Otsu Thresholding for Crack Segmentation in Aircraft Fuselage Surface Images

Project info

• Field of studyAerospace and Aviation
• LevelUndergraduate thesis (S1)
• Method typeImage Processing
• Estimated build time± 3 working days
• Toolsgraythresh, imbinarize, bwlabel, regionprops (Image Processing Toolbox)
• PriceUS$19 - US$116

Research goal

Segment fuselage surface image with Otsu thresholding and assess the influence of grey-level threshold and illumination intensity.

Variables

Example result

• 0.455Otsu threshold
• 0.829Effectiveness metric
• 0.99Dice coefficient
• 0.998Pixel accuracy
• 8.451Foreground (%)
• 5Connected objects

The Otsu threshold of 0.455 separates objects from background in Fuselage surface image with a Dice coefficient of 0.990 and pixel accuracy of 0.998; the foreground covers 8.5% of the image in 5 components.

Computed from illustrative synthetic data, not from a real study.

About Otsu Thresholding and Segmentation

Otsu thresholding automatically chooses an intensity threshold separating object and background by minimising within-class variance. It is used as a first segmentation step, followed by cleaning and labelling objects.

Steps

1. 1Read the grayscale image and check its histogram
2. 2Compute the threshold with graythresh
3. 3Binarise with imbinarize
4. 4Clean the result (bwareaopen, imfill)
5. 5Label objects with bwlabel and measure with regionprops
6. 6Evaluate segmentation with the Dice coefficient if a reference exists

Assumptions to check

• The image histogram is roughly bimodal (object and background)
• Illumination is uniform; otherwise use adaptive thresholding
• Objects are large relative to noise

Reading and reporting the results

Report the threshold, the number of objects, and the size of segmented objects plus a quality measure such as Dice. Explain the method's limits under uneven illumination.

Example MATLAB code

A short starting script with this project's variable names. The delivered version is adapted to your data and fully commented.

rng(2);
[xx, yy] = meshgrid(1:256);
truth = ((xx-60).^2 + (yy-70).^2 < 25^2) | ((xx-150).^2 + (yy-80).^2 < 30^2) | ((xx-110).^2 + (yy-190).^2 < 35^2);
grey_level_threshold = 0.3 * xx / 256;                                          % uneven background gradient
fuselage_surface_image = mat2gray(0.25 + 0.4*truth + grey_level_threshold + 0.06*randn(256));

level = graythresh(fuselage_surface_image);
bw = imbinarize(fuselage_surface_image, level);
bw = bwareaopen(bw, 100);
bw = imfill(bw, 'holes');
bwa = imbinarize(fuselage_surface_image, adaptthresh(fuselage_surface_image, 0.4));                  % adaptive threshold for comparison

dice = @(a, b) 2*nnz(a & b) / (nnz(a) + nnz(b));
fprintf('Otsu level = %.3f\n', level);
fprintf('Dice (Otsu) = %.3f, Dice (adaptive) = %.3f\n', dice(bw, truth), dice(bwa, truth));
L = bwlabel(bw);
stats = regionprops(L, 'Area', 'Centroid');
fprintf('Objects found: %d, areas: %s\n', numel(stats), mat2str([stats.Area]));

figure;
subplot(1,3,1); imshow(fuselage_surface_image); title('Image');
subplot(1,3,2); imshow(bw); title('Otsu + cleaning');
subplot(1,3,3); imhist(fuselage_surface_image); hold on; xline(level, 'r', 'LineWidth', 1.5);

Packages and prices

Choose how much help you need. Each higher package includes everything in the lower ones. Prices exclude 11% VAT.

• Payment: 50% down payment before work starts, balance after the deliverables are handed over
• Tax: Prices exclude 11% VAT
• Data and academic integrity: Results are computed from your own research data; you write and stand behind your manuscript, we help with the analysis and your understanding of it
• Revisions: Up to 2 revisions are included in the price (Package 3: 3 rounds)

Frequently asked questions