Looking for a suitable image and video processing method for seam tracking

Hi everybody, I am trying to implement a seam tracking algorithm using image processing (please check attached files); I have tried many methods that are working but I still think that I can do better (yet my codes aren't always working properly and I get some bugs). I attach images and if you want to see a video of the process please contact me ([e-mail]) I want two detect the disks and mark a point between them as the seam. ROI is shown in the circle and the pixels that are on the left side of the red line, are not a part of ROI and must be ignored and I need to track the green point in a live video. (If you think you can help me, please email me to send a video file.) P.S: I attached the file of my code, please check and leave a comment if you know a better way to track the seam position.

[Image Analyst]

Attach a short video in a .zip file, or else attach an image for which your algorithm does not work well.

[Image Analyst]

What exactly is the disc? Is it the dumbbell-shaped blob in the middle? What do you need to measure? The weighted centroid of the dumbbell for each frame?

Answer 1

[yanqi liu]

sir，please check the follow code to get some information

for more information,you can read the book 《计算机视觉与深度学习实战》，Thank you！

% This code opens all the frames of the video and pre-precesses them
clc
clear all
close all
%%
% Read the video file
V = VideoReader('video.mpg');
fcount = 0;

while hasFrame(V)
    frame = readFrame(V);
    im = rgb2gray(frame);
    bw = im2bw(im, 0.8);
    [L,num] = bwlabel(bw);
    stats = regionprops(L);
    for j = 1 : num
        recti = stats(j).BoundingBox;
        % target filter
        if recti(4)>size(bw,1)*0.1 || recti(2) < size(bw,1)*0.3
            bw(L==j) = 0;
        end
    end
    [~,num] = bwlabel(bw);
    if num < 2
        bw = im2bw(im, 0.85);
        [L,num] = bwlabel(bw);
        stats = regionprops(L);
        for j = 1 : num
            recti = stats(j).BoundingBox;
            % target filter
            if recti(4)>size(bw,1)*0.1 || recti(2) < size(bw,1)*0.3
                bw(L==j) = 0;
            end
        end
    end
    bw = logical(bw);
    % other filter
    bw2 = imclose(bw, strel('line', round(size(bw,1)*0.2), 90));
    bw2 = bwareaopen(bw2, 100);
    [r,c] = find(bw2);
    [~,ind] = min(c);
    if ~isempty(ind)
        bw2 = bwselect(bw2, c(ind), r(ind));
        bw = logical(bw.*bw2);
    end
    [L,num] = bwlabel(bw);
    if num > 2
        bw = bwareaopen(bw, 3);
        [L,~] = bwlabel(bw);
    end
    stats = regionprops(L);
    rects = cat(1, stats.BoundingBox);
    [~,ind] = sort(rects(:,2));
    bw(L~=ind(1)&L~=ind(2)) = 0;
    bw = logical(bw);
    [L,num] = bwlabel(bw);
    stats = regionprops(L);
    cens = cat(1, stats.Centroid);
    frame2 = insertMarker(frame,cens,'x','color','red','size',6);     
    figure(1); imshow(frame2, []);
    fcount = fcount +1
    title(num2str(fcount));
    pause(0.00001);    
end

[yanqi liu]

% This code opens all the frames of the video and pre-precesses them
clc
clear all
close all
%%
% Read the video file
V = VideoReader('video.mpg');
fcount = 0;

while hasFrame(V)
    frame = readFrame(V);
    im = rgb2gray(frame);
    bw = im2bw(im, 0.8);
    [L,num] = bwlabel(bw);
    stats = regionprops(L);
    for j = 1 : num
        recti = stats(j).BoundingBox;
        % target filter
        if recti(4)>size(bw,1)*0.1 || recti(2) < size(bw,1)*0.3
            bw(L==j) = 0;
        end
    end
    [~,num] = bwlabel(bw);
    if num < 2
        bw = im2bw(im, 0.85);
        [L,num] = bwlabel(bw);
        stats = regionprops(L);
        for j = 1 : num
            recti = stats(j).BoundingBox;
            % target filter
            if recti(4)>size(bw,1)*0.1 || recti(2) < size(bw,1)*0.3
                bw(L==j) = 0;
            end
        end
    end
    bw = logical(bw);
    % other filter
    bw2 = imclose(bw, strel('line', round(size(bw,1)*0.2), 90));
    bw2 = bwareaopen(bw2, 100);
    [r,c] = find(bw2);
    [~,ind] = min(c);
    if ~isempty(ind)
        bw2 = bwselect(bw2, c(ind), r(ind));
        bw = logical(bw.*bw2);
    end
    [L,num] = bwlabel(bw);
    if num > 2
        bw = bwareaopen(bw, 3);
        [L,~] = bwlabel(bw);
    end
    stats = regionprops(L);
    rects = cat(1, stats.BoundingBox);
    [~,ind] = sort(rects(:,2));
    bw(L~=ind(1)&L~=ind(2)) = 0;
    bw = logical(bw);
    [L,num] = bwlabel(bw);
    stats = regionprops(L);
    cens = mean(cat(1, stats.Centroid), 1);
    frame2 = insertMarker(frame,cens,'*','color','green','size',6);
    figure(1); imshow(frame2, []);
    fcount = fcount +1
    title(num2str(fcount));
    pause(0.00001);    
end

[yanqi liu]

sir, use some tracking method to develop, such as compressive tracking, particle filter tracking

for more information,you can read the book 《计算机视觉与深度学习实战》，Thank you！

[yanqi liu]

yes，may be give me an email address，i would send the zip file

[yanqi liu]

sorry, the email can not seed success，so upload，please check