Improved estimation of aboveground biomass in wheat from RGB imagery and point cloud data acquired with a low-cost unmanned aerial vehicle system

Table 9 Accuracy assessment for the estimation of AGB in wheat from the selected input variables with SMLR and three machine learning techniques

Input variables	Technique	Calibration (N = 201)			Validation (N = 85)
Input variables	Technique	R ²	RMSE (t/ha)	AIC	R ²	RMSE (t/ha)	rRMSE (%)
VIs	SMLR	0.69	1.52	742.63	0.70	1.58	34.49
	SVR	0.72	1.49	696.91	0.65	1.70	37.19
	ELM	0.68	1.56	758.14	0.68	1.64	35.77
	RF	0.70	1.49	438.29	0.64	1.73	36.93
Canopy height metrics	SMLR	0.68	1.53	747.72	0.72	1.54	33.61
	SVR	0.70	1.52	697.85	0.73	1.49	32.66
	ELM	0.69	1.50	747.92	0.72	1.54	33.72
	RF	0.72	1.46	431.34	0.76	1.40	30.76
VIs and canopy height metrics	SMLR	0.72	1.43	717.41	0.75	1.46	31.77
	SVR	0.74	1.38	679.36	0.73	1.51	33.07
	ELM	0.72	1.44	714.49	0.73	1.50	32.77
	RF	0.76	1.33	375.06	0.78	1.34	29.31

The input variables for each technique were selected with SMLR. The accuracy metrics were calculated from calibration data and validation data separately. The number in bold for each column represents the maximum R², minimum RMSE, minimum AIC and minimum rRMSE, respectively

ISSN: 1746-4811