So I have this image:
There are 36 objects in total and roughly 10 different ones.
Is there a method to calculate how many different objects there are?

 Akzeptierte Antwort

Akira Agata
Akira Agata am 25 Jul. 2018
Bearbeitet: Akira Agata am 25 Jul. 2018

2 Stimmen

How about the following way?
% Load the image
BW = imread('bARgija.png');
% Remove small area (< 10 pixel)
BW2 = bwareaopen(BW,10);
% Obtain area and bounding box for each area
s = regionprops('table',BW2,{'Area','BoundingBox'});
% Set allowable area torelance (e.g 5 pixel)
tolPixel = 5;
% Assume same object has same area (with 5 pixel torelance)
r = tolPixel/max(s.Area);
[~,~,group] = uniquetol(s.Area,r);
% Show the result
color = jet(max(group));
figure
imshow(BW)
hold on
for kk = 1:height(s)
rectangle(...
'Position', s.BoundingBox(kk,:),...
'EdgeColor', color(group(kk),:),...
'LineWidth', 2);
end
The number of different objects can be obtained by max(group).
>> max(group)
ans =
12

4 Kommentare

Image Analyst
Image Analyst am 25 Jul. 2018
Bearbeitet: Image Analyst am 25 Jul. 2018
Very nice - clever. I was going to use some kind of machine learning method to figure out how many area clusters there are. I didn't know about uniquetol(). It looks like a nice function - perfect for using here.
I also learned about height() and width() from your code. However, I'm disappointed that the Mathworks chose to create a built-in functions with those names, instead of tableheight() and tablewidth() because I've often used variables with those names (height and width). I guess I'll have to change now.
João  Oliveira
João Oliveira am 27 Jul. 2018
Thank you for the answer. Really apreciated.
Paolo Piazza
Paolo Piazza am 24 Mär. 2020
This is really interesting, I'd have another question: is it possible to calculate how simillar/dissimilar the objects are in an image? for instance, if all objects were saws (or more less have the same shape).
Image Analyst
Image Analyst am 24 Mär. 2020
Bearbeitet: Image Analyst am 24 Mär. 2020
What I'd probably do is to align the objects, like with imregister() or imrotate() or something, then compute the Dice Sorensen Coefficient. It's been talked about here before so look it up. Also run my attached demo on it.

Melden Sie sich an, um zu kommentieren.

Weitere Antworten (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by