Studies On Soil Iron Oxide Hyper-spectrum Characteristics And Estimating Model In Hilly Paddy Soil In Southern China

With the rapid development of remote sensing technology, especially the near-infrared spectroscopy and hyperspectral technology, In recent years, Application of hyperspectral remote sensing technology to quickly got near iron soil information has become an important part of current remote sensing of soil science. As a trace element necessary for crops, iron on the growth and yield of crops played an important role. Technical analysis using hyperspectral remote sensing of soil characteristics and content of iron oxide is an urgent demand estimate scientific management of farmland, but also a necessary requirement to develop precision agriculture using hyperspectral remote sensing technology.This paper, based on the high spectral measurements and chemical analysis from Xingguo County and Jiangxi Gaoan, Jiangxi Province two paddy soil samples. Made 350~2500nm spectral reflectance, spectral logarithmic curve, the first derivative curve and the curve to the envelope curve 14 totals. The study found that in the visible wavelength 450~520nm, iron oxide content in soil and soil reflectance negative correlation, In the band 1200~1800nm, the spectral reflectance differences increase, with a significant increase in iron oxide content increased with soil spectral reflectance characteristics. Variance of different iron oxide content of the spectral reflectance analysis showed, 1500~1800nm band could paddy soils in the study area as iron band spectral response.Visible to near-infrared hyperspectral technology and modeling method is an important direction of the current combination of soil research in the field of near-Earth sensor could be used to quickly access and agricultural crops, soil management, accurate information on aspects of the iron oxide. In this study, the iron oxide content of the soil by hand and 14 curves data correlation analysis, On the other hand dig for data, Looking characteristic parameters including soil index desertification, soil line parameters, soil iron index and gone to the envelope curve of depth, width, area, symmetry, angles, including, Iron oxide content and soil characteristic parameters correlation analysis, At the same time using a regression analysis, multiple regression analysis, partial least squares regression modeling, Iron oxide content of soil were constructed to estimate the model.The results showed that: This study validation of a regression model, multiple regression model and partial least squares regression model, contrast, chose the high precision three models: A regression model was the preferred one dollar envelope curve model 1632 nm band established: Y=57.4(λ1632)125.53; Stepwise regression model was preferable to the envelope by a high degree of symmetry and area spectral parameters established, Its multiple regression model was Y=70.164-69.256S-0.341 A, Based on full-band modeling and validation samples ratio of 3, Partial least squares fit of the regression estimation highest model B. Among them, Partial least squares regression to estimate the highest accuracy of model B: Partial least squares regression to estimate the highest accuracy of model B: RC2 highest, was 0.92, RMSEC minimum of 4mg/kg, RMSEP minimum of 5.03mg/kg, RPD maximum in three models, was 2.92. Therefore, this article select the optimal model for estimating soil PLSR model high iron oxide content of the spectrum.