Remote Sensing Monitoring Models Of Soil Salinization In The Yellow River Delta Based On Featurespace

The Yellow River Delta is the typical of China's coastal saline soil area.Soil salinization has become an important factor in agricultural and economic development of the region.With its unique advantages,remote sensing technology has provided a simple and efficient way for soil salinization monitoring.Using remote sensing monitoring of soil salinization is mainly based on spectrum characteristics of salinized soil,but spectral characteristics of soil would be affected by atmosphere,soil moisture,vegetation and other factors,so we need further study to reduce their impact on the accuracy of extracting salinization information.Hyperspectral data opens up new avenues for quantitative relationship study on spectral characteristics of soil and salinity.This paper took Kenli County,Dongying city in Shandong province as the research object,obtained measured soil hyperspectral data,soil salt content data,cation-anion content of a variety of physical and chemical indexes of data and remote sensing data in the study area.To get the major factors affecting the salinization of soil,this paper dealt with measured physical and chemical properties data of saline soil and analysed correlation with the total salt content of each index.Analysed the measured soil hyperspectral data,studied characteristics of saline soil spectral and did correlation analyses of spectral reflectance and salinity data to guide the study on salinity-sensitive band of choice.After the preprocessing of ETM images of Landsat7 on August 16,2015,extracted the vegetation index,salinity index and wetness index.And the three indicators data are normalized.Built two dimensional feature spaces of vegetation index-salinity index,vegetation index-wetness index and salinity index-wetness index with the normalized data separately,and expressed the model with mathematical formulas.Used the measured salt content verification data to construct the model,then chose the best model.Apart from the model of the feature space,this paper built multivariate linear regression models based on a variety of spectra of soil salinity information for soil salinization inversion in the study area.Meanwhile,used the previous study on the BP neural network model to get inversion of salinization information in the same period,then did accuracy comparison in these three models to explore the feasibility and advantages of the three models.Through the comparative analysis of different models,revealed the feasibility of Landsat satellite multispectral data in soil salinization detection.Selected the optimal two-dimensional feature space for information extraction from ETM images from 2006 to 2015,and acquired change information of years of salinization in Yellow River Delta area.Analysed the salinization information over the years,and summarized the change information in study area.The theory of feature space has been verified in monitoring of soil salinization in arid areas,but in coastal saline area it has not yet been applied.This paper studied the salination in Yellow River Delta region with multiple two dimensional feature space models.The indexes used in models are simple and easy to extract.But compared with other remote sensing inversion models,the accuracy of the feature space model remains to be improved.Feature space model in non-arid areas still require further research.