Authors

Abstract

This study mainly investigates the feasibility of monitoring and estimating the RCC (Relative Chlorophyll Content) and the LAI (Leaf Area Index) of sorghum by coupling integrated learning model with UAV multispectral image, clarifies the quantitative relationship between RCC and LAI of sorghum and the vegetation index based on different spatial resolutions, and constructs a Monitoring and prediction model for the RCC and the LAI of sorghum based on the UAV multispectral image and the vegetation index at different spatial resolutions. The model constructed based on integrated learning, and using the stacking approach had good prediction accuracies at three spatial resolutions, with the stacking model predicting R2=0.87, MAE=18.27, and RMSE=22.23 for the RCC at spatial resolution of 0.017 m; R2=0.86, MAE=17.38, and RMSE=23.21 for RCC at spatial resolution of 0.024 m; R2=0.80, MAE=18.62, and RMSE=24.12 for RCC at spatial resolution of 0.030 m; R2=0.93, MAE=0.34, and RMSE=0.37 for LAI at spatial resolution of 0.017 m; and R2=0.89, MAE=0.44, and RMSE=0.55 for LAI at spatial resolution of 0.024 m. The model established by combining the vegetation index and integrated learning can quickly and accurately monitor and predict RCC and LAI of sorghum, which provides a scientific methodology and theoretical basis for scientific monitoring and predicting RCC and LAI of sorghum in the field.

Abstract in Chinese

本研究主要探讨了将集成学习模型与无人机多光谱影像耦合，监测和估算高粱叶绿素相对含量和叶面积指数的可行性，明确了基于不同空间分辨率的高粱叶绿素相对含量和叶面积指数与植被指数之间的定量关系，构建了基于无人机多光谱影像和植被指数的不同空间分辨率高粱叶绿素相对含量和叶面积指数的预测模型。基于集成学习和堆叠方法构建的预测模型在三种空间分辨率下均具有良好的预测精度，其中堆叠模型在空间分辨率为 0.017m时预测叶绿素相对含量的R2=0.87，MAE=18.27，RMSE=22.23；在空间分辨率为 0.024m时预测叶绿素相对含量的 R2=0.86，MAE=17.38，RMSE=23.21；空间分辨率为 0.030 m 时叶绿素相对含量的 R2=0.80，MAE=18.62，RMSE=24.12；空间分辨率为 0.017m 时叶面积指数的 R2=0.93，MAE=0.34，RMSE=0.37；空间分辨率为 0.024m时叶面积指数的 R2=0.89，MAE=0.44，RMSE=0.55。植被指数与集成学习相结合而建立的模型能够快速且准确地监测和预测高粱的叶绿素相对含量和叶面积指数，为科学监测、预测田间高粱的叶绿素相对含量和叶面积指数提供了科学的方法和理论依据。