vertices and faces from the .ply file (as below) but got the error: Subscript indices must either be real positive integers or logicals.
vertices = [0.289329 0.230469 0.550781
0.285156 0.230469 0.543497
0.285156 0.229169 0.550781
0.277344 0.230469 0.549533
0.285156 0.229169 0.550781
0.285156 0.230469 0.543497
0.277344 0.230214 0.550781
0.285156 0.229169 0.550781
0.277344 0.230469 0.549533
0.276716 0.230469 0.550781]
However, it works well with the positive data.
Can you please help me out? I would really be very grateful to you. Thanks.
An example (part) of my data:
ply
format ascii 1.0
comment file created by Microsoft Kinect Fusion
element vertex 2959173
property float x
property float y
property float z
element face 986391
property list uchar int vertex_index
end_header
0.300781 0.625023 1.113281
0.296898 0.628906 1.113281
0.300781 0.628906 1.117165
0.695368 0.660156 1.496094
0.691406 0.659174 1.488281
0.691406 0.660156 1.491550
0.695368 0.660156 1.496094
0.699219 0.658886 1.496094
0.691406 0.659174 1.488281
..........................................................
..........................................................
..........................................................
3 2959146 2959147 2959148
3 2959149 2959150 2959151
3 2959152 2959153 2959154
3 2959155 2959156 2959157
3 2959158 2959159 2959160
3 2959161 2959162 2959163
3 2959164 2959165 2959166
3 2959167 2959168 2959169
3 2959170 2959171 2959172
Sleh Eddine Brika (view profile)
Niels Schroeter (view profile)
David Legland (view profile)
Hi Andy,
I am not sur what you want to do exactly; one possiblity is to identify faces that are located within a polygonal mesh. In this case, the you will need to detect if all vertices are within the mesh. There is for the moment no such function in geom3d, but I think there is on FEx. Another possibility is to display faces with different transparencies. In that case I think the simpler is to manipulate direclty the graphical handles of the meshes.
Hope this helps ?
Andreas Horn (view profile)
Dear David, this is a really awesome library!
However, I'm specifically looking for a function that removes internal faces in a complex patch (exactly the functionality as described here: http://meshlabstuff.blogspot.com/2009/04/howtoremoveinternalfaceswith.html). Do you have any idea how I could do this within Matlab?
Thanks so much for any hint!
Andy
dingkun lian (view profile)
dingkun lian (view profile)
dingkun lian (view profile)
David Legland (view profile)
Hi Weihua,
Extracting the position of 3D pixels in a 3D image is not complicated (usign for example the find function), for managing self occlusion there is no straightforward solution in geom3d... One possible solution would be to reconstruct the set of 3D position (using for example marching cubes), and then consider intersections of reconstructed structuers with rays parallel to the direction of the view. By sorting intersections with respect to relative position on ray, a virtual view of the structrues could be obtained. This could be rather computerintensive however.
UTA (view profile)
Hi David
Thank you for your code. I also interesting to extract the pixels positions of the 3D object which in a specific view direction. That is to say, I want to get the 3D pixels locations after consider the selfOcclusion. Would you please give me some help on this?
Thanks ahead!
weihua
David Legland (view profile)
Hi Alex,
unfortunately there is for the moment no such functionality if the geom3D toolbox. But you may check the following (very recent !) submission: http://www.mathworks.com/matlabcentral/fileexchange/55803surfacesintersect
If you transform a 3D polygon into a set of triangular faces, I suppose you can use "surface interesect" for computing intersection points.
Regards,
Alex Lifshitz (view profile)
Hi David,
Thank you for the explanation and correcting the code.
I have another question. Is it possible somehow to clip a 3d polygon by another shape, for example a cylinder rather than just a plane?
Thanks
Alex
David Legland (view profile)
Hi Alex,
The computation of the volume of a mesh is correct: you need to consider signed volume to take into account non convex meshes. You can try with a torus for example... As a quick workaround, you can consider using abs(det), but the function will return correct results only for convex (or starshaped with respect to origin, I guess) meshes.
Actually, the bug was in the "clipConvexPolyhedronHP" function. A new face is created during the process, and its orientation was not kept consistent with the rest of the mesh. I have updated the package, that should solve the pbm.
regards,
Alex Lifshitz (view profile)
Following my previous comment below it seems that there is a bug in the implementation of meshVolume function.
The function divides the volume into tetrahedra. Then the total volume equals to the sum of the volumes of all tetrahedra. The code computes the volume by calculating the determinant of the vertices.
vols(i) = det(tetra) / 6;
It turns out that sometimes this determinant can be negative due to a different order of the rows (vertices coordinates).
The correct formula should include absolute value of the determinant.
vols(i) = abs(det(tetra)) / 6;
That solves the problem.
Alex Lifshitz (view profile)
Hi
I am interested to calculate a volume of the ellipsoid after slicing it with an arbitrary plane. I am using the following code (here for simplicity I will use a sphere and cut it in the center).
[x,y,z] = sphere(30);
[f,v,c]=surf2patch(x,y,z,z,'triangles');
meshVolume(v, f)
This calculates a volume of a sphere as expected (4/3*pi)
Now, I slice it in half.
P = createPlane([0,0,0], [0, 0, 1]);
[v1 f1] = clipConvexPolyhedronHP(v, f, P);
Now if I calculate the volume, I get incorrect value. If I draw it, I see that the reason for this is that the surface along which the sphere was cut, remains open.
drawMesh(v1,f1,'facealpha',0.5)
My question is how can I close the shape such that the volume can be calculated correctly.
Thanks!
David Legland (view profile)
Hi Michael,
sorry, there was a bug in the polyhedronCentroid function... I have just submitted a revised version that should solve this.
David Legland (view profile)
Hi Umair,
to compute the volume, you need to specify how the vertices are connected. There is no "simple" method, but you can search for alpha shapes for example.
Umair Bin Asim (view profile)
Is there any way here, with which we can find the volume of an arbitrary 3D shape defined by a set of 3D points?
Michael Strohmeier (view profile)
Hi David,
thank you for your answer.
I have to bother you with another issue.
Wenn i compile polyhedronCentroid.m i get the error 'Undefined function or variable "vt".' Where do i get the data for "vt"?
Thank you in advance,
Michael
David Legland (view profile)
Hi Michael,
unfortunately I think this will not be possible... There is already a vectorization procedure for managing all faces of the mesh in a global way. Then it seems difficult to process also lines in a vectorized way.
