how to extract features from a satellite image will all feature matrices having same size?

i want to classify a satellite image based on its NDVI values. i have to classify the image in three classes: wheat, fodder and others. when i extract wheat (ndvi>0.5), fodder(ndvi<=0.5 and >0.3) and others(ndvi<=0.3) pixels from image, the three column matrices have different sizes due to which i cant apply patternnet to it. plz help me by telling some efficient way of feature extraction. plz also tell me if it is possible to train the neural network first for a sample of the image and then use it to classify the whole image? i would really b obliged if any one answer. badly need a quick response

Answer 1

[Greg Heath]

Since each pixel has a different classification, the rxc image is columninzed and transposed to an Nx1 row.

 input = image(:)';

The corresponding Nx3 target matrix contains 3-dimensional columns fro the 3-dimensional unit matrix eye(3).

Hope this helps.

Thank you for formally accepting my answer

Greg

[saima zia]

thanks a lot sir for clearing one of my confusions.

i have a 50 by 50 matrix. so according to ur suggestion i columnized it and i got 2500x1 row by:

input=image(:);

now i want to ask that should i arrange the pixels in input like first all the wheat pixels then fodder pixels and then others?

Also according to my input image, what will be the size of my target matrix? and how to create it?

how to assign each pixel its class?

plz answer so that i can carry out my research work.

[Greg Heath]

sizes already given

 help ind2vec

 help vec2ind

For examples search using

 greg ind2vec vec2ind

[saima zia]

Thank u sooooooooo much respected sir, i have done the classification.

Really obliged and thankful.

Answer 2

[ashish bha]

respected sir , i am working on satellite image classification using BBO with CS algorithm with Urban , vegetation, barren , water , rocky , can you pl help me image classification of its satellite image if you can pl : [e-mail] Thanking you.