how can I calculate hsv value for the object of interest?

Sir, I am doing fruit classification system.I finished preprocessing and segmentation using k-mean and thresholding. I think to use HSV color space for color feature. I want to know how can I calculate hsv value for the object of interest (fruit image). I find hsv value but I got the value of the whole image including background. Thank you so much,sir.

Answer 1

[Image Analyst]

I do that in my HSV color segmentation demo in my file Exchange: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

Also search the forum for tag:fruit since fruit analysis seems to be one of the most popular subjects for student projects.

[Yuki Zue]

yes sir.I used k-means clustering for color segmentation.Now I want to try to create a single segmented image with the two clusters. Can I use this code?I think this code will be for three clusters.Please,give me guideline,sir.Thank you so much for your kindness. Code is:

colmat = [1 0 0; 1 1 0; 0 0 1];

segmented_image = reshape( colmat(Labels(:),:),Row,Col, 3);

[Image Analyst]

I don't think kmeans clustering would be a good idea for color segmentation for defects in fruit. For one thing, you'd want something that worked for any amount of area fraction defect from 0% to 100%, and kmeans will have trouble at the low and high end of that. Secondly the fruit will vary in color, and thus so will your cluster locations. This is not as good as having fixed ranges for acceptable color. If your Valencia orange fruit is brown but uniform, kmeans would say that it's a good fruit despite the bad color. And your code won't work. There is no way you'll have a defect detection algorithm with two lines of code. The code is not segmenting the image despite what you called the variable.

I attached a kmeans color segmentation demo, for what it's worth, in some of my other posts. It's not yet on this (my work) computer so I can't attach it.

[Yuki Zue]

I am done segmentation using k-means and thresholding methods,sir.I did two cluster in k-means.I mean that I want to find HSV values.So,I decided to do color feature extraction after k-means segmentation.To find HSV values, I want to create one segmented image by adding above two clusters.Because of doing this,I think I can get accurate hsv value. First,my HSV code: hsvImage=rgb2hsv(segmented_image); hImage=hsvImage(:,:,1); sImage=hsvImage(:,:,2); vImage=hsvImage(:,:,3); hvalue=mean(hImage); svalue=mean(sImage); vvalue=mean(vImage); But this can't give accurate hsv values.Then I added this code: colmat = [1 0 0; 1 1 0; 0 0 1];

segmented_image = reshape( colmat(Labels(:),:),Row,Col, 3); I think this is for three clusters.For two clusters where should I change,sir. I think I got hvalue as good result.svalue and vvalue got 1.0.is it ok,sir?I am just beginner.But I love MATLAB Language.Thankyou so much for your patience and guidelines.

[Image Analyst]

Well, like I said I don't think it's a robust method. I'm attaching my kmeans color segmentation demo and you can see how well it works on your images. You can also attach an image if you want so we can run your code.

To format your code, see this tutorial: http://www.mathworks.com/matlabcentral/answers/13205#answer_18099

[Tasneem Al-Tmimi]

I have check the SimpleColorDetectionByHue()File, It helps actually, but am confused how to detect the HSV values of only a segmented part not the whole image can you just explain this part please or just tell me where to find in the code since am lost here :D

[Image Analyst]

Take your segmented image - the binary image mask - and use that to extract the HSV values

huePixelValues = hImage(mask);
satPixelValues = sImage(mask);
valuePixelValues = vImage(mask);

[Tasneem Al-Tmimi]

Thank you that works with me :)

One last question, what is the accurate way to set my HUE, SAT, and VALUE intervals for different colors?

As an example, I did this for HUE , it works but not always

GREEN_RANGE = [54,216]/360;
BLUE_RANGE = [200,245]/360;
RED_RANGE = [288,360]/360;
YELLOW_RANGE = [36,50.4]/360;
BLACK_RANGE = [0, 0]/360;

Is there a reference I can refer to that has the different color shades and swatches range values ?

[Image Analyst]

Try the Color thresholder app on the Apps tab of the Tool Ribbon.

[Tasneem Al-Tmimi]

Actually I can't understand how to use this App :/

can you tell me the steps looks like what ??

[Image Analyst]

Click the "Load image" button and select an option, such as load your image from a file on disk. Then select the color space you want to threshold in. Then, after you got it like you want it, click Export and send the code to a script.

[Tasneem Al-Tmimi]

I did that, and I got like max and min value for each channel, the question here can I apply the same values for different pictures that have items with the same colors to detect ??

Thanks in advance

[Image Analyst]

You must have forgotten to click the export button. That will create a function that you can call on other images.

[Tasneem Al-Tmimi]

Thank you very much, this was such helpful

If I have one question about one of your exchanged file where shall I ask you ??

Because I am considered on that :)

[Image Analyst]

You can ask here in Answers. I check here a lot whereas I hardly ever go back to my File Exchange to look for comments.

[Tasneem Al-Tmimi]

I have a picture that thresholding it always gave me wrong result can you check it for me ??