Reason why this condition is generating NaN everywhere?

Hi all, my code is not working correctly. I suspect it is because of a condition I set in the final if statement, but I am not sure how to modify it to run correctly. This code takes velocity data from PIV and uses xcorr2 to find a characteristic length scale. When I call it outside the script, it fails to pass the final if statement, and everything turns into a NaN value. Does anyone know what I am doing incorrectly? I am a newbie so any help will be appreciated!

The first block is the function, and the second block is where I call it

code:

function [lambda,L] = length_scale(v_x,v_y,mesh)
    b = size(v_x);
    c = min(b);
    vx= v_x(1:c,1:c);
    vy= v_y(1:c,1:c);
    N = xcorr2(ones(c,c));
    corr_x= xcorr2(vx)./N;
    corr_y= xcorr2(vy)./N;
    [X,Y] = meshgrid(-c+1,c-1);
    [~,rho] = cart2pol(X,Y);
    for i = 1:c
        F(i)=mean(corr_x(i-1<=rho && rho<i) + corr_y(i-1<=rho && rho<i));
    end
    L=NaN.*ones(1,c);
    if (max(isnan(F)) == 0) && (min(F)<0)
        L=F./F(1);
        f=fit((0:c-1).', L(1:c).', 'exp1', 'Starpoint', [0 0]);
        lambda = -mesh/f.b;
    else
        lambda=NaN;
    end    
end

calling it:

count=1;
for i =1:cell_size
    
    velocity = (vel_x{i}.^2+vel_y{i}.^2).^0.5;
    avg_v(i) = mean(velocity, 'all',  'omitnan');
    std_vel(i) = std(velocity, 1, 'all', 'omitnan');
    [temp_lambda, Len(i,:)] = length_scale(vel_x{i},vel_y{i},mesh);
    lambda(count) = temp_lambda.*(pxsize);
% uncomment these lines when the length scale script works 
    count = count + 1;
end

[Abderrahim. B]

Hi!

Tried to run this code to see if I can help. Seems that the code you shared is incomplete.

cell_size = ?

vel_x = ?

vel_y = ?

Please share the correct code.

[Hriday Talreja]

@Abderrahim BELISSAOUI Hi, I stated above that these variables rely on PIV-generated data. Those are image data values. They are always different.

[Hriday Talreja]

@Abderrahim BELISSAOUI it is essentially a cell in which each entry in its column is a matrix corresponding to velocity measurements

[dpb]

What it is "essentially" is of little help -- unless you can supply a dataset that illustrates the problem, you're asking folks to do your debugging for you completely from scratch. That's just not a reasonable expectation for code given with no explanation of what it is trying to do nor even a single comment.

The condition

if (max(isnan(F)) == 0) && (min(F)<0)

looks flaky to me at best -- the first could more concisely be written as

if all(isfinite(F))

but if these are correlations, one wouldn't expect a negative necessarily.

Again, it's going to depend upon the data although it also appears you may not be defining everything although again without data it's too tough to try to debug.

Use the debugger and set breakpoints and see where your logic fails to produce what you expect.