在MATLAB中从多边形图像中检测正确数量的CORNER坐标

Question

I have a number of polygon images like a hexagon, a pentagon, any quadrilateral etc.. i need to generalize the detection technique to detect the RIGHT number of Corner coordinates.. no extra coordinates should be generated.

for eg:- the code should detect only 4 for a quadrilateral, 3 for triangle, 5 for pentagon and so on..

I used HARRIS corner detection to detect right corners by specifying the number of corners value but i cant use the same code for an image with different number of edges.

Reason for using the same code is i am trying to bulk process image -> Detect corners and print them... i cant change the code for each image.

Sample Images:-

Octagon:

Pentagon:

Answer 1

There is a function called corner that works very well given the right input parameters.

For instance setting an appropriate QualityLevel give accurate results:

clear
clc

A = imread('Octagon.jpg');
A_gray = rgb2gray(A);

figure;
Ca = corner(A_gray,'QualityLevel',.2)

The coordinates ar stored in Ca as a N x 2 matrix. Here N = 8.

imshow(A)

hold on

scatter(Ca(:,1),Ca(:,2),80,'filled','k')
hold off

B = imread('Pentagon.jpg');
B_gray = rgb2gray(B);

figure;
Cb = corner(B_gray,'QualityLevel',.2)

imshow(B)

hold on

scatter(Cb(:,1),Cb(:,2),80,'filled','k')
hold off

Outputs:

and

Yay!