How can i track points in two images?

Currently i fed two points to the first image,how i know where those points locate on a second image?

 if true
originalRGBImage = imread('a.png');
imshow(originalRGBImage) ;
hold on
% GEt boundary of the human 
[boundary_y,boundary_x] = find(~I) ;
idx = boundary(boundary_x,boundary_y) ;  
% Boundary points 
boundary_x = boundary_x(idx) ; boundary_y =boundary_y(idx) ;
imshow(I) ;
hold on
%draw boundary line
 plot(boundary_x,boundary_y,'b') ;
%Get Tracking Points from the user
uiwait(msgbox('Locate the starting point of the neck '));
[neck_X,neck_Y] = ginput(1);
loacationEar=plot(neck_X,neck_Y,'r+', 'MarkerSize', 10); 
uiwait(msgbox('Locate the starting point of your spine?'));
[spine_X,spine_Y] = ginput(1);
loacationSpine=plot(spine_X,spine_Y,'r+', 'MarkerSize', 10); 
uiwait(msgbox('Locate the point in your knee?'));
[knee_X,knee_Y] = ginput(1);
hold on;
loacationknee1=plot(knee_X,knee_Y,'r+', 'MarkerSize', 10); 
uiwait(msgbox('Locate the point in below the knee2?'));
[belowKnee_X,belowKnee_Y] = ginput(1);
loacationknee2=plot(belowKnee_X,belowKnee_Y,'r+', 'MarkerSize', 10); 
uiwait(msgbox('Locate the point in above the knee through the thie?'));
[aboveKnee_X,aboveKnee_Y] = ginput(1);
loacationknee3=plot(aboveKnee_X,aboveKnee_Y,'r+', 'MarkerSize', 10); 
I = im2bw(originalImage);
[boundary_y,boundary_x] = find(~I) ;
idx = boundary(boundary_x,boundary_y)
% Boundary points
boundary_x = boundary_x(idx) ; boundary_y =boundary_y(idx) ;
%draw boundary line
plot(boundary_x,boundary_y,'b');
  end

Image which i fed Points

Image which i need to find where those above points in locate?

[KALYAN ACHARJYA]

You want the coordinates of "Two Plus-Red Position"?

[cha bha]

I need to find new coordinates of the red markers, (I need to find where those red markers locate in second image)

[KALYAN ACHARJYA]

Are both images are of the same size?

[cha bha]

yes.same size. Actually its a 2 frames in a video,Only difference is human body lift more than first image

Answer 1

[Image Analyst]

Maybe try surf() in the Computer Vision System Toolbox. Feature based object detection

[cha bha]

I checked with surf,but couldn't find a way to give points manually to detect, Is there any way to provide points which need to track?

[Image Analyst]

Can you explain in words what defines the point you want to track? Like it's the highest non-white pixel at a particular column or whatever?

[cha bha]

In first image i marked two points using maker(+),I need to find and get those points loacations in the second image?(x and y values of those points) In other words, i need to track where those red markers in second image,

[Image Analyst]

With a new image, those points won't be there of course. What rule did you use to decide where to put the markers? Do you just want to take the highest red point (assuming all people wear a red shirt)?

Do you want to fit a quadratic along all of the top red points to see how straight the back is? Or a line to find it's angle?

[cha bha]

Actually I try to check whether this exercise formation correct or not,

1 Step - I take the video clip and extract it to frames

2 step - Then i take the first frame from saved frames

3 Step - User need to mark those two points using mouse

4 Step - I get those details and find whether body boundary is lies on above or below,

Step 5- Then I need take another frame and check it also satisfy those condition,Currently i dont have a way to detect those points in automatically in second image,

I tried to use KLT algorithm to this,But couldn't find a way to give points to video manually and then track it through the video,

[cha bha]

Yes i need to check whether back is in straight or not in second image also, But the t shirt color should not depend on it.

