MATLAB Answers

Fit image datapoints to curve

2 views (last 30 days)
Johannes Hope Denstadli
Johannes Hope Denstadli on 3 Feb 2020
Edited: Matt J on 3 Feb 2020
I'm trying to control a robot from an ultrasonic greyscale image. So far I've used Sobel filtering to get a binary image, and Hough transform to detect line segments.
Now I want to create a control output to make the midpoints of the detected Houghlines aline most accurately with the drawn arc. The robot can move in the X and Y direction as well as rotate. The reference values I get from the code beneath contains [x-coordinate of midpoint, y-coordinate of midpoint, angle of the line segment, distance from midpoint to arc]. Does anyone know how to obtain the optimal rotation and movement to aline the points with the arc?
I = imread('Weld_ultrasound.png');
I = rgb2gray(I);
BW2 = edge(I,'Sobel','nothinning');
[H,T,R] = hough(BW2,'Theta',-90:1:89);
P = houghpeaks(H,20,'threshold',ceil(0.2*max(H(:))));
lines = houghlines(BW2,T,R,P,'FillGap',5,'MinLength',250);
function ref = Houghlines2ref(lines)
min_dist = 950; % Set minimum distance
max_dist = 1050; % Set maximum distance
ref = []; % [angle,distance]
for i = 1:length(lines)
mid = (lines(i).point1+lines(i).point2)/2;
if sqrt(mid(1)^2+mid(2)^2) <= min_dist
ref = [ref; [mid,lines(i).theta, (sqrt(mid(1)^2+mid(2)^2) - min_dist)]];
elseif sqrt(mid(1)^2+mid(2)^2) >= max_dist
ref = [ref; [mid,lines(i).theta,(sqrt(mid(1)^2+mid(2)^2) - max_dist)]];
end
end
I = imread('Weld_Ultrasound.png');
I = rgb2gray(I);
figure, imshow(I), hold on
if isempty(ref) == 1
return
else
for k = 1:length(ref(:,1))
p0 = [ref(k,1),ref(k,2)];
[X,Y] = pol2cart(deg2rad(ref(k,3)),-ref(k,4));
xy = [p0; p0+[X,Y]];
plot(p0(1),p0(2),'rs')
plot(xy(:,1),xy(:,2),'LineWidth',1,'Color','green'); % Plot referance vectors
end
end
t = linspace(0,0.5*pi,128); % Plot arc
x = [0 1000*cos(t) 0];
y = [0 1000*sin(t) 0];
hold on
plot( x, y, 'r' )
end
Thanks
Original image;

  0 Comments

Sign in to comment.

Answers (1)

Matt J
Matt J on 3 Feb 2020
Edited: Matt J on 3 Feb 2020
You would use a 2D version of an Iterative Closest Point (ICP) algorithm. This FEX contribution appears to have one:

  0 Comments

Sign in to comment.

Sign in to answer this question.