How to know an area of object using MATLAB

http://imgur.com/DBiot3W

I am currently stuck in this part. What I want to know is the area of an object that is highlighted in the YELLOW region above.

Please do assist me.

Answer 1

[Image Analyst]

The image segmentation algorithm that you used to get that binary image is poor. You should improve it. Only once you can get a good binary image can you call regionprops and get meaningful data out. You're not at that point yet. Go here for "car detection" algorithms: http://iris.usc.edu/Vision-Notes/bibliography/contents.html Sorry, we can't help with your algorithm selection or implementation/development.

[Gary]

Any suggestion or idea on how to further improve the segmentation ?

I am a beginner in MATLAB.

Do enlighten me on what i am suppose to do next.

[Image Analyst]

See my Image Segmentation Tutorial: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

[Image Analyst]

Did that help? I see you are still asking so I figure you must not have run the tutorial . If you just want to know the total number of white pixels in the binary image you can do

numWhitePixels = sum(binaryImage(:));

[Gary]

Forgive me a little bit.Even if i run your tutorial , i still have to understand the code.

Sorry.

[Image Analyst]

I thought the tutorial was well commented. Almost every line or pair of lines has a comment explaining it. Anyway, the tutorial is for if you want to know the area of each blob in your ROI. If you just want to know the number of all white pixels in the ROI, you can use the code I gave above:

numWhitePixels = sum(binaryImage(:));

binaryImage(:) converts the image into a column vector so you don' tneed to use sum(sum()). Or you can do

numWhitePixels = sum(sum(binaryImage));

Do you understand the code now?

[Gary]

Thanks a lot. Will update this thread if i still stuck somewhere.

[Image Analyst]

You're welcome. Can you mark the Answer as Accepted?

Answer 2

[Walter Roberson]

Once you have created a mask that corresponds to only the object, use regionprops() with the 'Area' option.

[Gary]

how do you create a mask that corresponds to the object only ?

I am beginner in this matter.

Do enlighten me.

[Walter Roberson]

Is that your original image, or do you have a color image that you are starting with? Will the car always be in the same location? Will it always be a car, or could it be an SUV or a truck?

[Gary]

I have a color image to begin with. The car will always be in the same location. It always will be a car.

I am testing it out on car first.

Any idea ?

[Image Analyst]

Take a picture without a car there then subtract them. That will tell you if a car is there or not. I can't imagine why you'd need the area, but if you did for some reason, threshold and call regionprops asking for the Area.

Please describe the use case. And I don't just mean "I want to know if a car is there" (that's NOT a use case) but something like "This is part of a system to find open parking spots at my university" or something like that.

[Gary]

this is a part of the system to detect the car randomly in front of me ( being pedestrian).

[Image Analyst]

You mean like a portable computer vision assistant for blind people who need to cross the street?

[Gary]

Yes and I will like to know what is the error in this part of the code

if true
  %

    [B,L] = bwboundaries(image_final,'noholes');
hold on
s = regionprops(image_final,'BoundingBox','FilledArea','FilledImage');
for k = 1:length(B)
boundary = B{k};
area = s(k).FilledArea;
area_string = sprintf('%2.0f',area);
text(boundary(1,2)-5,boundary(1,1)+5, area_string,'Color','black','FontSize',2,'FontWeight','bold');
end
end

Sometimes it will yield the value of 1 , 2 or 3 instead of the real value of the area.

Any idea on how to improve the code.

[Image Analyst]

You must have very tiny blobs in there. If it's only one pixel then when the image is shrunk to fit on the screen, it may have been subsampled away and you might not even see it. Use bwareaopen() to remove blobs smaller than a certain size that you specify.

[Gary]

This is the image that I tested the code on it but it somehow yield the value of 3 for the area.

Any opinion onto why the value is not correct ?

[Image Analyst]

There are blobs in there with areas 1, 2, or 3 as well as some very very large areas of thousands of pixels. I don't understand why having a blob with area 3 is unexpected for you.

[Gary]

Oh. So how am i able to get the entire area of the white colour pixel in the image above ?

I am still relatively new to MATLAB.