determine pixel length and width of the body

hello master,

I have an image that has been processed using canny edge and want to count the number of pixels in each area, what command should i use?

Answer 1

[Image Analyst]

You do not have many closed areas. It looks like, for some reason, you did edge detection on a cow. I think you should probably use SegNet instead since Canny will not give you a closed contour around the outermost perimeter of the cow.

[ilham trishardanto]

can i use contour tracing to make closed area sir?

[Image Analyst]

Probably not using fcontour() for this image. My answer is still to figure out how to use Segnet (deep learning).

Why do you need to outline cows anyway?

If you want you can use drawfreehand() to manually trace the cow. See attached demos.

[ilham trishardanto]

I need an outline to get information in the form of the number of pixels that will be used to calculate the weight of the cow

[Image Analyst]

I suggest you try to find out how the object selection tool in Photoshop works. I drew a box around the cow and this is what it found. The black and white dashes outline the cow.

I think it does a good enough job to make a weight estimation. But I don't know the details of how it works.

[ilham trishardanto]

thanks sir, but i need information in the form of the number of pixels on the cow's body, the length of the cow's body will be obtained and the width of the cow's body in the form of pixels which will then be converted to cm and entered into the formula. and I do it in a large number of images, so it's not possible to do it manually with photoshop ^_^

[ilham trishardanto]

maybe use bwtraceboundary sir? because after get the closed area i can determine the pixel length and width

[Image Analyst]

You can try. If the cow is a different color than the background, you can use the Color Thresholder app on the Apps tab of the tool ribbon.

[ilham trishardanto]

can you help me to use region of interest to get body cow area, sir?

[Image Analyst]

Yes. Once you have the area of interest, the area of the cow from a side view, you'll have to calibrate it to real world units like centimeters. So you'll need to have an object of known size, at the same distance as the cow, in the field of view, like a meter stick or piece of photocopy paper or something like that. Then you measure that and compute the calibration scale factor in cm/pixel. See attached demo.

[ilham trishardanto]

thank you sir

Answer 2

[yanqi liu]

yes，sir，may be use bwlabel and regionprops to get the area information

may be upload your origin image，then we can do some analysis

[ilham trishardanto]

this is a original image sir

[yanqi liu]

clc; clear all; close all;
img = imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/845780/image.png');
J = rgb2hsv(img);
v = mat2gray(J(:,:,3));
bw = imbinarize(v,'adaptive','ForegroundPolarity','dark','Sensitivity',0.55);
bw = bwareafilt(bw,1);
bw2 = imopen(bw, strel('disk', 7));
bw2 = bwareaopen(bw2, 1e3);
bw2 = imclose(bw2, strel('disk', 11));
bw2 = imfill(bw2, 'holes');
stats = regionprops(bw2);
rect = stats(1).BoundingBox;
be = imdilate(bwperim(bw2), strel('disk', 7));
[r,g,b] = imsplit(img);
r(be)=255;g(be)=0;b(be)=0;
img2 = cat(3,r,g,b);
figure;
subplot(2, 2, 1); imshow(img);
subplot(2, 2, 2); imshow(v, []);
subplot(2, 2, 3); imshow(bw, []);
subplot(2, 2, 4); imshow(img2, []);
title(sprintf('长度为%d', rect(3)));

[Image Analyst]

Not bad, but we know that when he comes back he'll say "yeah but it doesn't work for this cow" and show us one completely different, like a black cow, or a cow with brown and white patterns on it.

[yanqi liu]

yes，sir，this method can not use in common，it use some stat segment method，so it is just a discuss demo

if meet other image，may be use DeepLearning model，such as segnet、unet to make some auto segment，but it aslo can not use to every other image

[Image Analyst]

Correct, just to build on that (for other people - I'm sure you know) with an example: for example if you trained only with side views of cows then came along with a front or rear view, then segnet might not find the accurate shape and of course the area wouldn't be right and thus the volume of the cow wouldn't be right.