Compute a colored Area on top of a curved surface

Contents

• background: a customer wanted to study a curved surface.
• get the image
• Use triangles to compute the area of the surface:
• image resoved at surface resolution
• find triangles, that have at least 2 red corners

background: a customer wanted to study a curved surface.

The surface is heated up by radiation. Image analysis should be used to compute the area above a certain temperature level. Here the mandrill serves instead of an IR-image

disp('This demo is meant to run in cell mode.')
return;

This demo is meant to run in cell mode.

get the image

load mandrill
im=ind2rgb(X,map);
sz=size(im);
ix=sz(1);
iy=sz(2);

% filter the image
h=fspecial('average',5);
filtim=imfilter(im,h);

% create reference surface
[x,y,z]=peaks(30);
sz=size(z);
sx=sz(1);
sy=sz(2);

% scaling
xfkt=sx/ix;
yfkt=sy/iy;

% plot image on surface
warp(z,filtim)
hold on;

% compute isoline for red level
% and plot 1st isoline only
rlevel=0.84
[c,h]=contour(filtim(:,:,1),[rlevel]);
idx=2:c(2,1);
plot3(c(1,idx)*xfkt,c(2,idx)*yfkt,idx*0+5,'b.')
axis([0 sx 0 sy -10 10])
title(['cut-level: ' num2str(rlevel)])

rlevel =
    0.8400

Use triangles to compute the area of the surface:

delaunay the surface, resize the image to the surface's points; make sure there are enough triangles covering the the color area of interest

tri = delaunay(x,y);
ptsim=imresize(filtim,[sx sy]);
rptsim=ptsim(:,:,1);
figure
mesh(rptsim)
title('smmothed red part of image')

image resoved at surface resolution

figure
warp(x,y,z,ptsim)
title('resized to surface resolution')
hold on

find triangles, that have at least 2 red corners

rptsim=rptsim(:);
rlevel=0.84;
ridx=find(rptsim>rlevel);
rtrue=zeros(sx*sy,1);
rtrue(ridx)=1;

count=rtrue(tri(:,1));
count=count+rtrue(tri(:,2));
count=count+rtrue(tri(:,3));

% try also 1 or 3 red corners: its fun to see
idx=find(count>=2);
rtri=tri(idx,:);

% and display them on top of the image
trisurf(rtri,x,y,z)


% compute the area of the triangles under
% consideration, formula taken from a
% textbook: Area=1/2*c*hc, hc=b*sin(alpha).
ntri=length(rtri);
triarea=0;
for k=1:ntri,
    xt=x(rtri(k,:));
    yt=y(rtri(k,:));
    zt=z(rtri(k,:));
    A=[xt(1) yt(1) zt(1)];  % corner vectors of triangle
    B=[xt(2) yt(2) zt(2)];
    C=[xt(3) yt(3) zt(3)];
    va=B-C;                 % side vectors of triangle
    vb=A-C;
    vc=A-B;
    a=sqrt(va*va');         % side lengths of triangle
    b=sqrt(vb*vb');
    c=sqrt(vc*vc');
    if abs(a*b*c)<1e-8      % avoid 1/0
        triarea(k)=0;
    else                    % compute area per triangle
        hh=(vb*vc')^2/((vb*vb')*(vc*vc'));
        triarea(k) = 0.5*c*b*sqrt(1-hh);
    end
end

title(['Area of triangles in square pixels: ' num2str(sum(triarea))])

Published with MATLAB® 7.0.1