[Image Analyst]

You can get the top row like this

for col = 1 : size(rgbImage, 2)
    thisColumn = rgbImage(:, col, 1);
    topRow(col) = find(thisColumn <= 255, 1, 'first');
end
plot(topRow, 'b-', 'LineWidth', 2);
grid on;
xlabel('column', 'FontSize', 20);
ylabel('Top most row', 'FontSize', 20);

You might need to examine the top row pixels to find out where the color changes substantially, like from shirt to shorts or shirt to neck or hair.

[cha bha]

Is there a way to do with above mention scenario using KLT Algorithm?

if true
 videoFileReader = vision.VideoFileReader('deadlift.mp4');
videoPlayer = vision.VideoPlayer('Position',[1,10,1280,720]);

 objectFrame = videoFileReader();
% objectRegion = [264,122,93,93];
% 
figure; 
imshow(objectFrame);
% objectRegion=round(getPosition(impoint));
% Show initial frame with a red bounding box.
objectImage = insertMarker(objectFrame,'+',objectRegion,'Color','red'); 
figure;
imshow(objectImage);
title('Red box shows object region');
points = detectMinEigenFeatures(rgb2gray(objectFrame));
pointImage = insertMarker(objectFrame,points.Location,'+','Color','white');
figure;
imshow(pointImage);
title('Detected interest points');
tracker = vision.PointTracker('MaxBidirectionalError',1);
initialize(tracker,points.Location,objectFrame);
while ~isDone(videoFileReader)
      frame = videoFileReader();
      [points,validity] = tracker(frame);
      out = insertMarker(frame,points(validity, :),'+');
      videoPlayer(out);
      disp(points);
end
  end

I try with this,but not able to do add points manually to this code,Is there way to add points manually to this KLT algorithm?

[cha bha]

https://in.mathworks.com/matlabcentral/answers/402726-how-can-i-track-points-in-two-images#comment_572451 I tried with above code but didnt get any result on top of the image.I get somwthing like this,

How to check whether below shown part was straight or not?

Answer 2

[Image Analyst]

Try this:

clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
clearvars;
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 20;

fullFileName = 'D:\Matlab\work\Tests\image2.png';
[folder, baseFileName, ext] = fileparts(fullFileName);

%=======================================================================================
% Read in demo image.
rgbImage = imread(fullFileName);
% Shrink it to speed it up
% rgbImage = imresize(rgbImage, 0.75);
% Get the dimensions of the image.
[rows, columns, numberOfColorChannels] = size(rgbImage);
% Display the original image.
subplot(2, 3, 1);
imshow(rgbImage, []);
axis on;
caption = sprintf('Original Color Image, %s', baseFileName);
title(caption, 'FontSize', fontSize, 'Interpreter', 'None');
drawnow;
hp = impixelinfo(); % Set up status line to see values when you mouse over the image.

% Set up figure properties:
% Enlarge figure to full screen.
set(gcf, 'Units', 'Normalized', 'OuterPosition', [0 0.05 1 0.95]);
% Get rid of tool bar and pulldown menus that are along top of figure.
% set(gcf, 'Toolbar', 'none', 'Menu', 'none');
% Give a name to the title bar.
set(gcf, 'Name', 'Demo by ImageAnalyst', 'NumberTitle', 'Off')

% Convert to a grayscale image by taking the min of all 3 color channels.
grayImage = min(rgbImage, [], 3);
% Display the original image.
subplot(2, 3, 2);
imshow(grayImage, []);
axis on;
caption = sprintf('Gray Scale Image, %s', baseFileName);
title(caption, 'FontSize', fontSize, 'Interpreter', 'None');
drawnow;
hp = impixelinfo(); % Set up status line to see values when you mouse over the image.

% Compute the histogram
subplot(2, 3, 3);
histogram(grayImage, 'BinWidth', 1);
grid on;

% Find the big central blob
mask = grayImage < 200;
% Fill the image.
mask = imfill(mask, 'holes');
% Extract just the biggest blobs.
mask = bwareafilt(mask, 1);
% Display the mask image.
subplot(2, 3, 4);
imshow(mask, []);
axis on;
caption = sprintf('Thresholded image');
title(caption, 'FontSize', fontSize, 'Interpreter', 'None');
drawnow;
hp = impixelinfo(); % Set up status line to see values when you mouse over the image.

