MODIS(Moderate-Resolution Imaging Spectroradiometer) remote sensing data have higher radiometric sensitivity,but its lower spatial resolution always causes the pixel’s impurity,normal classification methods of land cover can not solve this problem.In order to achieve the classification of land cover a line spectral mixture model was used in MODIS data pixel unmixing of Zhengzhou area.The preprocessing of MODIS data,line spectral mixture model(LSMM),and methods for solving image end-member’s reflectance were also discussed.The classification result was compared with the class map derived from Landsat ETM+ data.The RMS shows that the pixel unmixing method has good results and the MODIS data can be effectively applied to remote sensing dynamic monitoring,and land cover classification.