Tristan Chaplin (view profile)
Hi David,
Thank you very much, it works just as I'd hoped.
Thanks,
Tristan
Michael Strohmeier (view profile)
Hi David,
your library is very useful! Thank you for sharing.
I am using intersectLineMesh3d and have to loop the function for several lines in order to get all intersection points.
I was wondering if it is possible to modify the code in order to use multiple lines...like you did for intersectLineSphere recently (last comment of Tristan Chaplin).
This would be very helpful since a loop around the function is too slow for my purpose.
Thank you very much!
Michael
David Legland (view profile)
Hi Tristan,
Yes this is possible! I just updated a new version, with a modified version of the intersectLineSphere function. There is an example in the header of the function that demonstrates how to use it with multiple lines.
regards!
David
Tristan Chaplin (view profile)
Hi David,
Thanks for this library, it has been very useful for me.
I was wondering, is there a away to use intersectLineSphere with multiple lines? Similar to the way intersectLinePlane works? It's too slow for my purposes to use a for loop and do each line one by one.
Thanks,
Tristan
David Legland (view profile)
Hi Dale,
in fact, both 'line' and 'ray' are infinite. If P0 is the origin point and V is the direction vector, both shapes represent a set of points satisfying P = t * V + P0. In the case of a line, t ranges from inf to +inf, whereas in the case of a ray, t ranges from 0 to +inf. Note also that the LINE or RAY variables represent only origin point and direction vector, and not the 'type' of the shape. The distinction between line or ray is made only by calling the appropriate function.
If you want to find intersection with a finite line (a "line segment"), you can get the position 't' of the intersection point on the line from the 'linePosition3d' function. If value is between 0 and 1, the points lies on the line segment comprised between P0 and P0+V.
hope this helps,
David
Dale Fried (view profile)
Hi David,
Thanks for sharing this great library. I have used it quite a bit already, and am now trying a new application.
I am missing what is the functional difference between intersectLinePolygon3d and intersectRayPolygon3d.
I would expect that the line function would understand a finite length, but the documentation just mentions an origin and a direction vector. Should that read, instead, "direction vector giving the length of the line segment"?
What am I misunderstanding?
Thanks.
Dale
Krisztian Szucher (view profile)
muna majeed (view profile)
Hi David,
Really great tool and very good! It helped me with a lot of problems. Thank you for sharing it with us.
please, I ask you can I crop or segment a mesh of type obj into part by this tools, please I need that
David Legland (view profile)
Hi Yiyu,
actually I modified this function yesterday also... I have used another approach, that makes it much faster, and it now returns multiple polygons as well. You can try the new version here: https://github.com/dlegland/matGeom/blob/master/matGeom/meshes3d/intersectPlaneMesh.m . Maybe you can compare with yours? If you have any advice/hint i'm still interested!
regards!
Yiyu Hong (view profile)
Hi David,
I modified your intersectPlaneMesh function code and make it work to find multiple contours.Do you need the code?
David Legland (view profile)
@Homa Rasouli:
for computing surface area of a polygonal mesh, you need an array with vertex coordinates, and an array that stores vertex indices of each face. Such arrays can be obtained from the 'isosurface' function, for example. Otherwise, you can read them from mesh file formats such as OFF or PLY.
Homa Rasouli (view profile)
I have problem in computing the surface of triangular mesh. The function works only with triangle mesh as an input. How should I compute the vertices?
David Legland (view profile)
@SM:
you're right, I use XYZ convention: the three angles correspond to extrinsic rotations performed around X, Y and Z axes, in that order. Note that the result row vector is ordered as [thetaZ thetaY thetaX].
See also the "eulerAnglesToRotation3d" function for details on the reverse operation.
regards, David
SM (view profile)
Hi David,
What convention are you following in the Rotation3dToEulerAngles file? I am not completely familiar with the equations, but it looks to me like this is an XYZ convention.
Thanks, Sem
David Legland (view profile)
Hi Johannes,
the volume of polyhedron is defined with respect to the orientation of faces. If faces point outwards, volume is positive, otherwise it is negative. You can simply consider the absolute value of the result. Be careful to ensure that orientation of all faces is consistent: you can use "checkMeshAdacentFaces" for that purpose.
David Legland (view profile)
Hi Dexter,
for drawing a 3D polyhedron, you can use the "drawMesh" function. It uses a classical FaceVertices representation. For creating the polyhedron, you may have a look at the 'steinerPolytope' function, it seems it may fit your needs.
regards,
David Legland (view profile)
Hi Dexter,
for drawing a 3D polyhedron, you can use the "drawMesh" function. It uses
Johannes Neumann (view profile)
Hi David,
thanks for your quick reply. I am using the following code now:
K = minConvexHull(tmpVerts,myEPS)';
tmpInd = unique(cell2list(K));
[K, vol] = minConvexHull(tmpVerts(tmpInd,:),myEPS);
I modified your minConvexHull to pass the volume from qhull and cell2list is cell2mat essentially.
Concerning the volume: I guess it should be
abs(det(tetra) / 6)
in your meshVolume function? Otherwise, I do not get positive volumes in general.
Dexter (view profile)
Hej, David:
Thanks for your sharing of this wonderful library.
I am working on draw a polyhedron, the first Brillouin Zone in Physics. Which specific function should I look in this library for my purpose?
Best wishes!
matteo (view profile)
Hi David,
thank you so much for your work (the intersectLineSphere), I am working on my master thesis and you really saved me.
David Legland (view profile)
Hi Matthew,
you can check the 'intersectLineMesh3d' function. It uses line representation in the form [x0 y0 z0 dx dy dz], where (x0,y0,z0) is the origin of the line and (dx,dy,dz) represents a direction vector of the line. Meshes are represented as vertices + faces, with vertices being Nvby3 array of vertex coordinates, and faces being either Nfby3, Nfby4 array of vertex indices, or cell arrays of vertex indices allowing more generic faces.
There is a function for reading meshes in OFF format ('readMesh_off'). Otherwise, you may want to check other contributions to import (and eventually convert) you files. See graph toolbox for example.
Matthew Brandsema (view profile)
Hello, I just found this toolbox and I am wondering if you can tell me if what I want to do is possible with these functions.
I have a vector equation of a line in 3D space, and I want to determine the position that it intersects with a mesh.
The mesh will be a mesh created in an external modeling program. (I can obtain all the information about its vertex coordinates from opening up a .obj file and extracting the information. )
I also need to know the normal vector of the mesh at the point of intersection.
