Applying Neural Networks to Hyperspectral and Multispectral Field Data for Discrimination of Cruciferous Weeds in Winter Crops

Table 5

Classification matrix obtained from STEPDISC analysis (from 2007 to 2010), and from MLP neural network (for 2010) considering cross-validation and using multispectral wavebands and spectral vegetation indices for cruciferous weeds, and wheat and broad bean crops.

Sampling years	Classification methods	Observed	Predicted spectra
		spectra	Broad bean	Wheat	Cruciferous weeds
				(% of correct classification)
2007	STEPDISC	Broad bean	100	0	0
		Wheat	3	92	5
		Cruciferous weeds	0	0	100
2008		Broad bean	—*	—	—
		Wheat	—	100	0
		Cruciferous weeds	—	0	100
2009		Broad bean	—	—	—
		Wheat	—	99	1
		Cruciferous weeds	—	1	99
2010		Broad bean	99	0	1
		Wheat	2	96	2
		Cruciferous weeds	2	1	97
2010	MLP	Broad bean	97	1	2
		Wheat	0	100	0
		Cruciferous weeds	1	2	97

*Data not sampled.