Research On The Problem Of Rape Nitrogen Optimum Measuring Position Based On Spectral And Imaging Data

Chlorophyll as the main material of plants’photosynthesis is an important indicator to plants’growth and nutritional status.It is of great significance that rapid and non-destructive testing for the chlorophyll content and optimal leaf position for chlorophyll diagnosis on the detection of plants growing.However,the traditional lab chemical measurements are time-consuming and destructive,which cannot meet the needs of the rapid non-destructive testing.Hyperspectral imaging technology setting spectrum technology and imaging technology in one,is an effective method in rapid and non-destructive measurement of plant chlorophyll,and has been more widely used.In this study,rape leaves as materials;hyperspectral imaging technology was used to establish a model to rapidly detect SPAD values in eight leaves of rapeseed leaves and the visualization of SPAD values distribution.The optimum measuring position leaves of SPAD meter was determined based on SPAD map.The main research contents and achievements are shown as follows:（1）Based on hyperspectral imaging technology,rapid non-destructive detecting SPAD values in eight leaves of rapeseed leaves was completed.Hyperspectral images of leaf samples were acquired in the Vis/NIR region（400nm-900 nm）and their spectral data were extracted from different position.Partial least squares（PLS）was applied to select the optimal spectral preprocessing method from first derivation(1^stDer)、second derivation(2^stDer)、SG-smoothing、standard normal variate transformation（SNV）、Detrending and multiplicative scatter correction（MSC）.Eight effective wavelengths were selected by SRA from spectral data after the two best preprocessing method--1^stDer and SNV.The linear model（PLS）and nonlinear model---least squares-support vector machine（LS-SVM）were respectively established basing on the full wavelengths and effective wavelengths.The best result was the LS-SVM model based on the full wavelengths with Rc=0.8444 and RMSEc=3.2051 for modeling set and Rp=0.8437 and RMSEp=0.8437 for prediction set.（2）Visualization of SPAD values of leaves was realized.Utilizing the optimal model--SNV-LS-SVM models built by the full wavelengths,SPAD values for each pixel in hyperspectral image were predicted.The visualized maps of distribution could demonstrate how the SPAD values varied intuitively from sample to sample.（3）A method exploited to elect optimum position of nitrogen measuring was proposed.SPAD distribution map was applied to achieve SPAD values of different leaf positions and sites.Correlation analysis and regression analysis were applied to analyze SPAD value and chlorophyll a,b,a+b content.The results demonstrated that the fourth leaf from the top and its top were optimal leaf positions and sites of nitrogen measuring,which was consistent with conclusion of the SPAD measured value.It indicates that the option,leaf position of nitrogen measuring by using hyperspectral imaging technology is feasible,which expands the scope of the application of hyperspectral imaging technology in precision agriculture.