Can this be done???
Fritz (view profile)
David (view profile)
David Legland (view profile)
@Abdulrahman,
sorry, I missed your comment... There is no direct function for testing if a point is inside a 3D circle. What you can do is 1) extract the plane containing the circle, 2) test if the point belong to the plane, and 3) if yes, compute coordinates of point in 2D plane, and test if the 2D point is within the 2D circle (you may need geom2d also).
hope this helps ?
regards
David Legland (view profile)
Hi Jin,
yes, there aer numerical precision issus with this function... I have submitted a new version, that is more tolerant for finding intersection at edges of faces. However sometimes too many intersections are obtained. You can remove them by using the 'mergeClosePoints' function from the geom2d toolbox.
regards,
David
Jin (view profile)
Hi David,
Really great tool! It helped me with a lot of problems. Thank you for sharing it with us.
Recently I'm using intersectLineMesh3d quite often. Here is an example I feel not all the intersections are found.
V =
0.5036 0.3333 0.3063
0.4102 0.1667 0.2098
0.1742 0.3820 0.1983
0.3505 0.4296 0.1688
0.3183 0.3623 0.3165
0.4246 0.2660 0.1180
P =
2 3 6
2 3 5
3 4 6
3 4 5
1 2 6
1 2 5
1 4 5
1 4 6
LINE =
0.3538 0.3302 0.2503 0.8359 0.1600 0.5250
This command
[inters pos inds] = intersectLineMesh3d(LINE,V,P) only gives me one intersection, but it should have two instead. Could you help me with that? Did I misunderstand the meaning of input and output? Thank you so much.
David Legland (view profile)
@Johannes,
The new set of faces refer to indices of vertices actually used by the faces. So it is necessary to use the following syntax:
[V2 F2] = mergeCoplanarFaces(vertices, K);
drawMesh(V2, F2);
The demo file is somewhat outdated, I will update it, and fix the doc of the mergeCoplanarFaces function.
Concerning the tolerance value, it is used for comparing normalised normal vectors of planes containing faces. There can be some numerical issues, but using 1e4 should be fine.
regards,
David
Abdulrahman (view profile)
David
Is there a function to find 3d points within a 2d circle !?
Great submission THANK YOU
Johannes Neumann (view profile)
Dear David,
I was very happy to find geom3d as it provides a multitude of tools i need. However, I am concerned to have discovered some oddities in mergeCoplanarFaces(). I have Voronoi polyhedra that are clipped to fit inside e.g. a cube. Here's an mwe:
%%% code %%%
close all;
clear all;
clc;
v1 = [0.7945 0.2065 0.3770
0.7851 0.2019 0.3754
0.6767 0.1483 0.3603
0.6747 0.1566 0.3578
0.3994 0.0620 0.8142
0.3305 0.4372 0.8683
0.5055 0.3719 0.8659
0.3032 0.2532 0.8922
0.3026 0.1366 0.8615
0.0841 0.2280 0.8028
0.1441 0.3496 0.8355
0.1113 0.2820 0.8271
0.8097 0.2518 0.7474
0.8414 0.2398 0.4822
0.7089 0.3694 0.8072
0.7858 0.3168 0.3958
0.3059 0.4221 0.8716
0.1471 0.3505 0.8370
0.2376 0.3283 0.8869
0.2138 0.3265 0.8775
0.3857 0.0514 0.7753
0.5503 0.0944 0.4249
0.2406 0.1644 0.5060
0.3526 0.4066 0.3743
0.3545 0.4235 0.3775
0.2192 0.2369 0.4857
0.1307 0.1782 0.5772
0.0560 0.1827 0.6723
0.0610 0.1760 0.6639
0.4537 0.5529 0.4090
0.2929 0.4950 0.4783
0.5109 0.6260 0.6706
0.1959 0.4789 0.6226
0.1916 0.4725 0.5978
0.5015 0.6228 0.6758
0.5646 0.5852 0.4243
0.4909 0.5704 0.4153
0.5468 0.6032 0.4606];
v2 = [ 0 1.0000 1.0000
0 0.8719 0.8367
0 0.7113 0.9914
0.0336 1.0000 0.7575
0.1827 0.9701 1.0000
0.1884 0.9592 1.0000
0.1394 1.0000 0.8965
0.1859 1.0000 0.9718
0.1910 1.0000 0.9642
0 1.0000 0.7134
0.1239 1.0000 0.8761
0.1670 1.0000 1.0000
0 0.7024 1.0000
0.1259 0.8741 1.0000];
vertices = v1;
% triangular faces can be obtained from convhulln
K = convhulln(vertices);
% display polyhedron
figure(1); clf; hold on;
set(gcf, 'renderer', 'opengl');
drawPolyhedron(vertices, K);
K2 = mergeCoplanarFaces(vertices, K, 1e4);
% draw the polyhedron
figure(2); clf; hold on;
set(gcf, 'renderer', 'opengl');
drawPolyhedron(vertices, K2);
%%% code %%%
For v1, mergePlanarFaces only merges faces if the PRECISION is set to values > 1e5. This startles me, since this particular polyhedron is not modified by me, but comes straight from convhull. For v2, the result seems to be completely garbled.
Am I doing something wrong here?
Cheers,
Johannes
David Legland (view profile)
Hi Chien Ting,
thanks, this is fixed! Yes, if you have any more updates, I an include them. Do not hesitate to contact me by email (on my author page)
regards,
David
Chien Ting CHIN (view profile)
I've been using your toolboxes for years, thanks for some great tools.
In drawEdge3d.m, line 29, I believe is a bug, it would never get executed. The correct line might be:
elseif nargin >= 6;
Also I've complete the documentation of drawEdge3d.m for my group's use. If you want see/use my version, let me know!
Thanks for some great tools!
Chien Ting CHIN (view profile)
Sorry I pasted the original line. The correct line might be:
elseif nargin >= 6;
Jdeen (view profile)
David Legland (view profile)
@John,
thanks for reporting! After checking, it appears that in your example, the plane is passing exactly through one of the vertices of the mesh. This can sometimes happen, and I do not know yet how to handle these cases. A simple workaround is to slightly change the position of the plane (or the mesh), by adding a random value at the origin:
plane = createPlane( randn(1,3)*1e3, [ 0 1 1 ] );
Hope this helps ?
