Abstract: Soil salinization is a process of global land degradation, which hazards the environment. It is caused by inefficient irrigation and the excessive use of water. It reduces the productivity of land. Xinjiang is the typical area of arid and semi-arid region. The monitoring of soil salinization timely and effectively is not only beneficial to the production of agriculture, but also in favor of sustainable development of agricultural land. The purpose of this study is how to improve the classification accuracy. In this paper, the author uses Spectral Angle Mapper method extracting the information of soil salinaztion. In order to improve the accuracy of classification, the author determines the appropriate soil and vegetation spectral library. The appropriate soil and vegetation spectral library decide the accuracy. The use of field measurements of soil spectral reflectance is used to study the soil spectral characteristics. It is combined with hyperspectral data of the Chinese environmental and disaster monitoring and forecasting of small satellites and classified soil salinization based on hyperspectral image. In the first part of this paper, according to the definition of degree of soil salinaztion, four classed of soil are classified, namely non-salined soil, slight-salined soil, moderate-salined soil and heavy-salined soil. The data from four classes of soil can be converted to fourteen transforms of soil spectral reflectance. There are fifteen transforms of soil spectral reflectance. They are the original and the converted fourteen transforms of soil spectral reflectance. According to the findings of the correlation analysis of fifteen transforms of soil spectral reflectance with soil salt content, regression analysis are done. The equations are chosen to estimate soil salt content. And root mean square error (RMSE) is employed to verify the accuracy of the equations. The best equations of estimating soil salt content are decided. In the second part of this paper, the author decides the soil spectral library according to the result of spectral characteristics. The vegetation spectral library is based on field measurements and actual survey. In the final part of this paper, the author uses the SAM method to classify the hyperspectral image based on the soil spectral library and vegetation spectral library. Such a classification is proved good, laying the foundation for the region's hyperspectral applications, giving warning to regional farmers for appropriate farming methods, provided the data for the region's sustainable development. This article identifies the soil and vegetation spectral library of the study area. This helps to further study on spectral characteristics in this region. Environmental and disaster monitoring and forecasting of small satellite is designed and developed by our country. China centre for resources satellite data and application provided for our researchers free. HSI hyperspectral data is also China's the only hyperspectral remote sensing images. This study attempts to contribute to our development of hyperspectral remote sensing images. To speed up the satellite for the applications of environmental monitoring. It is better for the country's environmental monitoring services. The development of hyperspectral remote sensing images will bring a new opportunity and challenge to remote sensing technology. Our researchers should strive for the development of hyperspectral remote sensing technology. The development of hyperspectral remote sensing will promise a better future.

Key words: hyperspectral data, soil salinization, spectral characteristics, Weigan-Kuqa Oasis, Xinjiang