PCB Image Detection in GUI

Hi, I'm trying to create a GUI to detect the defects in PCB. The image shown is the reference image that I would wish to create but I failed to search for any references to create the green boxes as detection and determine the type of defects in GUI. I'm wondering if there are any references I could refer to?

[KIVANÇ mutlu]

I made comparison of fixed pcb images but XOR comparison does not work on samples from different locations. I need to first identify the PCB location using the fiducial reference. I want to teach fiducial points. Then he will search for these points in certain areas and determine his coordinate. According to the first reference, the periodic, that is, sliding of the pcb is determined.

[Image Analyst]

KIVANC, how/where did you get his images? I don't see that he posted them.

Anyway, did you try imregister()? And can you post/attach his images?

[Lee Hao Nam]

Hey, I have managed to created the Boundary box around the defect using the lines of code below:

[L, num] = bwlabel(img3);
bboxes = regionprops(img3,'BoundingBox','centroid');

for k = 1:length(bboxes)
    CurrBB = bboxes(k).BoundingBox;
    rectangle('Position',[CurrBB(1),CurrBB(2),CurrBB(3),CurrBB(4)],'EdgeColor','g','LineWidth',2)
end

The current issue I am facing is how do I classify the types of defects

[Image Analyst]

I don't know what img3 is, but you need to look at the various features of each blob, like it's area, or perimeter, or circularity, or aspect ratio, or something like that.

[Lee Hao Nam]

The images are as attached below.

img3 is the imabsdiff of the 2 images below

[Image Analyst]

OK, so please describe in words, what are the features that describe the various classes, and give the class names, like "nick", "complete break" or whatever.

[Lee Hao Nam]

So the types of defects are such as:

1. Short

2. Open

3. Spurous copper

4. Mouse bite

5. Spur

The codes are as shown below up till the part where creating Bounding Boxes and texts

% Show Fig 1

a = uigetfile('*.*'); % Get all types of files

img1 = imread(a);

img1 = im2bw(img1);

figure()

imshow(img1);

% Show Fig 2

b = uigetfile('*.*'); % Get all types of files

img2 = imread(b);

img2 = im2bw(img2);

figure()

imshow(img2);

% Show Fig 3 of subtracted img1 and img2

figure()

img3 = imabsdiff(img1,img2);

imshow(img2);

[L, num] = bwlabel(img3);

bboxes = regionprops(img3,'BoundingBox','centroid');

% Obtain centroid of the boundary boxes

centroids = cat(1,bboxes.Centroid);

hold on

for k = 1:length(bboxes)

CurrBB = bboxes(k).BoundingBox;

rectangle('Position',[CurrBB(1),CurrBB(2),CurrBB(3),CurrBB(4)],'EdgeColor','g','LineWidth',2)

H= text( centroids(:,1)-9, centroids(:,2)-17,'types of defect'); % goes by column first then row

set(gcf,'DefaultTextColor','red')

end

hold off

[edgar sanchez]

do you have the full project? Im doing this for school. Can you send it too me? the code.

[e-mail]

Today is the due date :)

[Image Analyst]

edgar, I'm not sure who you're asking. I guess you're asking Lee. I don't have his code (other than the partial chunk above) and I don't know if he finished or not. All he did so far was to subtract the test image from the template/reference image and find differences and put boxes around them. He never said (despite me asking) what constitutes the 5 different types of defects, so the classification into those defects is not shown in the above code. Moreover, the code expects perfect alignment between the two images, which is not likely to happen in a real world situation. I'm dealing with that myself all week with my current project where a robot is supposed to move an object and end up back in the same position every time. It's close, but not perfect. It's not aligned perfectly to the nearest pixel so there are "edge effects" that show up a little thin lines along the edges where they did not subtract perfectly and we need to handle those.

Answer 1

[Image Analyst]

You can use

hold on;
rectangle('Position', [x, y, width, height], 'EdgeColor', 'g');

to create green rectangles.

See this link for algorithms: http://www.visionbib.com/bibliography/contentschar.html#KK11219

For examples:

Bley, H.[Heinrich],

Segmentation and Preprocessing of Electrical Schematics Using Picture Graphs,

CVGIP(28), No. 3, December 1984, pp. 271-288.

Elsevier DOI BibRef8412

Fukada, Y.[Youji],

A Primary Algorithm for the Understanding of Logic Circuit Diagrams,

PR(17), No. 1, 1984, pp. 125-134.

Elsevier DOI BibRef8400

Thalmann, D.[Daniel], Demers, L.P.[Louis-Philippe], Magnenat-Thalmann, N.[Nadia],

Locating, Replacing, and Deleting Patterns in Graphics Editing of Line Drawings,

CVGIP(29), No. 1, January 1985, pp. 37-46.

Elsevier DOI BibRef8501

Okazaki, A., Kondo, T., Mori, K., Tsunekawa, S., and Kawamoto, E.,

An Automatic Circuit Diagram Reader with Loop-Structure-Based Symbol Recognition,

PAMI(10), No. 3, May 1988, pp. 331-341.

IEEE DOI Application, Circuit Diagrams. Recognize Symbols. BibRef8805

Fahn, C.S.[Chin-Shyurng], Wang, J.F.[Jhing-Fa], Lee, J.Y.[Jau-Yien],

A Topology-Based Component Extractor for Understanding Electronic Circuit Diagrams,

CVGIP(44), No. 2, November 1988, pp. 119-138.

Elsevier DOI Application, Circuit Diagrams. Automatic contour following based on the patterns that can appear. BibRef8811

[Lee Hao Nam]

alright, thank you for the references. Will try to get around it

Answer 2

[Giselle Roy]

hmm @Lee did you try using the CNN object detector, try classifying each defect based on it perimeter turn radius or if you have any other approach send it in, i may be able to solve it

Answer 3

[Abi]

Code please

[Image Analyst]

To learn how to write MATLAB code, invest 2 hours of your time here:

MATLAB Academy - Free 2 hour training

[DGM]

I don't know about you, but I learned as a child that saying "gimme" without actually saying what you want isn't really a successful strategy.

If you want to post an answer, post one. If you want to ask a question, ask one. This is neither. Decide what you want to do.