Abstract: 【Objective】 To select the object-oriented supervised classification method for remote sensing classification of spruce forest in Tianshan Mountain, and to select a method with the best classification effect in the hope of providing the basis for the investigation of forest land resources and dynamic monitoring and evaluation in this area. 【Methods】 Based on GF-2 remote sensing image data, the optimal segmentation scale of objects in the study area was screened by the combination of ESP scale evaluation tool and visual interpretation. On the basis of three different classification methods, remote sensing classification was conducted. 【Results】 The optimal segmentation scales of water body, road, other land, forest land and grassland in the study area were 390, 372, 316, 296 and 246, respectively. Secondly, under the optimal segmentation scale of local objects, compared with the nearest neighbor classification, combined with vector data classification and threshold classification, it was found that the Kappa coefficient and overall precision of the three classification methods were 0.760,7, 0.782,0, 0.840,6 and 0.814,8, 0.830,5 and 0.876,5, respectively. 【Conclusion】 The threshold classification method is superior to the other two classification methods, which indicates that it is feasible to select a better threshold classification method to solve the problem of forest land resources investigation in this area.