Determining crack length from images

Hello all,

I have been looking at several Q/As on here about detecting cracks and measuring their lengths from images (and eventually video). I think I have a good start, but I am struggling to segment the image so I can remove unnecessary information. I have used the image segmentation tool, but the cracks are so small that it is difficult to keep them in the binary image without a lot of extra stuff.

The end goal is to measure the length of the crack from its initiation point (in this case the center of the image), and then the angle from horizontal to calculate the horizontal growth of the crack (vertical does not matter for what I am doing). I have attached a sample image and the code I have developed so far, but I was wondering if anyone had insight into how I could better segment this image and make the measurements?

If there is a good way to identify the end point and draw a line from the origin to the end of the crack that would probably be perfect.

list = dir("/MATLAB Drive/Images/");
nFiles = numel(list)-2;

tlX_left = 275;
tlY_left = 850;
tlX_right = 2050;
tlY_right = 850;
width = 1600;
height = 300;

for i = 1:nFiles
    filename = fullfile(list(i+2).folder,list(i+2).name);
    I = imread(filename);
    [BW, maskedImage] = segmentImage(I);
    BW2_left = imcrop(BW, [tlX_left tlY_left width height]);
    BW2_right = imcrop(BW, [tlX_right tlY_right width height]);
    imshowpair(BW2_left,BW2_right, "montage")

end

function [BW,maskedImage] = segmentImage(RGB)
%segmentImage Segment image using auto-generated code from Image Segmenter app

% Threshold image with adaptive threshold
BW = imbinarize(im2gray(RGB), 'adaptive', 'Sensitivity', 0.000000, 'ForegroundPolarity', 'bright');

% Dilate mask with default
radius = 3;
decomposition = 0;
se = strel('disk', radius, decomposition);
BW = imdilate(BW, se);

% Create masked image.
maskedImage = RGB;
maskedImage(repmat(~BW,[1 1 3])) = 0;
end

Answer 1

[Pratyush Swain]

Hi Alexander,

I can infer from the information given that there is a need for better segmentation of image in order to remove noises and small objects. One aspect where the image can be processed better is by leveraging the "bwareaopen" function.It helps to remove remove small objects from binary image.

Result after using "bwareaopen" function:

Previous Result:

After segmentation, we can apply hough transforms to image in order to detect the crack lines:

    % Existing implementation
    I = imread(filename);
    [BW, maskedImage] = segmentImage(I);
    BW2_left = imcrop(BW, [tlX_left tlY_left width height]);
    BW2_right = imcrop(BW, [tlX_right tlY_right width height]);
    % -------------------------
    
    RGB_left = imcrop(I, [tlX_left tlY_left width height]);
    RGB_right = imcrop(I, [tlX_right tlY_right width height]);
    
    % Using bwareaopen to eliminate small noises
    BW2_left = bwareaopen(BW2_left, 100); % Adjust the size threshold as needed
    BW2_right = bwareaopen(BW2_right, 100);

    figure,
    imshowpair(BW2_left,BW2_right, "montage")  

    % Extracting Hough Lines
    lines_left = HoughLines(BW2_left);
    lines_right = HoughLines(BW2_right);
    
    %Draw hough lines for BW2_left
    figure, imshow(RGB_left), hold on
    max_len = 0;
    for k = 1:length(lines_left)
       xy = [lines_left(k).point1; lines_left(k).point2];
       plot(xy(:,1),xy(:,2),'LineWidth',2,'Color','red');
    
       % Plot beginnings and ends of lines
       plot(xy(1,1),xy(1,2),'x','LineWidth',2,'Color','green');
       plot(xy(2,1),xy(2,2),'x','LineWidth',2,'Color','green');
    
       % Determine the endpoints of the longest line segment
       len = norm(lines_left(k).point1 - lines_left(k).point2);
       if ( len > max_len)
          max_len = len;
          xy_long = xy;
       end
    end
    % Display max length on left hand side image
    disp("max length on left side: ")
    disp(max_len)
    hold off
    img_frame = getframe(gca);
    BW2_left = img_frame.cdata;
    
    % Draw hough lines for BW2_right
    figure, imshow(RGB_right), hold on
    max_len = 0;
    for k = 1:length(lines_right)
       xy = [lines_right(k).point1; lines_right(k).point2];
       plot(xy(:,1),xy(:,2),'LineWidth',2,'Color','red');
    
       % Plot beginnings and ends of lines
       plot(xy(1,1),xy(1,2),'x','LineWidth',2,'Color','green');
       plot(xy(2,1),xy(2,2),'x','LineWidth',2,'Color','green');
    
       % Determine the endpoints of the longest line segment
       len = norm(lines_right(k).point1 - lines_right(k).point2);
       if ( len > max_len)
          max_len = len;
          xy_long = xy;
       end
    end
    
    % Display max length on right hand side image
    disp("max length on right side: ")
    disp(max_len)
    hold off
    
    img_frame = getframe(gca);
    BW2_right = img_frame.cdata;

    % Show final result
    figure,
    imshowpair(BW2_left,BW2_right, "montage")    

Extracting Hough lines implementation:

function lines = HoughLines(BW)
    % Canny edges
    BWc = edge(BW,'canny');
    [H,T,R] = hough(BWc);
    
    % Hough Peaks
    P  = houghpeaks(H,5,'threshold',ceil(0.3*max(H(:))));
    
    % Hough lines (parameters can be adjusted) 
    lines = houghlines(BW,T,R,P,'FillGap',5,'MinLength',7);
end

Final result:

These results can be improved more with better segmentation techniques and parameter tuning.

For more information on "bwareopen" function and "Hough Transforms", please refer to following links:

1.https://www.mathworks.com/help/images/ref/bwareaopen.html

2.https://www.mathworks.com/help/images/hough-transform.html

Answer 2

[Preetham Manjunatha]

Here is the MATLAB code to calculate/estimate the crack area, width and length. In addition, this package assumes the crack is segmented either using morphological method or multiscale gradient-based or deep learning semantic segmentation methods. This package estimates the crack area, width and length (pixel scale can be provided to estimate these physical quantities).