Remote Sensing Estimation Of Anji White Tea Yield Based On Multi-source Data

Anji White Tea is produced in Anji County,Huzhou City,Zhejiang Province.It is a national geographical indication product.The annual yield estimation of White Tea is an important part of production,but the current yield estimation method relies on expert experience and lacks a large-area rapid plot-scale remote sensing yield estimation method.This problem limits the overall planning of tea picking,tea pickers recruitment,spring tea frying,and sales.It also severely affects the management efficiency of Anji White Tea from picking to market circulation.Based on the above-mentioned problems,this project anticipates the estimation of Anji White Tea production elements:tea garden plot distribution,picking period and picking yield distribution model as the monitoring carrier to conduct remote sensing estimation research on Anji county White Tea production.This paper includes three studies:extraction of tea garden area in Anji County,prediction of tea garden plot-scale picking period,and remote sensing estimation of single picking yield.The process of this research is to extract the White Tea planting area of Anji County through Sentinel-2 satellite images before and after the tea garden pruning period in 2019;using temperature as the driving factor,weather forecast data and DEM digital elevation model as the means of improvement,establish the tea garden plot scale picking period Prediction model;using ASD portable ground object spectrometer and UHD185 airborne hyperspectral imager,combined with field-measured yield,chlorophyll,and leaf area data,establish a tea tree yield estimation model based on vegetation index.The research results of this paper not only provide theoretical support for the remote sensing yield estimation of Anji White Tea at the plot scale,but also promote the digitization of tea output information and innovate tea garden production management models.The main tasks of this paper are:(1)Extraction of tea garden area in Anji County:First,analyze the spectral characteristics and vegetation index characteristics of Anji White Tea before and after the pruning period,and then combine the classification methods such as decision tree and support vector machine to classify and extract the tea garden area in the study area Finally,combining the field survey data and the sampling data not used for classification extraction,the accuracy analysis of the classification extraction results was carried out to obtain a high-precision tea garden distribution map in Anji County.The classification accuracy reached 99.04%,and the Kappa coefficient was 0.96.Provide data for forecasting the picking period of tea gardens in Anji County.(2)Prediction of the picking period of tea garden plots in Anji county:Firstly,by collecting and analyzing the meteorological data of Zhejiang Province over the years and the picking time of Anji White Tea in various counties and cities,the meteorological factors that have a significant correlation with the growth of White Tea are obtained to establish the forecast of the picking period of Anji White Tea For the model,the R2 values of the prediction models for the mining period and the end of harvesting period are 0.676 and 0.705,respectively,and the Sig values are all less than 0.05,indicating that the model is significantly correlated.Then through the weather forecast data and DEM digital elevation model,the picking period forecast model is improved and upgraded,so that it has iteratively updated forecast results and temporal and spatial distribution characteristics,and combined with the tea garden distribution map of Anji County obtained in(1),the picking The forecast results are specific to the tea garden scale.The root mean square error is used as an evaluation index to analyze the error between the picking period forecast data and the field survey data.The RMSE of the forecast model for the mining period and the picking end period are calculated to be 1.44 respectively.And 2.13,it shows that the forecast model can predict the plots of tea plantations in Anji County with high precision.The sampling and verification of the tea gardens’ mining and end picking times in Zhejiang Province are conducted.The root mean square errors of the prediction data and the field survey data are 1.92 and 2.54,respectively,indicating The prediction results of the model are also valid for the expansion of the spatial scope in Zhejiang Province.(3)Remote sensing estimation of single-picking yield:First,the UAV is used to obtain hyperspectral remote sensing data before and after a single-picking of tea,and the ground ASD,leaf area and chlorophyll measured data are simultaneously obtained,and the research has established a method based on leaf area index and chlorophyll index.A hyperspectral tea yield estimation model based on remote sensing.The research results show that the yield estimation model based on leaf area index R750/R710 has the lowest root mean square error of 122.836,and based on the chlorophyll index CARI,the yield estimation model has a minimum root mean square error of 163.825,which can meet the accuracy requirements of Anji white tea yield estimation.In order to reduce the cost of obtaining remote sensing data from UAVs and enable the production estimation model to be widely used in actual agricultural production,this study further explored the correlation between RGB vegetation index and tea production,and established the regression equation through the pigment index NGI,The root mean square error of the test is 134.346,laying a theoretical foundation for large-scale,low-cost white tea production estimation.