Cotton Science

Estimation of chlorophyll content in cotton canopy using UAV multispectral imagery and machine learning algorithms

Zhao Xin, Li Zhaoyang, Wang Hongbo, Liu Jiangfan, Jiang Wenge, Zhao Zeyi, Wang Xingpeng, Gao Yang

Cotton Science, 2024, 36(1): 1-13. DOI: 10.11963/cs20230026

植被指数 Vegetation index	一元线性回归模型 Simple linear regression model	建模集 Modelling dataset			验证集 Validation dataset
植被指数 Vegetation index	一元线性回归模型 Simple linear regression model	R²	RMSE/(mg·L^-1)	RPD	R²	RMSE/(mg·L^-1)	RPD
NDVI	y=56.16x-26.688	0.542	1.314	1.477	0.384	1.627	1.274
RVI	y=0.987x+9.175	0.547	1.306	1.486	0.341	1.683	1.232
NDCI	y=40.712x-8.574	0.258	1.671	1.161	0.287	1.749	1.185
RECI	y=0.988x+10.163	0.547	1.306	1.486	0.341	1.683	1.232
GCI	y=0.987x+11.236	0.549	1.275	1.455	0.392	1.732	1.384
NDRE	y=65.713x+36.574	0.373	1.712	1.239	0.316	1.752	1.265
SAVI	y=37.463x-26.687	0.541	1.313	1.477	0.384	1.274	1.627

Table 5 Simple linear model of vegetation index and SPAD and the model validation

Other figure/table from this article

Fig. 1 Overview map of the experimental area
Table 1 Basic soil nutrient content of experimental field
Table 2 Multispectral camera sensor parameters
Table 3 Calculation formula for vegetation index
Table 4 Statistical description of chlorophyll content in cotton canopy leaves
Fig. 2 The correlation coefficient between vegetation index and chlorophyll content
Fig. 3 Distribution of measured cholorophyll content and the predicted chlorophyll content value from machine learning regression model A-F are methods based on LASSO regression, RR, PLSR, KNNR, RFR, and SVR, respectively.
Table 6 Estimation results of machine learning regression
Fig. 4 Inversion mapping of cotton canopy chlorophyll content using the RFR model