Peanut Growth Monitoring And Yield Estimation Based On UAV Multispectrum

In order to facilitate rapid and accurate monitoring of crop growth information and estimation of crop yield in agricultural production in this study,peanut,a major crop in China,was selected for research.Using drone multispectral images,the peanut growth vegetation index was obtained.Each vegetation index was used as an independent variable,and the chlorophyll content,leaf area index,soil organic matter content index,and yield of the peanut during its seedling,flowering,podding,and mature stages were used as dependent variables,The optimal fitting models for predicting peanut growth and yield at various growth stages were constructed,and the relevant models were tested using independent data.The specific results are as follows:(1)Among the monitoring models for the chlorophyll content of peanut during the four growth stages,the monitoring model based on NGRDI,NDVI,GNDVI,and NDRE has the best effect in order,with R~2of 0.657,0.583,0.581,and 0.457,respectively.The monitoring results at the early growth stage are relatively ideal,and it is not suitable for monitoring at the later stage.(2)Among the four leaf area index monitoring models for peanut growth stages,the monitoring model based on NGRDI,NGRDI,NGRDI,and NDVI has the best effect,with R~2being 0.659,0.638,0.613,and 0.589,respectively.The LAI monitoring at flowering and podding stages is more suitable.(3)Among the monitoring models for soil organic matter content during the four growth stages of peanut,the monitoring model based on NDRE,NDRE,OSAVI,and DVI has the best effect,with R~2being 0.607,0.682,0.601,and 0.653,respectively.Monitoring of soil organic matter content should be carried out during the middle growth stage of peanut.(4)Using coefficient of variation,a comprehensive monitoring index for peanut growth was created.Multiple linear regression,partial least square regression,and support vector machine regression were used to establish a comprehensive growth monitoring model for peanut at four growth stages.Among these models,it was found that support vector machine regression performed best at seedling stage,with an R~2value of 0.734.At flowering stage,partial least square regression performed best,with an R~2value of 0.752.During the pod setting stage,support vector machine regression performs best,with a maximum R~2value of0.747.Support vector machine regression performed best at maturity,with an R~2of 0.755.(5)In the peanut yield estimation model,the correlation between the vegetation indices and yield in the early growth stage is poor,making it difficult to estimate the yield;However,in the middle and late stages of growth,peanut yield can be effectively estimated.Monitoring models based on NDVI,NGRDI,DVI,and SAVI at pod setting and maturity stages are the best,with R~2of 0.505 and 0.692,respectively.Other univariate or multivariate linear regression models have poor results.Integrating the models established for each growth period,it is concluded that NGRDI is more suitable for monitoring chlorophyll content and leaf area index,and NDRE is suitable for monitoring soil organic matter content.The early monitoring results of chlorophyll content and LAI showed that these data were highly accurate,but the late model did not achieve the desired results;The method of monitoring soil organic matter content is not significantly affected by the growth period.Through the model proposed in this article,the growth status of peanuts can be monitored and diagnosed in real time,providing an effective reference basis for peanut yield estimation and other related work.