[Objective] Vegetation coverage in Bayanbulak grassland of Xinjiang Autonomous Region was graded and the regional vegetation coverage was determined in order to provide theoretical reference for the prevention pasture degradation. [Methods] With the help of American SOC_710 portable visible/the near infrared spectrum imaging spectrometer, the normalized difference vegetation index(SOC_NDVI) was adopted to establish the MODIS spectral estimation model between NDVI(MODIS_NDVI) and actual vegetation coverage(VC) using linear regression analysis method. The grading and determining of regional vegetation coverage were completed on the inversion coverage result of the model. [Results] The prediction accuracy of MODIS spectral estimation model was high with a standard error of 11.55%, the overall accuracy reached 88.92%. The 5 vegetation coverage levels had extreme high coverage, high, moderate, low and extreme low coverage, and these coverage proportion were: 18.87%, 25.61%, 31.28%, 13.86% and 10.38%. [Conclusion] The vegetation coverage on the whole study area was high in the north and west, and low in the east. The research area was covered mainly by middle and high coverage grades; proportions of other coverage grades as low, extreme low and extreme high, were small.