MATLAB: Does using ginput on 3d surfaces not return the correct point of contact

3dginputMATLABselect3dsurfaces

Why does using ginput on 3d surfaces not return the correct point of contact?

If I use ginput on a 3-D plot, such as a surface, the coordinates returned are not the coordinates where the mouse touches the surface. In fact, the coordinates do not appear to make any sense at all. Why is this happening?

Best Answer

When you click on the figure window with the mouse, you are actually defining a line traveling into the screen along the line of sight. The 'CurrentPoint' property of the axis stores the endpoints of this

line, where the endpoints are defined by the axis limits in each dimension. The endpoints are defined in three dimensions (i.e., each endpoint is a 3-element vector); the first row of the 'CurrentPoint' value represents the endpoint farthest away from the viewer, while the second row represents the endpoint closest to the viewer.

Since the GINPUT function was intended for use with 2-D plots, it only returns two values, the x and y coordinates of the endpoint farthest from the viewer. That is, GINPUT returns the (1,1) and (1,2) elements of the 2-by-3 matrix returned by

    get(gca, 'CurrentPoint')

Note: In two dimensional plots, since the line travels directly perpendicular to the X-Y plane, this accurately represents the (x,y) coordinates of the intersection of the line with the plane.

Instead of using the GINPUT function, you can use the SELECT3D function to select 3D points on surface and patch objects. This function can be found on MATLAB Central at:

http://www.mathworks.com/matlabcentral/

Related Solutions

MATLAB: Specifying 3D points with the mouse

In short, don't use ginput(), use get(ax,'CurrentPoint') instead.

Basically the current position of the mouse, combined with the view matrix defines a stabbing vector. It will be up to you to interpret which plane that stabbing vector should intersect. Here is an example which prints out a coordinate in the view axes whenever the mouse moves. You could instead assign it to a 'onButtonPress' callback if you like. I'll be honest and say I can't quite remember which plane the view matrix uses by default, but it sounds from your question like you're on top of interpreting the result of the matrix transformation and are more interested in the replacement of ginput() with something more useful:

function example
figure('WindowButtonMotionFcn',@(src,evnt)printPos())
plot3(rand(5,1),rand(5,1)*20,rand(5,1)*5,'.'), view(3)
      function printPos
          clickedPt = get(gca,'CurrentPoint');
          VMtx = view(gca);
          point2d = VMtx * [clickedPt(1,:) 1]';
          disp(point2d(1:3)')
      end
end

MATLAB: I would like to offer a new algorithm to Matlab. Whom should I contact

Contact nobody.

Seriously, just because you have a neat new algorithm, the odds of The MathWorks wanting to add it to the language or to a specific toolbox is VERY small. That requires a broad utility, for code that almost always, they want to write themselves. Remember, code that is a part of MATLAB needs to be maintained by The MathWorks. Will you be around 10 or 20 years from now when there are problems to be resolved?

As I said, it requires broad utility, because when they provide code, it will become part of MATLAB FOREVER. This means your algorithm must be useful to many people, to justify them distributing it, writing it, maintaining it, debugging it, testing it.

Does TMW want to pay you for your algorithm, on your recommendation alone? Based only on a paper that you wrote? Sorry, but almost surely, no.

TMW employees do very much read the literature. They keep up to date. I know these things to be fact. If they see something described that THEY decide is of value to a broad spectrum of users, and it is something they can write and maintain, something that is a robust idea, they have people who will write that code, and do so professionally. But they only have a limited set of resources. So they need to carefully choose what they will provide. Again, once part of MATLAB, it is part of MATLAB FOREVER, maintained FOREVER. Ok, almost forever.

(I have code written in MATLAB almost 30 years ago. It still works, and at TMW, they are proud to hear that. That means that they chose carefully and cogently what to write, what to provide, looking to the future, as well as to the past.)

So where does this leave you?

1. You can contact The MathWorks, with a feature request. Again, this will rarely result in your code becoming part of MATLAB, or a toolbox, because while you may think it to be neat and useful, they have very different standards.

2. Far better is to write the code yourself. Then post it on the MATLAB Central File Exchange (or Github, which the FEX can access.) Note that you cannot post p-coded files. So if you decide you want to keep some details of your algorithm secret, don't bother. And these sites are for FREE use. You cannot ask for payment for use, a licensing fee, etc.

3. You can write the code yourself, as a set of tools for use. Then charge others for your code. Now you will need to deal with distributing the code, enforcing your conditions, preventing others from copying your code, because the internet lets people do anything they want. So you will need to encrypt your code, perhaps in the form of p-code, perhaps as compiled code. But even then, a determined hacker can do a lot to your code.

But contacting The MathWorks with an offer of your great idea, and asking them to pay you for inclusion into MATLAB? While they will probably politely decline your offer if you hear anything at all, I'd bet that at least someone will quietly smile at the temerity of your request.

Sorry, but that is how it must work. All of this said by someone with dozens of highly useful tools, posted on the file exchange. Personally, I think much of the work I have done would be worth including in MATLAB, but I am a bit biased. So I accept the MATLAB Central File Exchange as a great place to post work, where anyone can make use of it.

Best Answer

Related Solutions

MATLAB: Specifying 3D points with the mouse

MATLAB: I would like to offer a new algorithm to Matlab. Whom should I contact

Related Question