In computational proteomics, inferring the peptide sequence from its tandem mass spectrum is an important issue. Several algorithms have been proposed to solve this problem. However, few algorithms make good use of the intensity information of the ions. In this paper, a novel scoring strategy is proposed based on $k$NN technique for identifying peptide by use of tandem mass spectra. First the intensity feature vector is defined to represent the total intensity distribution of ions with different types. Then a hyper surface with a novel distance is constructed. A dataset of intensity feature vectors is established by use of the identified spectrum and all the vectors are mapping to the points on the hyper surface. Finally, a scoring strategy based on $k$NN technique in the hypersurface spaceis proposed for re-evaluating the peptide identification results. Experimental results demonstrate that the proposed method improves the accuracy of peptide identification algorithms.