•  
  •  
 

Turkish Journal of Agriculture and Forestry

Abstract

The swift, effective, and noninvasive assessment of citrus chlorophyll levels is important for the timely diagnosis of citrus growth status, and the decision-making of orchard management. In this study, unmanned aerial vehicles (UAVs) with multispectral remote sensing imagery were applied to the mandarin orange, through which five vegetation indices [Medium Resolution Imaging Spectrometer Terrestrial Chlorophyll Index (MTCI), Red-edge Chlorophyll Index (CIre), Nitrogen Reflectance Index (NRI), Red, and Normalized Difference Vegetation Index (NDVI)], which better reflect citrus photosynthetic activity and productivity, were screened. The stepwise linear regression (SLR) method, support vector machine (SVM) regression method, and radial basis function neural network (RBFNN) technique were employed to develop the chlorophyll inversion model for navel oranges, and the precision and relevance of the inversion model were evaluated using the coefficient of determination (R2), root mean square error (RMSE), mean relative error (MRE), and estimation accuracy (EA). The results revealed that the MTCI and CIre exhibited highly significant correlations with the citrus chlorophyll content, with correlation coefficients of 0.830 and 0.829, respectively. Additionally, the NRI, Red, and NDVI also showed certain correlations with the citrus chlorophyll content, with correlation coefficients of –0.547, –0.481, and 0.470, respectively. Moreover, the four metrics used to evaluate the SLR method fell short, with an R² value trailing behind that of both the SVM method and the RBFNN by 9.6% and 15.0%, respectively. The training concentration effect of the SVM method was the best, with an R2 = 0.738. In the actual test, the effect decreased, with an R2 = 0.714, and the RBFNN method had the best effect in the test concentration, with an R2 = 0.759. Additionally, the citrus chlorophyll content during the coloration period could be simulated based on the RBFNN model, with R2, RMS, RMRE, and EA values of 0.759, 0.053, 0.188, and 0.975, respectively. The results indicated good applicability of the RBFNN model in determining the citrus chlorophyll content, which could provide accurate guidance for the management of citrus orchards.

Author ORCID Identifier

HU LI: 0009-0002-6423-0537

SHIJIANG ZHU: 0009-0002-1007-3978

WEN XU: 0009-0004-7835-1190

ZHONGYONG YANG: 0009-0009-3005-2032

XINBO LIU: 0009-0004-5122-6509

WEIQI ZHANG: 0009-0008-4804-6637

DOI

10.55730/1300-011X.3270

Keywords

UAV remote sensing, chlorophyll, vegetation index, RBF neural network method, support vector machine

First Page

345

Last Page

355

Publisher

Scientific and Technological Research Council of Turkey (TUBITAK)

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Download

Share

COinS