Feasibility Study of Sugarcane Yield Prediction Using NDVI, CIred edge Indices Associated with Volume of Digital Surface Model (DSM)

Abstract

The objective of this study was to investigate the feasibility of predicting sugarcane yield in the field using some vegetation indices including NDVI and CIred edge associated with volume of digital surface model (DSM). In this study, 30 samples were used, consisting of 30 varieties of 1 replicates from the same plot. Therefore, there are only 1 sample in each sample. Each sample is 6 × 8 square meters. The images of sugarcane field were captured using a multispectral camera, covering 5 wavebands i.e. Red, Green, Blue, NIR and Red Edge, installed on an unmanned aerial vehicle. The sugarcane was then harvested and weighed to determine the yield. Images from each wavelength were analyzed by Pix4D Mapper software and calculated for vegetation indices (NDVI and CIred edge) and the volume of DSM. Correlations between the three indices and actual sugarcane yield were observed. The results showed that the NDVI, CIred edge and volume of DSM with correlated with actual yield at the correlation coefficient (r) of 0.77, 0.78 and 0.87, respectively, and the Root Mean Square Error (RMSE) of 3.18, 3.11 and 2.46, ton/rai, respectively and the Root Mean Square Error of Cross Validation (RMSECV) of 3.37, 3.32 and 2.61, ton/rai respectively, The tolerances are quite high due to the harvesting of photos near the harvest. Photo analysis does not cut out non-cane pixel parts. Therefore, the error is quite high. suggesting a feasibility to predict the cane yield Further study is necessary to improve the prediction accuracy.