How can the two stages of feature extraction and classification be linked?

i want to make classification [using Multylayer perceptron] to the galaxies images to classify 3 type of galaxies(elliptical, spial, irregular).

I made a feature extraction using the PCA for about 10 images for each of 3 types of images and put them in a matrix .. The first 40 columns in the matrix express the properties of the elliptical galaxies and the second 40 columns express the properties of the spiral galaxies and the third 40 columns express the properties of the irregular galaxies. and named this matrix (P)which is input . And then i write the classification code with ANN which i found in matlab is:-

load('P.mat')

load('T.mat')

net=feedforwardnet([10 12]);

net=train(net,P,T);

view(net)

_____________________

i want to know what is T(target) ? what can i put in T ?

i want the system make classification.

can any one explain to me what i can do plz?

i want to make train and test the images >> what i can do?