Remote Sensing Inversion Of Nitrogen Nutrition Status In Apple Canopy Of New Shoots Stage

Nitrogen is the essential nutrient for the growth and development of apple tree,the rapid and convenient diagnosis for crop nitrogen content is of great significance to crop precision management and high quality and high efficiency production.The remote sensing technology is based on its spectral reflectance characteristics,wide range of observations,quick aging,abundant information and other advantages enter the field of crop nutrition diagnosis.The increasing development of remote sensing technology makes the large area of nutritional status inversion possible.This study took the Shandong province Yantai Qixia as the research area,with of apple canopy which new shoots growing flourishing as the research object,the SPOT-5 satellite remote sensing image,the ground measured hyperspectral data,DEM elevation model data and so on were adopted and the software such as ERDAS,ENVI5.1,ArcGIS,DPS,LIBSVM were applied to the inversion of the apple canopy reflectivity,building up the apple canopy N-Content sensitivity index,setting up and screening apple canopy N-Content inversion model.The apple orchard extracted from the study area was carried out according to the inverse reflectance of apple canopy,and the N-element content of apple orchard in the research area was retrieved by the spatial inversion model of N content which has built up.The main findings are as follows:(1)Canopy reflectance inversion of apple tree which new shoots growing flourishingThe SPOT-5 remote sensing images of apple tree which new shoots growing flourishing were pretreated by atmospheric correction,correction of positive injection and geometric rectification,and the terrain correction of the image was reduced by SCS+C model,which obviously reduces the influence of atmosphere and topography.The inverse reflectance of the apple canopy was obtained by using the linear model to decomposed the mixed image element.Comparing the inverse apparent reflectivity with the surface reflectance,the results showed that the image was similar in the area of relatively flat terrain,and the obvious difference can be seen in the areas with large topography,The shadow image after the terrain radiation correction was compensated by the reflectivity of green,red and near infrared bands and vegetation indices.Comparing the inversion reflectivity of the apple canopy in the experimental sample area with the apple apparent reflectivity,the inverse reflectance of apple surface and the inversion reflectivity of apple canopy,the inversion reflectivity of apple canopy is closer to the measured reflectivity.The error of the inversion reflectivity and the measured reflectivity decreases progressively after the correction,compared with each band error,the red Band error decreases from 137.222% to 0.475%.(2)The N sensitive index was constructed and screened185 spectral indices were constructed based on the inversion reflectivity of apple canopy,and the correlation analysis was carried out with the status of N-content in apple canopy.(R3+R4)*(R4-R3),(R3/R4)/(R3-R4),lnR4/(R4-R3),(R4/R3)*(R3+R4),R4-R3,(R4-R3)/eR4,R2-R3 and R4/R2 altogether 8 indices whose correlations to a significant level were Screened out as sensitive indices,The sensitivity indices are mainly composed of red-band(0.61~0.68um),near-infrared bands(0.78~0.89 UM)and Shortwave infrared bands(1.58~1.78 um).(3)The construction and test of N-content inversion model for apple canopyThe 8 sensitive indexes were selected as the independent variables,the N content of apple canopy was measured as the dependent variable.and the stepwise regression model is established.finally the X1(i.e.(R3/R4)/(R3-R4)),X2(i.e.lnR4/(R4-R3)),X3(i.e.(R4/R3)/(R4+R3)),X4(i.e.(R4-R3)(/eR4)were screened out as the four independent variables,the model as follows: y=-170.481+13.807* x1-21.448*x2+203.379* x3-235.571* x4;3 highly significant levels of sensitivity indices X1(i.e.(R3+R4)*(R4-R3))X2(i.e.(R3/R4)/(R3-R4)),X3(i.e.lnR4/(R4-R3))would be applied To build SVM model as the independent variables,with multiple training to determine the type of SVM 4(4 indicates V-SVR),The kernel function type is 2(2 = RBF function).The decision coefficient of stepwise regression model is 0.361,the decision coefficient of support vector machine model reaches 0.785 which is much higher than the stepwise regression model;the RPD of support vector machine model is 2.145,which is higher than the stepwise regression model,the result shows that the support vector machine regression model has a good estimation ability for apple canopy nitrogen inversion.(4)Extracting apple orchard information and the inversion of nitrogen nutrition statusAccording to the apple canopy inversion reflectivity,measured canopy inversion reflectance,the DEM data and the NDVI data in the research area,the information of apple orchard in the research area were extracted and the accuracy evaluation of classification is up to 90.03%.Using support vector machine model to study the spatial inversion of n-nutrient status in apple orchard and obtain the spatial distribution of n nutrition status in the research area,that can provide macroscopic basic data for the efficient management of apple orchard.In the spatial layout,the N-element content of orchard in the research area shows the rank of 1th(2.6-2.9%),2th(2.9-3.2%),mainly in the southern Shewopo town and the northwest Sujiazhuang town;the northeast is dominated by the 1th rank.The southwest is dominated by 2th rank;the 4th(3.5-3.8%)rank are scattered.Comparing the inversion rank with the measured value of the sample area,the measured canopy Nitrogen level was the same as that of N content level,indicating that the spatial inversion distribution map is instructive to the macroscopic management of the apple orchard in the research area.