Sobel and Canny Edge Detection for Recognising Object Contours on a Sorting Robot Camera

Project info

• Field of studyRobotics and Mechatronics
• LevelUndergraduate thesis (S1)
• Method typeImage Processing
• Estimated build time± 3 working days
• Toolsedge, imgaussfilt, bwdist (Image Processing Toolbox)
• PriceUS$22 - US$141

Research goal

Apply Sobel and Canny edge detection to the conveyor object image while tuning the edge threshold and the smoothing sigma to assess processing quality.

Variables

Example result

• 0.998Sobel F1
• 1Canny F1
• 1.444Sobel edge pixels (%)
• 1.62Canny edge pixels (%)
• 0.3Canny high threshold
• 0.239Mean gradient magnitude

On the noisy Conveyor object image image the Canny detector reaches an F1 of 1.000 against ground-truth edges versus 0.998 for Sobel; Canny marks 1.6% of pixels as edges (high threshold 0.300).

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

About Sobel and Canny Edge Detection

Sobel and Canny edge detection find object boundaries in images through sharp intensity changes. Sobel is simple and fast, while Canny is more accurate with thinning and double thresholds. They are used in image processing for inspection, measurement, and segmentation.

Steps

1. 1Read the image and convert to grayscale if needed
2. 2Reduce noise with a Gaussian filter
3. 3Apply edge with the sobel method
4. 4Apply edge with the canny method and set thresholds and sigma
5. 5Compare results visually and quantitatively (precision and recall)
6. 6Choose parameters according to application needs

Assumptions to check

• The image has edge contrast sufficient against noise
• Illumination is fairly uniform
• Threshold parameters are tuned on a representative image

Reading and reporting the results

Show a comparison of both methods and explain the parameters used. If a reference image exists, report edge precision, recall, and F-score.

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(1);
[xx, yy] = meshgrid(1:256);
clean = double((xx-128).^2 + (yy-128).^2 < 60^2) + 0.5*double(xx > 180 & yy > 40 & yy < 100);
conveyor_object_image = mat2gray(imgaussfilt(clean, 1) + 0.08*randn(256));    % noisy image
edge_threshold = [0.05 0.15];                                           % Canny thresholds

gt = edge(clean, 'canny');                                      % reference edges
e_sobel = edge(conveyor_object_image, 'sobel');
e_canny = edge(conveyor_object_image, 'canny', edge_threshold, 2);

names = {'Sobel', 'Canny'};  E = {e_sobel, e_canny};
for k = 1:2
    e = E{k};
    prec = nnz(e & (bwdist(gt) <= 2)) / max(nnz(e), 1);
    rec = nnz(gt & (bwdist(e) <= 2)) / nnz(gt);
    fprintf('%s: precision %.3f, recall %.3f, F1 %.3f\n', names{k}, prec, rec, 2*prec*rec/(prec + rec));
end

figure;
subplot(1,3,1); imshow(conveyor_object_image); title('Noisy image');
subplot(1,3,2); imshow(e_sobel); title('Sobel');
subplot(1,3,3); imshow(e_canny); title('Canny');

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