% Find the top row in each column
topRow = ones(1, columns); % Preallocate.
for col = 1 : columns
  thisColumn = mask(:, col);
  tr = find(thisColumn, 1, 'first');
  if ~isempty(tr)
    topRow(col) = find(thisColumn, 1, 'first');
  end
end
subplot(2, 3, 5);
plot(topRow, 'b-', 'LineWidth', 2);
grid on;
xlabel('column', 'FontSize', 20);
ylabel('Top most row', 'FontSize', 20);

% Plot over original image.
subplot(2, 3, 6);
imshow(rgbImage, []);
axis on;
caption = sprintf('Original Color Image, %s', baseFileName);
title(caption, 'FontSize', fontSize, 'Interpreter', 'None');
drawnow;
hp = impixelinfo(); % Set up status line to see values when you mouse over the image.
hold on;
plot(topRow, 'b-', 'LineWidth', 2);

Now that you have the top row you'll have to examine it to find the part of it that identifies the back, like where it starts and stops. Can you do that?

[cha bha]

I have no idea how to find the start and end point which i need,It depend on the image also.I need to know where the neck start and spine start,

Or is there any way to find back goes a straight line before the neck start using this top row, Thanks for your help and guidance

If i able to get a and b point it was easy to do my calculation

[Image Analyst]

Well what is your idea? Do you want to develop your program yourself or do you just want to hire someone to do it for you? I'm assuming you want to write the program and would rather not have someone do everything for you. So what do you think might work? Maybe assume the person is wearing a shirt of some color and that the shirt is along the top pixels for a longer stretch than the head or pants. So go along the top pixel and look at the pixel just below. Find out what colors are there. The color that is there the most you can assume is the shirt. Then extract the shirt pixels and fit them to a line or quadratic with polyfit(). What do you think? Do you have a better idea?

[cha bha]

Ok i will check with that, I have another idea not sure its possible, First need to get a correct graph for correct formation, Then get the graph from above result(which need to decide correct or not) After that compare that graph design with the correct graph design and check whether it has a similar graph path or not,Is it Possible to do or not?

[Image Analyst]

Sure you can compare (your step 3) the actual shape (step 2) to the correct or preferred shape (step 1), but you still have to identify the parts of the plot that you want to use in the comparison. In other words, you still need to identify the back and exclude the head, even if you do plan on comparing with a reference/correct shape later in step 3. So you still need to flesh out the details of step 2 a little bit. And you need to define features of the back elements you extract since the number of those won't necessarily be the same number of elements as the reference curve. For example you need to coefficients of the polynomial or whatever.

[cha bha]

Ok i will try with that, Another thing was,I tried to track user given point through the video,I able to track manually given point,How can i track several point?

This was the code to track one point given using marker.

if true
 videoFileReader = vision.VideoFileReader('deadlift.mp4');
videoPlayer = vision.VideoPlayer('Position',[1,10,1280,720]);

 objectFrame = videoFileReader();
% objectRegion = [264,122,93,93];
% 
figure; 
imshow(objectFrame);
[neck_X,neck_Y]= ginput(1);
% imshow(pointImage);
% points = detectMinEigenFeatures(rgb2gray(objectFrame));
pointImage1 = insertMarker(objectFrame, [neck_X,neck_Y],'+','Color','white');
figure;
imshow(pointImage1);
title('Detected interest points');
tracker = vision.PointTracker('MaxBidirectionalError',1);
initialize(tracker,[neck_X,neck_Y],objectFrame);
while ~isDone(videoFileReader)
      frame = videoFileReader();
      [points,validity] = tracker(frame);
      out = insertMarker(frame,points(validity, :),'+');
      videoPlayer(out);
      disp(points);
end
  end

I need to add several point to this,How can i do it?