Research On Summer Corn Quality Monitoring And Forecasting Model Based On Near-ground Hyperspectral Remote Sensing

As an important technical means in precision agriculture,hyperspectral technology is widely used in many aspects such as crop growth monitoring,crop yield estimation,and quality inspection.Using remote sensing technology,the predecessors have carried out a large number of researches on inversion and yield estimation of physicochemical indicators,but less research on summer corn.For the study of corn quality,former scholars mainly used near-infrared spectroscopy analysis technology to measure the spectral information of grains,construct prediction models of grain proteins,and rarely use canopy hyperspectral information to carry out grain protein estimation research.Therefore,this study takes summer corn as the research object,and analyzes the reflection characteristics of summer corn canopy spectrum and the response law of canopy spectrum and physical and chemical indicators under field management conditions of different varieties,different planting densities and different nitrogen fertilizer application rates.Based on the identification of five sensitive bands for physical indicators,combined with the hyperspectral vegetation index,an estimation model for crop indicators based on vegetation index,and an estimation model for yield and grain protein based on vegetation index and physical indicators.The main work of this study is as follows:(1)Comparing the canopy spectral curves of summer maize under different field management,the spectral differences between the varieties with the big trumpet stage are the most obvious.The differences of the varieties are mainly reflected in the compactness of the plant leaves;in the range of 350nm-1350 nm,the planting density is positively correlated with the canopy spectral reflectance;in the range of 750nm-1350 nm,with the increase of nitrogen application in the field,the spectral reflectance of crops gradually increased,but in the visible light band showed the opposite law.(2)Comparing the response characteristics of the canopy spectrum under different physical and chemical indicators,it is found that the five selected physical and chemical indicators have similarity in the response of the canopy spectrum,and all show a positive correlation;the correlation between the physicochemical index and the canopy spectrum was analyzed to clarify its sensitive band.Based on the hyperspectral vegetation index,a PLSR prediction model for the physicochemical index was constructed.The results show that the model has a good effect.The five physical and chemical indicators achieve the best results when the number of vegetation indices is n = 20,15,20,25,25.(3)The correlation between yield and vegetation index and physical and chemical indicators is analyzed,and AGB has the highest correlation with yield;based on the vegetation index and physical and chemical indicators,a yield estimation model is constructed.The model’s decision coefficient R2 is 0.604,which is an increase of 0.101 compared with the traditional linear regression model.The determination coefficient of the fitting result between the measured value and the predicted value is 0.584,which has a good prediction effect.(4)By analyzing the correlation of grain protein with spectral information and physical and chemical indicators,it was found that grain protein has good correlation with LNC and near-red band,respectively.The vegetation index based on the near-red wave band was selected,combined with the LNC during the heading and spinning stage to construct an estimation model for grain protein,in which the coefficient of determination of the model was 0.454.The determination coefficient of the fitting result between the measured value and the predicted value is 0.530.