Droplet Dynamic Dimension determination image processing

Hello, I have a code there is a droplet separating from a injector and it flows and change shapes. When you run the code it asks you to crop the image around the area which the droplet is moving and then the rest of the code executes. I used to use a circle for recognition of the droplet, now since the droplet is changing shape, my code is not working and I want to count pixels of my pixel! can someone help me with this? The part that says CODE BELOW in a comment mode I want to add my image. I am only attaching one image of 175 images. I just need a hint how to do such thing except from circling the droplet which is from some image forward is not circular anymore.

analysis part.

clear 
clc
close all


AutoThreshold = true;
CropImage = true;
Shadowgraphy = true;                    %// For OH chemiluminescence = false
ExportTracking = false;
StartTime = 50;                         %// Start time of video data in ms

%//-------------------------------------------------------------------------------------------------------------------------------------//

folder = pwd;

if AutoThreshold == false
    
    prompt = {'Enter Threshold value (0-1):'};
    boxTitle = 'Threshold';
    definput = {'Shadowgraphy: 0.2 / OH chemiluminecsence: 0.5'};
    threshold = str2double(inputdlg(prompt,boxTitle,[1 50],definput));
end

imageFiles = [dir([folder '\*.jpg']);dir([folder '\*.png'])];
numFiles = length(imageFiles);
data(1:numFiles,1:3) = NaN;

for trackingNo = 0:9                %// check if old tracking exist and make new folder
    
    if trackingNo == 0 && exist([folder '\Tracking'], 'dir')
        exportPath = ['\Tracking_' num2str(trackingNo+1)];
    elseif trackingNo == 0 && ~exist([folder 'Tracking'], 'dir')
        exportPath = ['\Tracking_' num2str(trackingNo+1)];
    end
    
    if trackingNo > 0 && exist([folder '\Tracking_' num2str(trackingNo)], 'dir')
        exportPath = ['\Tracking_' num2str(trackingNo+1)];
    end
    
end

mkdir(folder, exportPath);

for x = 1:numFiles

    %// Calculate actual time
    if Shadowgraphy == true
        
        time = (x-1)/2000*1000+StartTime;
    else
        
        time = (x-1)/1600*1000+StartTime;
    end
        
    %// Load image & select cropping position
    A = imread([folder '\' imageFiles(x).name]);
       
    if CropImage == true && x == 1
   
        cropFigure = figure('Name','Crop Image: Draw a rectangle','NumberTitle','off','units','normalized','outerposition',[0 0 1 1]);
        imshow(A);
        crop = imrect;
        addNewPositionCallback(crop,@(p) title(mat2str(p,3)));
        fcn = makeConstrainToRectFcn('imrect',get(gca,'XLim'),get(gca,'YLim'));
        setPositionConstraintFcn(crop,fcn); 
        position = wait(crop);
        cropPosition = getPosition(crop);
       
        close(cropFigure);
    end
    
    %// Crop image
    if CropImage == true
        
        cropPosition = round(cropPosition);
        A = imcrop(A,cropPosition);
    end
           
    %// Treshold image
    if AutoThreshold == true
        if Shadowgraphy == true
            
            threshold = isodata(A);
        else
            
            threshold = graythresh(A);
        end
    end
    
    %//---------------CODE BELOW-------------------------------------------------------       
    A1 = A;
    minGL = min(A1(:));
    maxGL = max(A1(:));
    meanGL = mean(A1(:));
    
%      [X, Y] = param_detect02(175,A1,150,maxGL,0,0,0,0.1);
%      
%         A = im2bw(A,threshold);
%        A = bwareafilt(A, 1);
%         BW1 = double(im2bw(A,0.15));
%        [accum, circen, cirrad] = CircularHough_Grd(A1, [15 2000]);
%     BW1 = imfill(A1, 'holes'); 
% %    labeledImage = bwlabel(A);
%        A1 = edge(A1,'sobel',threshold);
    
    %// Save data
    data(x,1) = time;    
    data(x,2) = x;  %radius of circle
    data(x,3) = 0;  %x position of circle
    data(x,4) = 0;  %y position of circle

     %//----------------------------------------------------------------CODE ABOVE------------------------------------------------------
   
