Matlab code to find the Wave height
15 Ansichten (letzte 30 Tage)
Ältere Kommentare anzeigen
I want to find the water wave height and wave length by analysing the video. I write the code for that :
Is this correct?
videoFile = 'wave_video.mp4'; % Replace with your video file name
video = VideoReader(videoFile);
numFrames = floor(video.Duration * video.FrameRate);
frameRate = video.FrameRate;
displacement = zeros(1, numFrames);
% Define a threshold value for detecting the wave (adjust this as needed)
threshold = 100;
for k = 1:numFrames
frame = readFrame(video);
grayFrame = rgb2gray(frame);
% Threshold the grayscale image to create a binary mask
binaryImage = grayFrame > threshold;
% Sum the binary image to get an estimation of the wave's displacement
displacement(k) = sum(binaryImage(:));
end
time = (0:numFrames-1) / frameRate;
plot(time, displacement);
xlabel('Time (s)');
ylabel('Displacement');
title('Water Wave Displacement Over Time');
How to find the unit of that height.
3 Kommentare
Mathieu NOE
am 10 Sep. 2024
maybe if you could share some video file (a short one) , someone can work on your problem (unless its obvious there's already an answer somewhere)
Antworten (2)
Walter Roberson
am 11 Sep. 2024
image_height = min(sum(cumprod(~binaryImage, 1),1) + 1);
or possibly
image_height = min(sum(cumprod(~flipud(binaryImage), 1),1) + 1);
depending on whether the image starts from the top or from the bottom of the array.
The logic here assumes that the portions without the wave are binary 0. The cumprod(~binaryImage) fills the leading zero elements of each column of the array, producing a "top hat" sort of effect. sum() of that gives the number of leading zeros in each column. min() of that finds the shortest entry -- corresponding to the place the hat was highest.
Perhaps max() would be needed instead of min(), depending how the data is arranged in the images.
Mathieu NOE
am 11 Sep. 2024
hello again
I am by no means an image processing expert, but sometimes I go outside my comfort zone ... hopefully my suggestion is not completely wrong.
One point, all the results given below are "in pixels" dimnesions, so as I commented in first place , there is still some work to do to convert these results into physical units. A very crude first approach would be to identify a know object in the picture and use its dimensions (known) to apply a conversion rule. a more sophisticated approach would require to estimate the position and orientation of the camera , but Ileave that to real professionnals ....
so what can I offer today ?
1/ I tried first to detect the line between water and border (arrow). So there is a lower and upper threshold used on the RGB triplet to convert the image into a binary matrix (water or not). To ease my work I cropped the image to what is inside the red rectangle
2/ then extract the lower boundary of the "non water" objet (ie the border) , detrend it , remove some outliers , and some smoothing on top , and here we get for every frame a more or less accurate image on the water line (for each frame) - this is plotted in the bottom subplot here - the top suplot is the binarized image with the separation in red
3/ from there you can compute some values like mean frequency (interest ?) and evidetly the max / min (= amplitude) of the wave
% mean period (in pixels)
mean_period_in_pixels(k)= 1/meanfreq(y3ds,1); % Fs spatial = 1 pixel
% max wave amplitude (in pixels)
amplitude_in_pixels(k)= max(y3ds) - min(y3ds);
full code :
videoFile = 'wave_video1.mp4'; % Replace with your video file name
video = VideoReader(videoFile);
numFrames = floor(video.Duration * video.FrameRate);
frameRate = video.FrameRate;
% Define a threshold value for detecting the wave (adjust this as needed)
threshold1 = 160;
threshold2 = 190;
for k = 1:numFrames
frame = readFrame(video);
ind1 = (frame(:,:,1)>threshold1) & (frame(:,:,2)>threshold1) & (frame(:,:,3)>threshold1);
ind2 = (frame(:,:,1)<threshold2) & (frame(:,:,2)<threshold2) & (frame(:,:,3)<threshold2);
binaryImage = ind1 & ind2; % here we get the white/gray border
% crop the bottom of the pictures (to remove the light reflection spots)
binaryImage = binaryImage(1:200,150:500);
[y,x] = find(binaryImage); % get x,y coordinates
[x3,y3,x4,y4] = top_bottom_boundary(x,y); %find the separation line between water and border
y3 = filloutliers(y3,'linear','movmedian',50); % remove some outliers
y3d = detrend(y3,'linear'); % remove linear trend (due to camera orientation)
y3ds = smoothdata(y3d,'sgolay',50); % smooth a bit the line
% some plots for debug and demo
if k<3 % plot as many as you want
figure(k),
subplot(211),imagesc(binaryImage)
