Inversion Of Cotton Main Growth Parameters Based On Unmanned Aerial Vehicle(UAV) Remote Sensing Image

Rapid and nondestructive access to agricultural information can realize the monitoring of crop growth and the prediction of crop yield.It is the necessary premise for managers to obtain crop growth information and make management decisions.It is also the important basis for accurate management of agricultural production.Cotton,as an important economic crop,plays an important role in the national economy.The leaf area index is an important parameter to predict cotton growth and yield.The process of measuring cotton leaf area index on the ground is complex and inefficient.If the UAV flight platform can be used for rapid monitoring,it is of great significance to achieve accurate management of farmland.In this paper,the multi spectral image of cotton canopy is obtained by using UAV remote sensing monitoring platform,and the inversion model of cotton leaf area index is constructed to realize the rapid estimation of its value.Based on the response of crops to the spectral characteristics of different bands,using the vegetation index obtained from the experimental image and the measured cotton leaf area index on the ground,the empirical model method and BP neural network model method are used to inverse the cotton leaf area index,and the accuracy of the inversion model is evaluated to select the best inversion model.The main research contents and conclusions are as follows:(1)Experimental data collection.The remote sensing platform composed of Dajiang M600 UAV equipped with Sequoia camera is used to collect the image of the test area,and then the image of the test area is spliced by pix4d mapper image processing software to meet the test requirements.(2)Empirical model fitting analysis.Using empirical model formula,linear,exponential and logarithmic models of NDVI-LAI,GNDVI-LAI NDRE-LAI and RVI-LAI are constructed respectively for fitting analysis,and decision coefficient(R~2)and root mean square error(RMSE)are used for accuracy evaluation of inversion model to find the best LAI inversion model.Through comparative analysis,it is found that the RVI-LAI index model has the best estimation results,with a determination coefficient of0.9113,followed by NDVI-LAI index model,GNDVI-LAI index model and NDRE-LAI index model,with a determination coefficient of 0.8138,0.5844 and 0.3478,respectively.(3)BP neural network model fitting analysis.Using BP neural network method,five BP neural network fitting models,NDVI-LAI,GNDVI-LAI,NDRE-LAI,RVI-LAI and VI-BP are constructed.Finally,the decision coefficient and root mean square error are used to evaluate the fitting accuracy of the model.The results showed that the VI-BP neural network model composed of the input layer of 4 vegetation index had the best fitting effect,with a decision coefficient of 0.9093 and a root mean square error of0.3755.(4)Model comparison and optimization.The better RVI-LAI index model,NDVI-LAI index model,GNDVI-LAI index model,NDRE-LAI index model,VI-BP neural network model and RVI-BP neural network model are used to predict 60validation samples,and the prediction accuracy is evaluated by decision coefficient(R~2)and root mean square error(RMSE).It is found that the determination coefficient of RVI-LAI index model is 0.9113,which is better than that of VI-BP neural network model in terms of both determination coefficient and root mean square error.Therefore,the index model based on RVI and LAI has the best effect in inversion of leaf area index.