| Crop biomass is an important index reflecting crop growth,crop breeding and management,and also one of the key factors affecting crop yield and yield.Potato plays an irreplaceable role in food security.Using remote sensing technology to obtain potato biomass information and estimate yield can provide decision-making information for farmland production management.The main research contents and conclusions of this paper are as follows:(1)Estimation of potato biomass based on UAV digital imageBased on the digital image data of the drone,calculate the digital image index,obtain the color spectrum data and the corresponding color space vegetation index through the color conversion space HSI.Correlation analysis was performed using image vegetation index and above-ground biomass of potato,and the correlation index was screened for significant correlation,and combined with univariate analysis,multiple linear regression,partial least squares and other methods to construct above-ground biomass estimation models.At the same time,the same operation was performed on potato underground biomass.1)The results of aboveground biomass estimation showed that in the mature stage,the model accuracy of multiple linear regression method was the highest,modeling R~2 and RMSE were 0.658 and131.158kg/mu respectively,and validation R~2 and RMSE were 0.587 and264.001kg/mu respectively.2)The results of underground biomass estimation show that in the mature period,the model accuracy of multiple linear regression method is the highest,modeling R~2 and RMSE are 0.533 and 315.203kg/mu respectively,and validation R~2and RMSE are 0.417 and 718.776kg/mu respectively.(2)Estimation of potato biomass based on drone hyperspectral imagesAnalyze the spectral reflection characteristics of UAV hyperspectral images under different conditions,analyze the correlation between hyperspectral information and biomass in different bands,and select the best original spectrum,construct the hyperspectral image vegetation index,and correlate with biomass Analysis,screen out significant correlation vegetation index,use univariate analysis method,multiple linear regression method and partial least square method to build biomass estimation model,and analyze by sub-method and growth period.Different analysis results show that in the mature stage,the model accuracy of multiple linear regression method is the highest,modeling R~2 and RMSE are 0.703 and 106.831kg/mu respectively,and validation R~2 and RMSE are 0.74 and 107.068kg/mu respectively.(3)Potato coverage extraction based on UAV digital imageThis chapter constructs a new vegetation coverage extraction method(H-A)based on the vegetation determination flow chart and other methods,and compares it with vegetation index method and maximum likelihood supervised classification method.The results show that H-A method has high accuracy and is a reliable vegetation coverage extraction method.(4)Potato yield estimation based on UAV imageBased on the estimation results of biomass and coverage,combined with the measured output values,the yield estimation method using the parameter(-biomass)-yield model is used to construct a yield estimation model. |