John Anderson (view profile)
Hi David, great submission,
I think I have noticed a slight problem with calls to intersectPlaneMesh.m and wasn't sure if this had previously been flagged.
When I exectue the following to calculate the intersection of a plane with a sphere I find that when N = 10 I get the error
Error using horzcat
Dimensions of matrices being concatenated are not consistent.
Error in intersectPlaneMesh (line 85)
polyEdgeInds = [polyEdgeInds currentEdgeIndex]; %#ok<AGROW>
Error in testIntersectingMesh1 (line 27)
polys = intersectPlaneMesh( plane, vertices, faces );
However, when N=15 there are no exectuion problems
Code:
% create a plane
plane = createPlane( [ 0 0 0 ], [ 0 1 1 ] );
% create sphere
N = 10;
[ x, y, z ] = sphere( N );
% triangulation
tri = DelaunayTri( x(:), y(:), z(:) );
%vertices
vertices = tri.X;
%faces forming convex hull
faces = convexHull( tri );
cla;
% draw solid
patch( 'Vertices', vertices, 'Faces', faces, 'facecolor', 'y', 'edgecolor', 'y', 'facealpha', 0.1 );
hold on;
% intersecting polygon
polys = intersectPlaneMesh( plane, vertices, faces );
% draw polygon
drawPolygon3d(polys, 'color', 'b', 'LineWidth', 2);
% fill polygon
fillPolygon3d( polys, 'r', 'faceAlpha', 0.2 );
Best wishes,
David Legland (view profile)
Hi Germano,
thanks, I will check it! I remember I used normalisation before cross product for improving stability, but I have to check again. Anyway, should be fixed in next release!
Germano Gomes (view profile)
Great library, i use it all the time, thanks.
A small correction for:
function normals = faceNormal(nodes, faces)
if isnumeric(faces)
% compute vector of first edge
v1 = nodes(faces(:,2),1:3)  nodes(faces(:,1),1:3);
v2 = nodes(faces(:,3),1:3)  nodes(faces(:,1),1:3);
% normalize vectors
% v1 = normalizeVector3d(v1); % remove
% v2 = normalizeVector3d(v2); % remove
% compute normals using cross product
normals = cross(v1, v2, 2);
normals= normalizeVector3d(normals); % correction GTG 2014
else
normals = zeros(length(faces), 3);
for i=1:length(faces)
face = faces{i};
% compute vector of first edges
v1 = nodes(face(2),1:3)  nodes(face(1),1:3);
v2 = nodes(face(3),1:3)  nodes(face(1),1:3);
% normalize vectors
% v1 = normalizeVector3d(v1); % remove
% v2 = normalizeVector3d(v2); % remove
% compute normals using cross product
normals(i, :) = cross(v1, v2, 2);
end
normals= normalizeVector3d(normals);% correction GTG 2014
end

Always better to normalize at the end otherwise the final normals are not normalized because of roundoff etc errors in the cross computation, this also has the advantage of eliminating 2 calls to normalizeVector3d, making faceNormal a tad faster.
samart (view profile)
Khaled Khairy (view profile)
After some minor initial tweaking, I find this a very helpful mission. Despite its size all parts seem to work seamlessly together. Thanks David.
Shaoshuai (view profile)
Very helpful! Thanks for sharing!
J.R.! Menzinger (view profile)
Very useful.
David Legland (view profile)
Hi Cedric,
thank for your interest!
Actually the centroid function is provided in the "geom2d" library, available on the Fex as well. I try to minimise dependencies between both submissions, but apparently some are still remaining...
You can also replace the centroid call in the demo script by the following:
faceCenter = mean(vertices(faceVertices, :), 1);
This should work as well.
Regards,
David
Cedric (view profile)
Hi David,
Thank you for your code!
I have a small problem.
I am not able to compile the DrawSoccerBall.m demo function.
there's a parameter missing when calling centroid function since the definition of centroid function is function centroid = polyhedronCentroid(vertices, faces)
Error message when I compile:
Undefined function 'centroid' for input arguments of type 'double'.
Error in drawSoccerBall (line 56)
faceCenter = centroid(vertices(faceVertices, :));
David Legland (view profile)
@Yong Min Yoon:
I just realized that I already implemented a "polygonArea3d" version, that is included in the geom3d archive... So I think this should do the job!
Regards,
David
David Legland (view profile)
@Yong Min Yoon:
It's great if ou could make the fucntion work better! Could you sent me the code so that I can include it in a future release ?
For polygon area, there is a function "polygonArea" in geom2d, that works for planar polygons (beware that it returns a signed area). You can transform 3D polygon to planar polygon by using "projPointOnPlane" function. I think this should be ok for computing area ?
regards
Yong Min Yoon (view profile)
@David
I modified your intersectPlaneMesh function code and make it work to find the multiple contours.
As a next step to make up my algorithm, I want to know that if there's a area computation of the polygon(the closed contour) in your works or not.
If you let me know, this will be great helpful for me. Thank you
Yong Min Yoon (view profile)
@David
Your code is totally what I want. So, I'm just going to try to fix your code. Can you explain what the problem is in your function?
David Legland (view profile)
@Yong Min Yoon:
first of all sorry for the delay to reply;
The function you are working with at the moment is not fully implemented. It expected to returns a set of polygons, but it currently returns only one. I have plan to extent to multiple polygons, but it will require some work (one idea would be to recursively link adjacent vertices until the first vertex is reached, then to start a new polygon). So, no easy solution for the moment, but stay tuned...
Yong Min Yoon (view profile)
@ David Legland
Hi, first of all, I'm very thankful for your codes. I have some questions in your "intersectPlaneMesh function".
I try to make the some cross section information of the vertebrae. Your code will be used to make up that.
when I used that codes with a vertebrae mesh data, 2 contours are expected to show the cross sections. However, only one contour was shown in the figure.
I think that your polys variable which is return in intersectPlaneMesh, is used to represent the multiple contours to show intersected region by the plane.
If it is right, can you explain how to set the multiple contours by the plane?
Images are in the below URL
URL: http://gall.dcinside.com/board/view/?id=medicalscience&no=162489
P.S. intersectionsPoints in your intersectPlaneMesh function seems to exactly show intersected points on the mesh data.
Evan (view profile)
@David:
Thanks for the response. What you've provided here has already been extremely helpful anyway.
