Abstract: Signature selection is very important in the target recognition, especially for the wavelet moment invariants which have three parameters (m,n,q). The absolute resovable signature selection algorithm under the multi-class mode for target recognition is put forward and the proper range for m,n,q values during the image digitalization processing is given, based on the signature selection in two class mode. The experiments show that this method can be successfully used in the recognition of multi-class targets, whether the multi-class targets are different, similar or mixed types.