Estimating Soil Salinization Based On Hyperspectral Technology In The South Of Xinjiang

Hyperspectral remote sensing is a new type of method for monitoring soil salinization, and carries on the corresponding ground remote sensing test is the important basis for guaranteeing its monitoring accuracy. Near-ground spectral measurement data for the interpretation of remote sensing images is very important and valuable reference data, also is the important foundation of remote sensing data interpretation, can help researchers to understand various spectral characteristics. Therefore, the research object of the project is the southern of Xinjiang, which salinization area is wide and salinization degree is heavy, This study of soil in southern xinjiang, xinjiang area as the research object, the salinization of soil on the basis of hyperspectral data. This paper studies the indoor measurement spectrum data processing and analys is, and soil spectral reflectance under different salinity gradient change, the relative difference of reflectance and reflectivity variation were analyzed and the correlation between soil salt, soil spectral reflectance and the relationship between soil salt, preliminar ily determine the characteristic of the salt band in 400 ~ 800 nm band range. The study found that salt can reduce soil reflectance, in 400 ~ 2 400 nm band range, reflectivity is not arranged by the discretion of the salt content present increasing or decreasing. The k-means clustering method are fuzzy soil spectral reflectance characteristics can be divided into four types of spectrum, combined with the PLSR method to classify modeling, can significantly improve the prediction accuracy and stability of the model, relying on human judgment or improved based on traditional soil classification insufficient, but also for spectral modeling and prediction of soil nutrients and other physical and chemical properties provide new ideas and methods. Through a variety of spectral preprocessing methods, the establishment of the inversion model of different water-soluble base ions, the best inversion model to determine various ions, and verify. Results show that different ions in the best inversion model used by different pretreatment method, choose the best method for spectral data preprocessing are beneficial to the model and the inversion accuracy improved. In addition to K+, Na+ and the poor correlation between the spectral reflectance and hardly inversion, the other five ions(HCO3-, Cl-, SO 42-, Ca2+, Mg2+) have good correlation, you can get a better inversion accuracy. The software will use ENVI5.1 hyperspectral data into multi-spectral data, by simulating Landsat 8 OLI of seven shortwave bands, build a variety of surface features typical spectral index, which were simulating Landsat 8 OLI seven bands together as input, respectively partial least square regression(PLSR) and support vector machine regression(SVR) to build soil salinity inversion model, the results of model inversion method to establish the accuracy of SVR is better than PLSR methods to achieve using hyperspectral data to simulate multi-spectral Landsat 8 data inversion salt.