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

Table 2

Hyperspectral and multispectral classification of cruciferous weeds, and wheat and broad bean crops using STEPDISC analysis according to the different sampling years.

Sampling; years	Spectral input data	Wavelengths (nm), multispectral bands, and spectral vegetation indices selected	Wilks’ Lambda	Exact-F	Overall classification; (%)	Cross validation; (%)
2007	Hyperspectral wavelengths	480, 405, 460, 705, 440, 680, 655, 595, 690, 515, 430, 445	0,000	244,5	99,3	98,0
	Multispectral bands and spectral vegetation indices	B, B/G, R/G, RVI, NIR, NIR/B, R/B	0,004	187,5	98,0	98,0
2008	Hyperspectral wavelengths	605, 690, 410	0,008	7387,5	100	100
	Multispectral bands and spectral vegetation indices	NIR/G, RVI, B/G, R/B	0,000	218990,5	100	100
2009	Hyperspectral wavelengths	725, 825, 815, 470, 430, 420, 410, 650, 665, 775, 540	0.131	517,8	99,2	99,2
	Multispectral bands and spectral vegetation indices	G, B/G, NIR/B, B, NIR/G, NDVI, R/B, R, RVI, R/G	0,111	681,9	99,3	99,1
2010	Hyperspectral wavelengths	735, 575, 485, 885, 410, 705, 525, 560, 750, 460, 405, 420	0,036	555,3	98,7	98,5
	Multispectral bands and spectral vegetation indices	B/G, R/G, NDVI, R/B, NIR/G, RVI, NIR/B, G, B, NIR, R	0,032	645,6	97,8	97,6