   
    %// Display every 10th image
    selector = 1:10:401;    
    if ismember(x,selector)
       
        subplot(4,5,ceil(x/10))
        imshow( A1);
        hold on
%           subplot(circen(1,1), circen(1,2), cirrad, 32, 'g');
        %title(['Image ' num2str(x)],'FontSize',16);
        title(['t = ' num2str(time) ' ms'],'FontSize',16);
    end
    
    if x == numFiles
        
        export_fig([folder exportPath '\Overview.png']);
    end
    
    %// Export images with tracking
    if ExportTracking == true   
        
        figure('Units','normalized','Position',[0 0 1 1])
        imshow(A1);
        hold on
        title(['Image ' num2str(x)],'FontSize',16);
        export_fig([folder exportPath '\Tracking_' num2str(x) '.png']);
        close;
    end
    
    %// Display progress in %
    Progress = round(x/numFiles*100);
end

%// Show & export regression plot
% figure('Name','Diameter Regression Plot','NumberTitle','off','units','normalized','outerposition',[0 0 1 1]);
% plot(data(:,1),data(:,2));
% 
% title('Diameter Regression');
% xlabel('Time [ms]'); 
% ylabel('Diameter [px]');
% export_fig([folder exportPath '\Diameter-Regression.png']);

%// Clear all NaN rows & export data
data = data(all(~isnan(data),2),:);
csvwrite([folder exportPath '\Diameter-Regression.txt'], data);

[darova]

Can you identificate the part you want to calculate the pixels?

[Sepehr Ariaei]

The second part (number 2). I would like to see what is the number of pixels or the area of number 2

[darova]

Can you upload a few more images for my experiments?

[Sepehr Ariaei]

Yes sure, Please find the link in Google Drive to the directory of the whole images.

https://drive.google.com/open?id=1Bo6_JwdW1SPh_4hUxaeAj5DPbVPT1T3Q

[Sepehr Ariaei]

And thank you for helping! I cannot appreciate enough

[darova]

Can't success. Don't have permission. Other ideas?

[Sepehr Ariaei]

I gave you access. You are Mykola Prykhodko right?

Answer 1

[darova]

Here is an idea:

• i cropped region of interest and reduced/dilated to separate each area

• i expanded each area separately in for loop

• then i want to ask user to pick area of interest

• choose correct threshold for your image

graythresh-0.3 graythresh-0.0

I1 = im2bw(I,graythresh(I)-0.1);    % binarize

[Sepehr Ariaei]

Thank you very much for helping me on this. There are some drawbacks. firstly it does not detect the last images where the droplet is a little fading and also at some parts when there is a overlap between the droplet and injector, it will count the pixels of the injector too. Moreover, can you give me a little idea of how can this selection of the part to count pixels can become more autonomous and me not picking the part?

[darova]

Ok. i made some improvements. Here are some notes

• Ideally after processing there are only two blobs: water and glass. I noticed that glass is always remains at the same place. So correctly identificate waterdrop i added checking

P     = [100 85];                   % coordinates of glass

• Sometimes when waterdrop is small reduce size of disk (dilation)

se1   = strel('disk',12,6);         % put here 8 instead of 12

• When waterdrop is large increase strel element to separate it from glass

se1   = strel('disk',16,6);         % put here 16 instead of 12

Just put this script inside folder with images. Let me know if something is wrong

[Sepehr Ariaei]

Dear Darova,

Thank you for your code it is working very good when you play around a little bit with the strel and other numbers! there is only one last problem which I promise it would be the last. I added an " if" to choose different disk size for different versions. The only problem is that when that shadow of water vapor appears after some image around 110 something forward, there is a problem. Instead of detection of the the droplet , it detects the water vapor pixels. How can I deal with that?

[darova]

Please show some of problems images

I assume that you tried to change threshold?

I1    = im2bw(I,0.07);              % binarize

[Sepehr Ariaei]

No actually I have not.

Please find the attached file

[darova]

this line

disp(['ERROR: ', str])

produces something like

ERROR image1
ERROR image2
...

You take all problem images to another folder and process them there

• No actually I have not.

Then you should try