I also had a small suggestion for computing vertex normals. If I'm reading it right, the function normalizes the facet normals before averaging them. But it turns out to be more accurate to weigh by area and angle (http://www.bytehazard.com/code/vertnorm.html). For area, I believe that just means simply not using normalizeVector3d when calling faceNormal. It shouldn't be too difficult to code the angle part on my own, but I'm just throwing it out there in case you were interested.
@Sven:
Awesome, thanks for the heads up.
Sven (view profile)
@Evan,
Please check out http://www.mathworks.com/matlabcentral/fileexchange/43013unifymeshnormals
It should do what you're looking for.
Sven (view profile)
@Evan,
I'm working on something to do exactly this. Will keep you posted here.
David Legland (view profile)
@Evan:
depending on how the meshes are constructed, the orientation of normals may be consistent (ie, all pointing outwards of the mesh), or not.
The function "faceNormal" does not make any assumption on this, and simply computes the normal of each face.
However, some other functions require correct orientation of the normals. This is the case for vertexNormals, for example.
I do not have perfect solution... I have written "checkMeshAdjacentFaces" to check normal orientation consistency of a given mesh, and it can be safe to use it also. But I do not have direct solution to reorient face normals in given direction.
hope this helps ?
Evan (view profile)
Hello David, I had a question about calculating normals. In the function faceNormal, you mention that no orientation is defined. But in vertexNormals, faceNormals is used to calculate the normals at the vertices. Won't this lead to some problems, or am I missing something?
I'm specifically interested in calculating normals of a surface (facet or vertex) that are oriented in the same direction, which I understand is difficult and nontrivial problem. Do you have any suggestions? I noticed the checkMeshAdjacentFaces function, but I'm not sure if that's useful for "reorienting" specific normals.
Yong Min Yoon (view profile)
Hi, first of all, I'm very thankful for your codes. I have some questions in your "intersectPlaneMesh function".
I try to make the some cross section information of the vertebrae. Your code was used to make that.
when I used that codes with a vertebrae mesh data, 2 contours are expected to show the cross sections. However, only one contour was shown if the figure.
I think that your return(polys) is used to represent the multiple contours to show intersected region by the plane.
Is my thought right? If it is right, can you explain how to set the multiple contours by the plane?
Images are in the below URL
URL: http://gall.dcinside.com/board/view/?id=medicalscience&no=162489
P.S. intersectionsPoints in your intersectPlaneMesh function seems to exactly show intersected points on the mesh data.
David Legland (view profile)
Hi Benjamin,
you're not wrong, there was a mistake in the function surfToMesh...
I have submitted a fixed version, that should appear soon.
Regards,
David
Benjamin (view profile)
Dear David,
I'm a bit of a novice user, so it's likely this is a flub on my part, but I figured it worthy of mention.
I've been attempting to use the cylinderMesh function, but with little success. I even tried the "Draw a rotated cylinder" example, but all I get is
Index exceeds matrix dimensions.
Error in surfToMesh (line 91)
z = z(1:n1, 1:n2);
Error in cylinderMesh (line 63)
[vertices faces] = surfToMesh(x, y, z);
I'm not entirely sure what is causing this, any ideas?
Benjamin
David Legland (view profile)
Hi Khaled,
thanks for reporting problem. I will check behaviour for points given in spherical coords. The function works fine with points given as cartesian coordinates (assuming sphere is centered), so I hope this could help using the function.
Regards,
David
Khaled Khairy (view profile)
With all respect for this valuable submission. I do have some slight critique. I downloaded geom3d to "quickly" check against my code for the determination of the spherical angle, as the function "sphericalAngle" promised. I only called sphericalAngle exactly as indicated in its help. I ended up wasting two hours of quality time ploughing through various fixes. [1] calls to functions whose names have been changed. Sorry , but I just stopped keeping track(some have already been mentioned in these comments) and [2] createPlane, was called from within sphericalAngle apparently with the wrong dimensions. [3] intersectLinePlane complained about matrix dimension mismatch on line 71. Honestly, I consider perseverance one of my strenghts, but I gave up at this point, as I can't say how much more time this really simple operation would/should cost me should I continue with geom3d. Sorry, but a smooth program flow is really important, especially in extensive code submissions such as this one.
David Legland (view profile)
Hi Orestis,
You can check the function 'intersectPlaneMesh', that computes the 3D polygon resulting from the intersection of a polygonal mesh with a plane. The algorithm used is simple, so there can be numerical issues. Then you can compute the area with the function 'polygonArea3d'. By considering several planes, you should be able to do what you want.
There is no function in geom3d for reading STL files. So you should either write yout own, or find one.
regards,
David
Orestis (view profile)
Dear David
Could you let me know if your library can be used i order to slice a 3d stl file with arbitrary planes?
The aim is to calculate the area bounded by the stl file in each slice.
With best regards
David Legland (view profile)
Dear Dmitri,
thank you for your feedback! Actually, many functions require the geom2d contribution to work properly. I try to make geom3d as independent and selfcontained as possible, but downloading geom2d as well is strongly recommended. I will update the different functions you mentioned to remove/reduce dependencies.
Regards,
David
Dmitri Kamenetsky (view profile)
Great package! Some minor bugs:
1. circleToPolygon function is missing. It is called in demoRevolutionSurface,
drawTorus and torusMesh.
2. vectorNorm is now vectorNorm3d. It is used in intersectLineTriangle3d, triangleArea3d.
3. drawEdge is now drawEdge3d. It is used in demoGeom3d.
Dmitri Kamenetsky (view profile)
One more bug: circleToPolygon function is missing. It is called in demoRevolutionSurface, drawTorus and torusMesh.
panagiotis (view profile)
@Sven
I am sorry , but I didn't know that point of view for plane creation. My background bibliography doesn't include such information.Thank you for informing me.
Sven (view profile)
@panagiotis: I highly recommend reading *again* the documentation for createPlane  it is very short and very clear, but your comment shows you still have a misunderstanding.
You seem to think that "3 points are a plane". This is not true. A plane is like a coordinate system  it has:
 An origin POINT
 An Xdirection VECTOR
 A Ydirection VECTOR
You can *create* a plane from 3 points, but a plane is *not* just 3 points. When you read the help for createPlane, you will see the first example which shows you *exactly how* to create a plane from 3 points:
PLANE = createPlane(Pt1,Pt2,Pt3)
You will see that PLANE is NOT EQUAL to [Pt1,Pt2,Pt3]. There are *very* good reasons for this which you will understand once you read the help for createPlane.
panagiotis (view profile)
@David
You are right David , but didn't thought at all that there would be a misunderstaning for defining the plane by three points!I think it is commonly accepted. Haven't heard before for direction of plane from the third point.
