Abstract: Because of the high mountains, steep slope and fragmentized fields, the spectrum characteristic of target on the ground in Guizhou province is complicated and is heavily disturbed by the phenomenon of "same spectrum different target" and "same target different spectrum". It is found that is difficult to reach the ideal result only by the traditional classifying method in the study on the application of CBERS data in the forest resources investigation of Guizhou province. Using the method of non-linear band ratio in-the study effectively reduces the disturbance. Analyzing through grey value, spectrum sectional curve, etc, has generally strengthened the contrast ratio of the image, and the differences between the various land-types,thus improved the classifying precision.