Suitable vegetation indices for predicting sugarcane Brix content in the field

Abstract

The goal of this research was to search for the optimal vegetation indices to predict the Brix content of sugarcane. The sugarcane of Khon Kaen 3 at age of 10 to 13 months was investigated. The models were optimized using regression equation by generating the relationship between four vegetation index (i.e NDVI GNDVI PVR and RVI) taken from multispectral image and measured °Brix content from ground sampling. The calibration model was developed using the sugarcane sample obtained from Sugarcane Breeding Institute, Khon Kaen university, while validation sets were the sugarcane sample from Bahn Phai Distric, Khon Kaen Province, and Saraburi province. The result showed that the overall °Brix content ranged between 16.8 and 24.8 ^oBrix. The vegetation index of GNDVI gave the best performance to be used as calibration model, providing the linear relationship of Y = -22.482x+31.832, where Y is predicted °Brix, x is GNDVI index. For the validation results, sugarcane in Bahn Phai Distric, Khon Kaen Province had the Bias value of 0.40 ^oBrix, and SEP value of 1.64 ^oBrix, meanwhile Saraburi province plant provided the Bias value of 2.80 ^oBrix, and SEP (standard error of prediction) value of 0.80 ^oBrix. The result indicated that multispectral imaging across the GNDVI index could be used as a real monitoring of cane quality in the field.