Research On Nutrition Information Of Facility Crops Based On Reflection Spectrum Image

Including two-dimensional spatial information,acquired based on hyperspectral imaging technology,the multi-band reflectance spectral data cube contains a large number of spectral data which can reflect the health of the crop.And the crop nutrient level is a major component of the crop health.Facility solanaceous crop pepper as the experimental object,this paper will focus on the description of the research that using reflectance hyperspectral imaging technology to acquire the crop NIR spectrum digital image at the leaf scale,finishing the analysis of image processing and spectral analysis,to extract the characteristic spectral data which can be used to reflect the plant leaf nutrition information,sorting data,then to established a model of crop nutrition information detection based on NIR reflectance spectroscopy.This study involved the uniform design of sample orientation multi-mineral nutrient elements stress cultivation.The effects of various mineral nutrient element interactions on the nitrogen fixation level of single element were considered,but the experimental detection target is locked to single nitrogen as the only standard parameter to reflect the level of crop nutrition eventually.This approach is not only simple and effective,but also has sufficient theoretical basis.After the actual statistical analysis,there was a significant linear correlation between the absorption and fixation amount of nitrogen and the application amount of nitrogen,phosphorus and iron nutrition.The results showed that the nitrogen content in crops could reflect the other elements supply in a certain range of nutrient stress,and this just explains the general reliability that the nitrogen level in the crop as the measurement standard of the crop overall nutrition level.The multi-collinearity among the independent variables was eliminated by ridge regression.The final near-infrared reflection hyperspectral crop nutrient detection model contains only 3 near-infrared band spectral reflectance as the independent variable of the prediction model: R1043(of Band 43),R1198(of Band 87)and R1662(of Band 229).As for the final prediction model,p = 0.002,which means the model is extremely significant.The coefficient of determination R2 = 0.857,indicating that the model has a very good prediction performance.According to the variable coefficient ratios,the gray image signals of the three variables selected inthe model were scaled and recombined into an NIR fusion image.Compared with the single feature band gray images,the fusion image with three feature bands has an obvious higher nutrition differentiation performance.