Anyway, came up with something new.
In createPlane , you take into account only the 2nd & 3rd point for the cross product.You don't create vectors like p1p2 & p1p3.It looks to me as if you define the [0 0 0] , point for all planes. The normal in some cases looks wrong to me and didn't find any help.. If I use the vectors , results are fine.
plane=[1 1 1 ,2 2 1, 3 3 1]
planeNormal(plane)
ans = 0 0 0
I think that this should be a vector like [0 0 a].
David Legland (view profile)
@panagiotis:
Apparently you did not really even try to get som help by yourself. Either "help planes3d", "help projPointOnPlane", or even "help geom3d" gives you the definition used in the library for representing planes.
Giving a low rating just because you did not want to read the doc is somewhat unfair...
By the way, many thanks to Sven for the support!
Regards,
David
panagiotis (view profile)
Sven,thank you for informing me. The truth is that I didn't got that deep in help..
As soon as I use the library again , I will reconsider the 1star rating!
Thanks!
Sven (view profile)
Hi Panagiotis, the help for projPointOnPlane says "... PLANE is a [N*9] array (see createPlane for details)". The help for createPlane is very clear about the definition.
Anyway, I'm glad that your issue has now been identified. One further note is that your "other" direction vector was specified as [0 1 1], which isn't "normalised" (ie, it has a magnitude of more than 1). This won't affect projPointOnPlane but it can affect other functions such as planePosition if you intend on using them.
Hope this makes your code work again, and if so, perhaps you might reconsider the 1star rating :)
panagiotis (view profile)
@ Sven
There was no such notice for this definition. I didn't know that the third point defines the plane's direction .
Sven (view profile)
@panagiotis:
Your definition of the plane does not make sense:
plane=[0 1 0 0 1 1 0 0 0];
Notice that elements 7:9 are all zero. Elements 7:9 give the plane's Yaxis direction vector. A vector of all zeros has no direction at all, therefore it is impossible to project a point onto this plane, because the plane doesn't exist.
panagiotis (view profile)
@Sven(strong defense for this library!)
point=[0 0 0];
plane=[0 1 0 0 1 1 0 0 0];
projPointOnPlane(point,plane)
ans = NaN NaN NaN
Sven (view profile)
@panagiotis:
>> projPointOnPlane([100 100 0],[0 0 0 1 0 0 0 1 0])
ans =
100 100 0
... seems fine to me. Please provide a reproducible error.
panagiotis (view profile)
The main problem I found is in projectionPointOnPlane in geom3d library. We tried to project a point on a plane on which the point already belongs to. The result that came up was NaN instead of the point's coordinates. That led to some problems with other functions like intersection..
Panagiotis
Sven (view profile)
@panagiotis: please describe the actual issue(s) you found. I would bet that if you found a reproducible error, David would very quickly address it. Just saying "there are errors with intersections" isn't actually very helpful, and doesn't let anybody improve the package above the 1star you feel it deserves at the moment.
panagiotis (view profile)
Dear David,
I am using your library for a couple of months and I have found some errors especially on projecting points and intersection..
Thought you might find it helpful.
Panagiotis
David Legland (view profile)
Hi Charles,
I have created a project page on sourceforge call "matgeom", that gather geom2d and geom3d. There are some forum facilities, so this should be easier to post data there I think.
http://matgeom.sourceforge.net/
Regards,
David
Charles (view profile)
Dear David,
I'm trying to use surfToMesh and then meshSurfaceArea function to calculate the approximate area covered by the dots with X and Y coordinates specified on XY plane. However, the area returns the value that does not make sense to me. I wonder if it's possible for me to attach the matrix here and get some help why it does not work in my code. The dots I have represents nearly a semicircle and the result gives me the area close to a full circle of that diameter. Is there some tolerance setting on meshing? Thank you very much!
Best,
Charles
Student Uni (view profile)
HI Sir
Basically i have to draw that 3D rectangle and divinde its faces in
different infinitesimal areas, and divide that rectangle in small volums.
then i need to take the coordinates of Vertices of areas and volumes and
also coordinates of their central point
clc
clear all
close all
for x=0:1:2;
for y=0:1:2;
plot3([x,x],[0,2],[y,y]);
hold on
end
end
for x=0:1:2;
for y=0:1:2;
plot3([0,2],[x,x],[y,y]);
hold on
end
end
for x=0:1:2;
for y=0:1:2;
plot3([y,y],[x,x],[0,2]);
hold on
end
end
i need something like this but this is square and i dont know how to get
the required co ordinates
I appreciate in advance your kind attention
Best Regards
HoaiNam LE (view profile)
Many thanks David!!!
Nam
David Legland (view profile)
@HoaiNam,
I have submitted a new version with some new functions you should find useful:
* functions 'reversePlane' and 'reverseLine3d', that avoid explicitely permutting the directions vectors
* function intersectPlaneMesh, that returns the intersection between a plane and a 3D mesh. Not a full clipping function yet, but this should help!
regards,
David
Yuri (view profile)
David, thanks a lot! Yuri
David Legland (view profile)
@Yuri,
The function sphericalAngle assumes the input points are located on the surface of the unit sphere, that is not the case in your problem.
I think you should use the "angle3Points3d" function, that computes the angle between 3 points in space, using the second point as center of the angle.
anglePoints3d(p1, j1, l1)
ans =
1.5708
HoaiNam LE (view profile)
Dear David,
Thanks for your very useful toolbox!
I've used some of your functions and I've found some things :
+ My problem :
I have a 3D point set. I create a minimum convex hull of this point set, using convhulln, so I have a list of facets of the convex hull.
I'd like to clip this convex hull by a plane. And this leads me to your function.
+ There 2 ways to do this with your toolbox :
1/ Use the function clipConvexPolyhedronHP. It works for my problem!
However if we want to have a side of convex hull (below or above to the plane) to clip, we must have a right plane input!!!
For example:
chX = convhulln(Pts); %Pts : 3D point set
% Select 3 points on the plane z=vClip
P=[0 0 vClip;
0 1 vClip;
1 0 vClip];
% Create plane z=vClip
plane=createPlane(P);
% This function createPlane gives me : plane=[0,0,0.7,0,1,0,1,0,0]
% Then the function clipConvexPolyhedronHP returns me a wrong polyhedron. So I permute the 2 direction vectors of the plane : plane=[0,0,0.7,1,0,0,0,1,0] and it returns me a polyhedron with the right side that I desire.
