Determination of Wheat Heading Stage Using Convolutional Neural Networks on Multispectral UAV Imaging Data

Table 4

Comparison of the accuracy of FHB preventive control time based on different monitoring models.

Monitoring model	Input band	Accuracy (%)
1D CNN + DT	550 + 660 + 730 + 790 nm	97.50
	550 + 660 + 730 nm	93.06
	550 + 660 + 790 nm	91.53
	550 + 730 + 790 nm	88.47
	660 + 730 + 790 nm	86.53
	DI	87.92
	DVI	91.39
	NDVI	89.72
	CIrededge	87.36
	GNDVI	87.36%
	TVI	89.36
NN	550 + 660 + 730 + 790 nm	67.22
	550 + 660 + 730 nm	67.22
	550 + 660 + 790 nm	67.22
	550 + 730 + 790 nm	63.75
	660 + 730 + 790 nm	56.94
	DI	56.94
	DVI	56.82
	NDVI	56.94
	CIrededge	59.17
	GNDVI	57.64
	TVI	63.33
SVM	550 + 660 + 730 + 790 nm	95.00
	DI	60.00
	550 + 660 + 730 nm	83.06
	550 + 660 + 790 nm	89.31
	550 + 730 + 790 nm	88.19
	660 + 730 + 790 nm	82.44
	DVI	71.81
	NDVI	72.22
	CIrededge	71.81
	GNDVI	60.97
	TVI	76.81
DT	550 + 660 + 730 + 790 nm	91.94
	DI	86.94
	550 + 660 + 730 nm	92.64
	550 + 660 + 790 nm	91.94
	550 + 730 + 790 nm	92.08
	660 + 730 + 790 nm	91.22
	DVI	89.17
	NDVI	88.33
	CIrededge	85.28
	GNDVI	89.17
	TVI	87.92