Re-labelling an image compared to another.

1 次查看（过去 30 天）

Robert Dupont 2015-11-5

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/253129-re-labelling-an-image-compared-to-another

Hello there,

I'm working on a project to create a measurement device in 3D using stereo-correlation.

We will take pictures of the samples with very small deformation between each other. The samples will have small circles drawn on them.

My problem is to follow those circles, from one image to another and from one camera to the other. To be more precise, I want to be able to follow each points through the experiment and calculate (with the matlab function "triangulate") their 3D location. Thus, I need to spot the center of each point and match them from one camera to the other, but also one image after the other.

So far I've done the part from one image after the other :

Propriete_Image_labellise = regionprops(Image_labellise,'Area','Centroid');

Where I get the size and the center of the different points.

centred = [];
for i = 1:size(Propriete_Image_labellise)
  if Propriete_Image_labellise(i).Area > 10
  centred = [centred;Propriete_Image_labellise(i).Centroid];
  end
end

Where I remove false points (here, points below 10 pixels in size).

for i = 0:size(centreg);
        while a == False
            %Test on the distance between two centers.
            if ((centreg(i,1)-centred(j,1))^2+(centreg(i,2)-centred(j,2))^2)^(1/2) > Ecart_maximum
                j = j+1;
            else
                temps2 = [temps2;liste2(j)];
                a = True;
            end
            %Test in case there is no match. Then a point (0,0) is created to match the other list. We then get two different list of the same size.
            if j> size(centred)
                temps2 = [temps2;0,0];
            end
          end
      end

Where I match the same points from two different images. with a distance criteria.

My problem is, from one camera to the other, I have too much differences, which makes the distance criteria invalid. The smaller distance could be from another point than the "real" match.

Then again, I might also want to use a speckle pattern, which makes this code kind of useless. Anyone more familiar with Matlab is able to share with me some function doing the work all by itself.

Thanks in advance.