hold on
plot(x3,y3,'r*');
hold off
subplot(212),plot(x3,y3d,'r*',x3,y3ds,'b');
end
% mean period (in pixels)
mean_period_in_pixels(k)= 1/meanfreq(y3ds,1); % Fs spatial = 1 pixel
% max wave amplitude (in pixels)
amplitude_in_pixels(k)= max(y3ds) - min(y3ds);
end
time = (0:numFrames-1) / frameRate;
figure
plot(time(1:k), amplitude_in_pixels);
xlabel('Time (s)');
ylabel('Displacement (in pixels) ');
title('Water Wave Displacement Over Time');
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [x3,y3,x4,y4] = top_bottom_boundary(x,y)
% based on FEX : https://fr.mathworks.com/matlabcentral/answers/299796-tight-boundary-around-a-set-of-points
%split data into classes and find max/min for each class
class_label = unique(x);
upper_boundary = zeros(size(class_label));
lower_boundary = zeros(size(class_label));
for idx = 1:numel(class_label)
class = y(x == class_label(idx));
upper_boundary(idx) = max(class);
lower_boundary(idx) = min(class);
end
% left_boundary = y(x == class_label(1));
% right_boundary = y(x == class_label(end));
%
% % left_boundary
% x1 = class_label(1)*ones(size(left_boundary));
% y1 = left_boundary;
%
% % right_boundary
% x2 = class_label(end)*ones(size(right_boundary));
% y2 = right_boundary;
% top boundary
x3 = class_label;
y3 = upper_boundary;
% bottom boundary
x4 = class_label;
y4 = lower_boundary;
% x_out = [x1(:);x2(:);x3(:);x4(:);];
% y_out = [y1(:);y2(:);y3(:);y4(:);];
end
1 Kommentar
Mathieu NOE
am 16 Sep. 2024
here a slight improvement over my previous code
the line angle correction is done by rotation and not by linear detrending (which is not a geometrical correction)
% correct tilting angle (due to camera orientation)
p = polyfit(x3,y3,1);
slope = p(1);
alpha = -atan(slope);
x3r = x3*cos(alpha) - (y3-y3(1))*sin(alpha);
y3r = (x3-x3(1))*sin(alpha) + y3*cos(alpha);
hope it helps
videoFile = 'wave_video1.mp4'; % Replace with your video file name
video = VideoReader(videoFile);
numFrames = floor(video.Duration * video.FrameRate);
frameRate = video.FrameRate;
% Define a threshold value for detecting the wave (adjust this as needed)
threshold1 = 160;
threshold2 = 190;
for k = 1:numFrames
frame = readFrame(video);
ind1 = (frame(:,:,1)>threshold1) & (frame(:,:,2)>threshold1) & (frame(:,:,3)>threshold1);
ind2 = (frame(:,:,1)<threshold2) & (frame(:,:,2)<threshold2) & (frame(:,:,3)<threshold2);
binaryImage = ind1 & ind2; % here we get the white/gray border
% crop the bottom of the pictures (to remove the light reflection spots)
binaryImage = binaryImage(1:200,150:500);
[y,x] = find(binaryImage); % get x,y coordinates
[x3,y3,x4,y4] = top_bottom_boundary(x,y); %find the separation line between water and border
y3 = filloutliers(y3,'linear','movmedian',50); % remove some outliers
% correct tilting angle (due to camera orientation)
p = polyfit(x3,y3,1);
slope = p(1);
alpha = -atan(slope);
x3r = x3*cos(alpha) - (y3-y3(1))*sin(alpha);
y3r = (x3-x3(1))*sin(alpha) + y3*cos(alpha);
y3r = y3r - mean(y3r); % remove dc value
y3rs = smoothdata(y3r,'sgolay',50); % smooth a bit the line
% some plots for debug and demo
if k<3 % plot as many as you want
figure(k),
subplot(211),imagesc(binaryImage)
hold on
plot(x3,y3,'r*');
hold off
subplot(212),plot(x3r,y3r,'m*',x3,y3rs,'b');
end
% mean period (in pixels)
mean_period_in_pixels(k)= 1/meanfreq(y3rs,1); % Fs spatial = 1 pixel
% max wave amplitude (in pixels)
amplitude_in_pixels(k)= max(y3rs) - min(y3rs);
end
time = (0:numFrames-1) / frameRate;
figure
plot(time(1:k), amplitude_in_pixels);
xlabel('Time (s)');
ylabel('Displacement (in pixels) ');
title('Water Wave Displacement Over Time');
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [x3,y3,x4,y4] = top_bottom_boundary(x,y)
% based on FEX : https://fr.mathworks.com/matlabcentral/answers/299796-tight-boundary-around-a-set-of-points
%split data into classes and find max/min for each class
class_label = unique(x);
upper_boundary = zeros(size(class_label));
lower_boundary = zeros(size(class_label));
for idx = 1:numel(class_label)
class = y(x == class_label(idx));
upper_boundary(idx) = max(class);
lower_boundary(idx) = min(class);
end
% left_boundary = y(x == class_label(1));
% right_boundary = y(x == class_label(end));
%
% % left_boundary
% x1 = class_label(1)*ones(size(left_boundary));
% y1 = left_boundary;
%
% % right_boundary
% x2 = class_label(end)*ones(size(right_boundary));
% y2 = right_boundary;
% top boundary
x3 = class_label;
y3 = upper_boundary;
% bottom boundary
x4 = class_label;
y4 = lower_boundary;
% x_out = [x1(:);x2(:);x3(:);x4(:);];
% y_out = [y1(:);y2(:);y3(:);y4(:);];
end
Siehe auch
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!