The present paper selected the spectral reflectivity of saline soil and vegetation of Weigan-Kuqa River Delta Oasis in the northern margin of the Tarim Basin in Xinjiang as objects, and used various spectral transforms to process the data with continum removed methods, derivate spectra, reciprocal, first order differential and root mean square etc, then analyzed the spectrum features and decided the most sensitive band ranges most relevant to salinization, and used field hyperspectral vegetation index, soil salinity index and measured synthetical spectral index to respectively establish hyperspectral quantitative models which could evaluate the soil salinization degrees. By comparing various spectral transformations of hyperspectral data the result showed that the first derivative of measured soil and vegetation hyperspectral were most sensitive to soil salinization degrees. The hyperspectral quantitative model based on measured synthetical spectral index could monitor soil salinization accurately and was better than the models simply based on vegetation index or soil salinity index. The research provided some scientific basis with soil salinization detection.