Multidimensional classification feature data set based on multi-temporal remote sensing image can be fully mining the information to improve the land cover survey classification precision.This article using multidimensional classification feature data set based on multi-temporal Landsat8 remote sensing image, through the different classification methods, extract the experimental area 11 kinds of land cover information and precision analysis in the Zhenlai County located in the one of the world's three largest saline-alkali soil. The results show that: 1) the SVM has good effect on saline-alkali soil special ecological environment′s land cover information extraction, overall accuracy87.77%, Kappa coefficient 0.8649; especial has good effect on saline-alkali soil , Production accuracy 98.34%, 2) Different classification methods′ accuracy from high to low is Support Vector Machine、Maximum Likelihood、Neural Net、Minimum Distance、Neural Net.3) the most common use of land in the study area include dry land、paddy field、saline-alkali soil, the west to dry land as the main、the saline alkali land, alkali foam and dry land alternative distribute in the central region , the east is given priority to with paddy field.