Estimation of Leaf Chlorophyll Concentration in Turmeric (Curcuma longa) Using High-Resolution Unmanned Aerial Vehicle Imagery Based on Kernel Ridge Regression

Abstract

High-resolution information is needed for precision agriculture to achieve precise management of inputs. High spatial and temporal resolution is requisite to get the actionable information for the timely response. The objective of the present study is to estimate the leaf chlorophyll Concentration using high-resolution (2 cm) images captured from UAV-mounted multispectral sensors for crop health monitoring. In this study, a hexacopter was flown at an elevation of 25 m to capture the images in green, red, red edge and NIR bands of turmeric plots grown at ICAR Research Complex, Northeast Hilly Region, India. A handheld SVC spectroradiometer having spectral range from 350 to 2500 nm was also used to collect the spectra of sample plants to support the UAV study. We evaluated an advanced machine learning algorithm kernel ridge regression combined with spectral information and ground-truth chlorophyll data to model the chlorophyll estimation. The multivariate analysis was also applied on spectroradiometer and UAV data, which recommended red band for chlorophyll prediction with R2 value greater than 0.6. We also found that kernel ridge regression is a robust method for developing chlorophyll estimation model with lesser training time. The results indicate that kernel ridge regression with a radial basis kernel function with four multispectral input bands can be utilized to evaluate the leaf chlorophyll concentration with an root mean squared error RMSE = 0.10 mg/g and regression coefficient R2 = 0.7452. However, this study is site specific and needs to be practiced in different crop sites in order to generalize this method for precision agriculture.