So do you have a tip to use the function createPlane to have a good result (so I dont have to permute the 2 direction vectors of the plane) ?
2/ Use the function polyhedronSlice. It doesn't work because it doesnt return only the intersection points of the plane and the polyhedron, but also some points that are wrong (see the image https://dl.dropbox.com/u/8702546/WEB/image_1.png).
However, if the function works, I can control the side of polyhedron that I want to clip by a simple command of Matlab!
So I feedback you this to correct the function polyhedronSlice!
I dont know if someone has already asked about these things...
Regards,
Nam
Yuri (view profile)
David, very nice and useful package. Thanks.
I have a question about sphericalAngle function. I defined the following plane:
>> p1 = [1, 1, 1]; p2 = [100, 100, 1]; p3 = [1, 100, 1];
>> pl = createPlane(p1, p2, p3);
Then a point outside the plane and the projection point onto the plane:
>> l1 = [40,40,40];
>> j1 = projPointOnPlane(l1, pl);
I'd expect the angle between l1, j1 and any point on the plane to be pi/2. However, this is what I get:
>> alpha_rad = sphericalAngle(l1, j1, p3)
alpha_rad =
1.5745 (~ pi/2)
>> alpha_rad = sphericalAngle(l1, j1, p1)
alpha_rad =
0 + 2.1073e08i (= 0)
>> alpha_rad = sphericalAngle(l1, j1, p2)
alpha_rad =
3.1416  0.0000i (= pi)
Can you, please, clarify the way the function works?
Thanks
marc (view profile)
Very nice. thanks
AP (view profile)
Great tools.
Henri SHI (view profile)
Myles (view profile)
Haochen Tang (view profile)
Hi David,
Thank you. I miss understand the definition. I thought it was the origin and the end point. Thanks a lot.
Best,
Haochen
Eduardo (view profile)
Hi David,
I´ll try to use some of that functions. First of all i want to do only one slice in order to get the intersection points, only after that i´ll do it with the other parallel planes.
Thanks.
Regards,
Eduardo
David Legland (view profile)
Hi Eduardo,
What you want to do seems possible, but to my knowledge there is no direct method.
There are some functions in the "meshes3d" subpackage that could help you, like intersectLineMesh3d or polyhedronSlice. You can also use intersectEdgePlane > you can obtain the set of intersection points between a plane and all edges. If you want to represent the slice as 3D polygon, you will still need to link points in appropriate order.
@Haochen:
Beware of the definition of 3D line. In the toolbox 3D line is represented by [x0 y0 z0 dx dy dz]. In order to define a vertical line you should use [0.6 0 0 0 0 10]. In this case you will get the correct result.
Regards,
David
David Legland (view profile)
Hi Eduardo,
What youwant to do seems possible, but to my knowledge there is no direct method.
There are some functions in the "meshes3d" subpackage that could help you, like intersectLineMesh3d or polyhedronSlice. You can also use intersectEdgePlane > you can obtain the set of intersection points between a plane and all edges. If you want to represent the slice as 3D polygon, you will still need to link points in appropriate order.
Regards,
David
Haochen Tang (view profile)
Hi David,
Thanks a lot for this great tool. I am using the intersectLinePolygon3d function. Some of the results seem not make sense. For example:
pts3d = [1 1 5;1 1 5;1 1 5];
line=[0.6 0 0 0.6 0 10];
[inter inside] = intersectLinePolygon3d(line, pts3d);
And I got:
nter =
0.9000 0 5.0000
inside =
1
Should we expect to get inter = 0.6 0 5?
Thanks so much,
Haochen
Eduardo (view profile)
Hi David
I´m trying to slice a triangular mesh. I´m not an expert in Matlab and i´m having some difficulties in implementing this. I have the edges, and i´m trying to get the intersection points with several parallel planes...
Can you give me a hand?
Thank´s.
Regards,
Eduardo
David Legland (view profile)
Hi Amanda,
I do not know similar library for scilab. I suspect you need to recode everything...
If you are looking for a free alternative, I can mention a port to octave of libraries geom3d+geom2d:
http://octave.sourceforge.net/geometry/index.html
regards,
David
Amanda (view profile)
Hi, David
Having been using this package for years. Really love it. Thanks for your gr8 job. Now I'm researching using Scilab to replace Matlab for some application. Searched scilab but didn't find geom3d available there. Any suggestion?
Amanda (view profile)
Tianjin Univ (view profile)
David Legland (view profile)
Hi Nir,
did you mean 3D polygon or 3D mesh ? It seems you speak about mesh, right ? In this case you case use function "intersectLineMesh" (contained in the updated submission). This should do the job.
@Benjamin: I have also added a function polygonArea3d in the update.
Nir (view profile)
Hey David,
It seems that the functions:
intersectRayPolygon3d
intersectLinepolygon3d
Don't work properly.
I have a rectangular 3D polygon (fairly simple, like an elongated cube) defined by both the vertices and the planes of it's sides.
A line or a ray, with a location in the middle of it in ANY direction should generate an intersection point. This does not work. Not for a line nor a ray.
Vertices:
0 0 0
0 0 2.7000
5.7000 0 0
5.7000 0 2.7000
5.7000 7.0000 0
5.7000 7.0000 2.7000
0 7.0000 0
0 7.0000 2.7000
Ray
Origin: 2.8508 3.5013 1.3500
Direction: 0.1 0.1 1
Help?
Hannes (view profile)
Thanks mate. Beer flies out to you :)
David Legland (view profile)
Hi Ben,
you can check the function "polyhedronSlice". It computes the 3D polygon that results from the intersection of a polyhedron with a plane. Example of use:
[nodes faces] = createCube;
plane = createPlane([.5 .5 .5], [2 3 4]);
poly = polyhedronSlice(nodes, faces, plane);
figure; hold on;
drawMesh(nodes, faces, 'facealpha', .5);
fillPolygon3d(poly, 'm');
There is no (not yet...) function for computing area of a 3D polygon. However it can be obtained by projecting on the intersecting plane:
poly2d = planePosition(poly, plane);
area = abs(polygonArea(poly2d))
Regards, David
Benjamin Sanderse (view profile)
Hi David,
Thanks for your very nice work. I am searching for a routine that can find the (area of a) polygon that results from the intersection between a plane and a cube. Would this be possible with your library?
