Fit a line on a plot (imagesc)
23 views (last 30 days)
I have a plot as the following:
Which comes from the command imagesc on a 101x101 square matrix. The matrix values are either 1 (yellow region) or 0 (blue region); however, their values aren't important actually. The only thing that matters is the line that seperates these two regions (quarter-circle-like line). Is there any way to estimate the equation of this line?
Thanks in advance!
LO on 14 May 2021
Edited: LO on 14 May 2021
contrast_line = diff(image);
imshow(contrast_line); % shows the line
thresholded_image=mean(contrast_line,3) > 240; % set here a value between 0 and 255 that would fit with your contrast line. It could be found perhaps with max(max(max(contrast_line)))
imshow(thresholded_image) % show the thresholded difference, in case you need to extract only the points above a certain threshold (or below, in case you set a < in the line above)
%% the other lines below require this function (see link here below)
%% based on this post (https://de.mathworks.com/matlabcentral/answers/512870-how-to-perform-robust-line-fitting-in-a-binary-image)
p_poly=polyfit(x,y,1);%polyfit for reference
'DisplayName',sprintf('R^2=%.2f (fit_line)',get_r_square(p_best ,x,y)))
'DisplayName',sprintf('R^2=%.2f (polyfit)',get_r_square(p_poly ,x,y)))
axis([1 size(thresholded_image,2) 1 size(thresholded_image,1)])
%Fit a line to the data. Use the RMS of the orthogonal distance as a cost
%function instead of MSE_y, as polyfit probably does.
%This works with fminsearch, which is sensitive to initial values that are
%far from the optimum, sometimes returning local optima.
%root mean square of the distance between the line defined by the two
%points in v and the points defined by x and y.
cost_fun=@(v) RMS(point_to_line_distance(pt, v([1 3]), v([2 4])));
%convert [x1 y1 phi2] to [x1 y1 x2 y2]
vr_2_v=@(vr) [vr(1:2) vr(1:2)+[cos(vr(3)) sin(vr(3))]];
%wrap the cost function and the converter
%initialize to around the center of the data
init=[mean(x) mean(y) 0*pi];
opts = optimset('MaxFunEvals',50000, 'MaxIter',10000);
fit_val = fminsearch(fun, init, opts);