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

Table 3

Hyperspectral and multispectral classification for cruciferous weeds, and wheat and broad bean crops using MLP and RBF neural networks according to the different sampling years.

Neuronal networks	Sampling years	Input data	Importance of variables	Neurons of hidden layer	Neurons of output layer	Overall classification; (%)
MLP	2007	Hyperspectral wavelengths (nm)	725, 720, 690	11	3	100,0
		Multispectral bands and spectral vegetation indices	B/G, R/B, B, R/G	4	3	100,0
	2008	Hyperspectral wavelengths	520, 585, 565	1	2	100,0
		Multispectral bands and spectral vegetation indices	NIR, G, NIR/B	6	2	100,0
	2009	Hyperspectral wavelengths	730, 595, 590	3	2	100,0
		Multispectral bands and spectral vegetation indices	B/G, G, NIR/B	5	2	99,4
	2010	Hyperspectral wavelengths	480, 490, 485	5	3	98,7
		Multispectral bands and spectral vegetation indices	B/G, R/G, RVI	8	3	98,1
RBF	2007	Hyperspectral wavelengths	705, 400, 710	10	3	88,7
		Multispectral bands and spectral vegetation indices	NIR/B, RVI, NIR/B	7	3	92,1
	2008	Hyperspectral wavelengths	620, 630, 605	4	2	100,0
		Multispectral bands and spectral vegetation indices	G	10	2	100,0
	2009	Hyperspectral wavelengths	610, 615, 655	9	2	94,8
		Multispectral bands and spectral vegetation indices	B/G, R/G, NIR/B	10	2	98,9
	2010	Hyperspectral wavelengths	415, 410, 420	10	3	80,4
		Multispectral bands and spectral vegetation indices	R/G, B/G, NIR/B	10	3	94,4