Regards, Ben
David Legland (view profile)
Hi Patrik,
no, you haven't missed it, and it could be useful. I will add it in a future release.
regards
Patrik Eschle (view profile)
distanceLine3d works on infinite lines a+t*b. Sometimes (collision detection) the distance between line segments AB and CD is required. A distanceSegment3d would be useful.
The same is true for points (distancePointSegment ).
Or did I just miss it?
Patrik Eschle (view profile)
Nicely written, good comments  thank you!
Shuhao Cao (view profile)
Sakshi (view profile)
David Legland (view profile)
@Qais,
you should install both geom2d and geom3d, typically in the same directory. Then you have to add directories to the path, using te 'addpath' function, or the 'File>Set Path' menu). You need to add at least the directories 'geom2d', 'geom3d', 'polygons2d', 'meshes3d', and eventually 'polynomialCurves2d'.
Some demos are provided with each archive, you can run them directly, and have an idea on the syntax.
I will also try to add some info on the matGeom web page (that gathers geom2d and geom3d), at http://matgeom.sourceforge.net .
Hope this helps
Qais AlKhazraji (view profile)
Dear David,
I am just start using Matlab in my research for geometrical computation. I appreciate if you would like to demonstarate how to use this library(3d or 2d) since each function I used give me syntax error about other functions as undefined. I am a biginner in Matlab and may be my question is dump question. I am very thankful.
David Legland (view profile)
@Khaled and Qais,
What I had in mind was to use the following procedure :
* compute intersection of line and polygon plane
* project polyingon vertices onto the plane (ie, compute their 2D coords)
* project intersection point onto the plane
* check if 2D intersection point is within the 2D polygon.
The following piece of code should do the job (assuming poly3d is Nby3 and line is 1by6):
plane = createPlane(poly3d(1:3, :));
inter = intersectLinePlane(line, plane);
poly2d = projPointOnPlane(poly3d, plane);
pInt2d = projPointOnPlane(inter, plane);
inside = xor(isPointInPolygon(pInt2d, poly2d), polygonArea(poly2d) < 0);
inter = inter(inside, :);
Both geom2d and geom3d are required. I will add functions 'intersectLinePolygon3d' and 'intersecRayPolygon3d' in the next release.
regards
Qais AlKhazraji (view profile)
Dear David:
I read your comment for Khaled question. It is not clear to me about what you said"The best is to use 3D plane, and eventually to project all points on the plane, then to use function for 2D polygons." Which points we should project on 3D plane and from my understanding line intersect a polygon in one point.
Thank you in advance
David Legland (view profile)
Hi Khaled,
to create a 3D polygon, you simply need to concatenate their coordinates into a Nby3 array, e.g. poly=cat(1,p1,p2,p3);
There are no fucntions for intersections of 3D line with 3D polygon. The best is to use 3D plane, and eventually to project all points on the plane, then to use function for 2D polygons.
Khaled (view profile)
Dear eng. David,
Thanks, this powerful toolbox is very helpful.
I tried to create a 3D polygon from four or three 3D points, and couldn't. So instead I tried to create a plane, but when finding an intersection with a 3D line the answer will not be accurate as the plane was polygon.
Is there a possible way to create a 3D polygon from four or three points? and
Is there a possible way to find a intersection between a 3D line and 3D polygon?
your efforts are appreciated.
Brandon (view profile)
Brady (view profile)
Incredibly helpful toolbox. Mostly intuitive function calls, and thorough documentation to make integration easy.
Siyi Deng (view profile)
very helpful.
Sven (view profile)
Excellent package for spatial manipulation. Extensive toolbox, that handles input intuitively. The only suggestion I have is, for completeness, to document some of the methods of calling functions for the more obscure functions. Eg., localToGlobal3d has one documented method of calling, but under the hood it parses input in many different ways. I find myself typing "edit" instead of "help" just to make sure I'm using things efficiently.
David Legland (view profile)
The format for cylinder is as follow: [xv yv zy R xc yc zc], with:
[xv yv zv] the direction vector of the cylinder,
R the radius of the cylinder
[xc yc zc] the origin of the cylinder.
the L parameter, which is given in help, is not used.
The cylinder is considered infinite, please use "linePosition3d" if you need finite cylinder.
Please note that this format is not consistent with the function "drawCylinder", for example. Therefore the function "intersectLineCylinder" is likely to change in a future release.
DL
Paklah Abd Rahman (view profile)
Hi,
I tried the "intersectLineCylinder" what are the definitions of: xa, ya, za, and L for cylinder??..Please anybody
David Legland (view profile)
Hi,
I have uploaded a new version, that fixes reported bugs, and have added some demos. Demos are available as published mfiles, and can be accessed through the file presentation above. They demonstrates some usage possibilities.
Most drawing functions follow the 'plot' syntax. Data are given first, then optional couples of parameter name and value. If not implemented, try using "h=drawMyShape(...)", then set up options using "set(h, 'color', 'k')" for example.
For bug reports and suggestions, the best is to sent me directly and email (can be found in my author page, replace "DOT" by ".").
DL
Jon Sporring (view profile)
This package aims at solving visualization problems, I often have. However, I have not been able to I'm trying to use this package, but experiencing a number of problems:
1. How are the colors controlled of the objects drawn?
2. I want to draw cylindrical representation of arrows, since quiver3 together with drawPlane does not work well, possibly due to lack of control of drawing order. Instead I want to draw thick arrows, and I'm using drawCylinder.m. However, in line 89 is used the nonexisting function local2Global.m. It appears to be replaceable by localToGlobal3d, if localToGlobal3d line 29 is replaced with "center = varargin{1};".
For suggestions to corrections, etc. what is the best procedure?
Thanks, Jon
Nathan Thern (view profile)
This library and its companion, geom2d, are shining jewels  indispensable for doing nontrivial calculations on objects in 2d and 3d space.
By the way, leave all the files in the geom3d directory and add the directory to your path. Then you can type "help geom3d" or "doc geom3d" and get properly linked help text in the command window or the help window. The Contents.m file is the proper "MATLAB way" of providing documentation.
us (view profile)
it is too bad that this (apparently) fullfledged software package does not come with an appropriate and modern help/intro/example module called a published mfile...
as of now, it is somewhat tedious for a naive user to have to wade through the help sections of the individual help files (or the densly written contents.m)  and it will (most likely) not give your package as much attention as it probably deserves...
us