Each substance has its own unique spectral information, so it is possible to identify substances based on spectral information. An identification method of spectral information divergence pigment based on statistical manifold is studied. The Riemann metric on the statistical manifold is used as a new metric in the information divergence. The spectral matching for four commonly used mineral pigments is carried out withutilization of the proposed method and traditional method, and the matching results are compared. The experimental results show that the method of spectral information divergence based on statistical manifold can solve the problem of geometric measurement of probability variation, and the accuracy of spectral information matching recognition is improved obviously.