Feature extraction for classification
1 visualización (últimos 30 días)
Mostrar comentarios más antiguos
Ushna Asif
el 20 de Sept. de 2017
Comentada: Ushna Asif
el 22 de Sept. de 2017
Hi there, I attached the images that I want to extract features for classification. these are zoomed images of different fruits. kindly help me how to find features on these images. I tried using extractHOGfeatures function but its not giving me correct classification.
2 comentarios
Image Analyst
el 20 de Sept. de 2017
What are the classes? How many classes are in each photo? One? If so, then one feature may simply be the color. Why do you think HOG should be what you want? Can you attach what you think the classification should look like?
Respuesta aceptada
Image Analyst
el 21 de Sept. de 2017
I would think that just color (Hue and Saturation) and texture might be enough. Of course if two images look the same but are different fruits, there is no telling, like you have an image of orange grainy things and one is an apricot and one is a nectarine.
3 comentarios
Image Analyst
el 21 de Sept. de 2017
I have examples in my File Exchange: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862&sort=downloads_desc
For example, convert to HSV color space and get the mean hue, saturation, and value.
hsvImage = rgb2hsv(rgbImage);
meanHue = mean2(hsvImage(:,:,1))
meanSat = mean2(hsvImage(:,:,2))
meanValue = mean2(hsvImage(:,:,3))
sdImage = stdfilt(hsvImage(:,:,3)); % Std Deviation of Value channel.
meanStdDev = mean2(sdImage);
You can make a feature vector for this one image like
featureVector = [meanHue, meanSat, meanValue, meanStdDev];
Más respuestas (1)
Akira Agata
el 21 de Sept. de 2017
I believe the following example will be your help. Using this method, you can extract 4096-dimensional feature vector for each image.
1 comentario
Ver también
Categorías
Más información sobre Image Data Workflows en Help Center y File Exchange.
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!