Hyperspectral Characteristics Of Rapeseed And Remote Sensing Monitoring Of Its Agronomic Parameters

The hyperspectral estimation method of rapeseed chlorophyll content is discusse in order to provide theoretical basis for efficient,non-destructive and large-scale monitoring of rapeseed chlorophyll and nitrogen content.The hyperspectral data and SPAD and NBI values of rapeseed at seedling stage,budding stage,flowering stage and poding stage are measured in Guanzhong area of Shaanxi Province.The removal spectra and first-order differential spectra of rapeseed at different growth stages are obtained.The variation rules of physical and chemical parameters,hyperspectral characteristics at different growth stages and hyperspectral curves of different SPAD and NBI are analyzed.The spectral index of the original spectrum,the spectral absorption characteristic parameters of the continuum removal spectrum and the "three-sided" parameters of the first-order differential spectrum are calculated,and the correlation between them and SPAD and NBI values is analyzed.Based on the original spectral characteristic band,spectral index,the continuum removal spectral characteristic band,the spectral absorption characteristic parameters,the first-order differential spectral characteristic band and the "three-sided" parameters are constructed.The SPAD,NBI ordinary and partial least squares estimation models are used to validate the accuracy of the models,and the inversion mapping is performed on UAV images.The main results are as follows:(1)At seedling and budding stage,the reflectivity of light absorption accelerated,but increased at flowering and poding stage.In the near infrared region,the leaf area index(LAI)and near infrared reflectance(NIR)of rapeseed increased with the growth and development of rapeseed.The SPAD values of different contents showed that the spectral reflectance of rapeseed decreased with the increase of SPAD values.From visible light to near infrared light,the reflectance increases due to the structure of plant cell wall and the overlapping of canopy leaves,and the spectral reflectance increases with the increase of SPAD value.The NBI values of different contents increased with the increase of nitrogen content,and chlorophyll and other pigments were synthesized continuously,photosynthesis was enhanced,and spectral reflectance was decreased.From visible light to near infrared light,the reflectance increases.with the increase of NBI value,because the increase of nitrogen content promotes the growth of cell wall structure of plant leaves and the increase of leaf area index of canopy leaves.Continuous removal spectra can enhance the difference of plant spectralcurves,highlight the variation of absorption Valley and reflection peak,and the first-order differential spectra can eliminate background noise.With the increase of SPAD and NBI,the "red edge" curve moves towards the near infrared direction,and the phenomenon of "red shift" appears.(2)DSI,RSI and NDSI spectral indices were calculated on the basis of the original spectra.The correlation between DSI,RSI and NDSI and SPAD was tested by P < 0.01 significance test,and the correlation coefficients were above 0.6,showing a significant correlation.The spectral index with the best correlation between each growth period and SPAD value was selected as the independent variable to construct the general and partial least squares regression model of rape SPAD value.Compared with the original spectrum,the estimation accuracy was significantly improved.The correlation between DSI,RSI,NDSI spectral index and NBI was tested by P < 0.01 significance test,and all were greater than 0.55.The spectral index with the best correlation between each growth period and NBI value was selected as independent variable to construct the general and partial least squares regression model of rape SPAD value.Compared with the original spectrum,the estimation accuracy was significantly improved.(3)Compared with the original spectrum,the absorption characteristics of the continuum removal spectrum were amplified.The spectral reflectance increased from budding stage,flowering stage,seedling stage to pod stage,and the maximum absorption depth and the area of absorption Valley gradually increased.The estimation models of SPAD values of rapeseed leaves at different growth stages were established by removing spectral characteristic bands and absorption characteristic parameters with continuum.The R2 values of the models were better than those of original spectrum and first-order spectral differentiation,and had better estimation ability of SPAD values.Compared with the original spectrum,the correlation coefficients between NBI and continuum removal spectra are improved,and the determination coefficients R2 of the estimated and validated models are increased.The NBI general and partial least squares regression models of rapeseed were constructed by selecting the absorption characteristic parameters which had the best correlation with NBI values as independent variables.Compared with the original spectra,the estimation accuracy was significantly improved.(4)In the 500-660 and 700-800 nm bands,the first-order differential spectra and SPAD of each growth period passed F=0.01 significance test,and the correlation from small to large was: pod stage,seedling stage,flowering stage and bolting stage.The "three-sided" parameters of the first-order differential spectrum of rapeseed in different growth stageswere extracted,and the "three-sided" parameters with the best correlation between each growth stage and SPAD value were selected as independent variables to construct the general and partial least squares regression model of rapeseed SPAD.The fitting degree of the model in each growth stage was good,and the accuracy of the model was better than that of the original spectrum,which indicated that the model had better prediction effect on rapeseed SPAD value.Compared with continuum removal spectra,the first-order differential spectra and NBI values of rapeseed were higher in estimating model and validating model decision coefficient R2 at bolting and flowering stages.The decision coefficient R2 of the seedling estimation model was 0.47,which was 0.13 higher than that of the original spectral seedling estimation model.NBI estimation models at different growth stages were constructed by using the first-order differential spectrum "three-sided" parameters.Compared with the model constructed by spectral absorption characteristic parameters removed by continuum,the estimation accuracy was improved by a small margin.(5)The remote sensing mapping of SPAD and NBI values using UAV images is tested,and the mapping results are better.It can be seen that the SPAD values and NBI values in the West are slightly higher than those in the east,and the mapping predicted values and ground measured values verify that R2 are 0.536 and 0.457,respectively,which are in line with the actual situation.