The Study Saline-alkali Soil Salinity Remote Sensing Inversion Model In The Arid Regions Based On BP Neural Network

Soil salinization is a major environmental issue in the world, and it is more serious in arid and semi-arid area. Acquiring accurate salinity information timely is important for monitoring and evaluating soil salinization. Traditionally, soil salinization monitoring selects fixed points to investigate in field, which wastes not only time but also manpower and can't show representative areas. It is impossible to realize large-area, real-time inspection. Remote sensing technique shows huge excellence in these aspects. Spectral technique is a new and effective approach in studying soil attributes. Spectral data in ground which is the base of band selection, validation and evaluation can build the links of ground, aviation and satellite remote sensing data.Taking Weigan—Kucha Oasis as the example area, this paper aims to explore saline-alkali soils salinity information remote sensing inversion model based on BP neural network in this semi-arid area. Firstly, the salinity of soil specimens is measured with soil solution method in laboratory and spectral data is gotten by CI700 in field. Analyzing the saline-alkali soils spectral data characteristic, this paper discusses the relations between saline-alkali soils spectral data and salinity, then selects the best band combination which can represent saline-alkali soils salinity spectral characteristic by means of multi-variables linear stepwise regression analysis and correlation analysis methods. That is spectral reflectance of 890nm, 800nm, 680nm, 590nm, 470nm, except remote sensing factors, the degree of groundwater mineralization, buried groundwater depth and the degree of topsoil mineralization are the three major factors which influence saline-alkali soils salinity, and these three factors are important input variables for this model.Being an important branch of intelligence computation, neural network model is a nonlinear mathematical model. This method may realize the mapping from eight dimensions variables to salinity information through training specimen data and adjusting the weights. Using neural network method to retrieve saline-alkali soils salinity is beneficial and can show the potentials of geography computation techniques in analyzing high quality data.This model contains two hidden layers, the first hidden layer including five nodes and the second hidden layer including three nodes. It has only one output layer, that is salinity information. The recycling model training makes this model approach to real mapping relation of datasets infinitely.This paper was divided into five chapters as follows: The first chapter expresses the main study background, goal, significance, then dissertates the evolvement of saline-alkali soils remote sensing inspection in the world, and some existing problems of the research on the saline-alkali soils using the 3S technology. The second chapter introduces the situations of the study area which cause soil salinization, including its nature environment, social economic situations and the status of soil salinization, The data source of the study area and the investigate in field.The third chapter dissertates the study method, content, technique route in this paper, and quantitively analyzes the relations among the collected spectral data in experiment, salinity and TM image data, and make use of multi-variables linear stepwise regression and correlation analysis method to select the best band combination which can show saline-alkali soils salinity characteristic.The forth chapter introduced the fundamental of NN model,the main factors to effecting saline-alkali soils salinity, then designs saline-alkali soils salinity information remote sensing inversion model based on back propagation neural network, and then this chapter gives the method of precision validation. In the last chapter, the results and the defects about this paper are discussed.