How can I improve my image processing code?

Dear Friends

Here are my codes to train a machine to some pictures which are located here titled: output_a______mat, then I want to match a new_feature_vector with the feature.mat

but the outbut has only 10% accurcy.

can you advise?

note:

1- run the pngoutput.m

2-run the imProc.m

3-run the grayEdge.m

4-run the colorProc.m

5-run the Datapreparation.m

6-run the labels.m

7-run the classification.m