The effect on feature performance caused by measuring scale has never been taken into consideration in the present methods of fully polarimetric high range resolution profile(HRRP) feature extraction based  on H/A/α decomposition. Therefore, the dynamic mutual information is introduced to design the evaluation criteria of measuring scale. Meanwhile, three different feature extraction methods are proposed based on mean scale, indefinite scale and pyramid scale respectively. The validity analysis of the feature subsets is performed based on the measured fully polarimetric HRRP data from two kinds of planes. The separability and stability of the subsets are validated by the recognition of multi-class plane.