Table 1.

Application of spectroscopy techniques for seed phenotype digitization. Abbreviations: DA, discriminant analysis; DBN, deep belife network; iPLS, interval partial least squares; LDA, linear discriminant analysis; LOD, limit of detection; MPLS, modified partial least squares; MPLSR, modified partial least square regression; NIR, near-infrared; PCA-SVM, principal component analysis-support vector machine; PLS-DA, partial least squares discriminant analysis; PLSR, partial least squares; SKNIR, single-kernel near-infrared.

Optical sensor	Pros and cons	Seed	Application	Method and performance	Ref
NIR	Affordable price High penetration depth Poor sensitivity for low concentrationsBroad and overlapping absorption bands	Soybean	Protein and oil content determination	SAS, NIR analyzer	[21]
			Amino and fatty acid determination	ISI program, PLSR, R2 = 0.06 ∼ 0.85	[22]
		Sunflower	Oleic acid determination	Regression analysis, R2 = 0.983	[23]
		Brassica napus	Tocopherol content determination	MPLS; R2 = 0.74	[24]
		Wheat	Fusarium-damaged kernels and deoxynivalenol identification	SAS, SKNIR, R = 0.72	[25]
		Sesame	Origin discrimination	IBM SPSS Statistics, DA, accuracy = 89.4 %	[26]
		Maize	Provitamin A Carotenoids Content determination	WinISI III Software, bayesian and MPLSR, R2 = 0.22 ∼ 0.75	[27]
			Viability evaluation	Matlab, PLS-DA, accuracy > 98 %	[28]
		Forage grass	Seed germination and vigor evaluation	R software, PLS-DA, accuracy = 61 ∼ 82 %	[29]
MIR	High specificity Few overlaps Low sensitivity for quantitative analysisHigh requirements for analysis models	Soybean	Isoflavones and oligosaccharides determination	Matlab, PLSR, R2 = 0.72,0.80	[30]
		Pea	total protein, starch, fiber, phytic acid, and carotenoids determination	Orange, PLSR, R > 0.71	[17]
		Brown rice	Aflatoxin contamination identification	Matlab, DA, accuracy = 90.6 %	[31]
		Peanut	Fungal contamination levels identification	TQ Analyst, PLSR, R2 = 0.9157	[32]
Fluorescence	High sensitivity High specificity Cluttered and weak signalBackground interference	Pea	Mineral nutrient (K, Ca, Mn, Cu, Zn, and Se) analysis	PyMca, R > 0.85	[33]
		Maize	Aflatoxin contamination identification	IBM SPSS Statistics, LDA, accuracy = 100 %	[34]
		Rice	Seed germination and vigor evaluation	DBN, R = 0.9792	[35]
		Brassica oleracea	Seed maturity and quality evaluation	–	[36]
Raman	High specificity Good signal-to-noise ratio Hardly disturbed by water molecules Expensive experiment materialsUnsuitable for fluorescent samples	Soybean	Crude protein and oil content determination	Matlab, PLSR, R2 = 0.916, 0.872	[37]
		Rapeseed	Iodine value determination	OPUS IDENT, R = 0.9904	[38]
		Kidney beans	Deoxynivalenol identification	Gaussian 03 package, LOD = 10-6M	[39]
		Maize	Aflatoxin contamination identification	SAS, PLSR, R2 = 0.941 ∼ 0.957	[40]
			Transgenic maize discrimination	Matlab, LDA, accuracy = 87.5 %	[41]
			Viability evaluation	Matlab, PLS-DA, accuracy > 95 %	[42]
THz	Fingerprint spectrum High sensitivity Mostly used for solution detectionHardly nondestructive detection	Maize	Moisture content determination	PLSR, R = 0.9969	[43]
			Viability evaluation	–	[44]
		Suger beet	Seed quality identification	Python, accuracy = 87 %	[45]
		Wheat	Seed quality identification	PCA-SVM, accuracy = 95 %	[46]
			Variety discrimination	Matlab, iPLS, R = 0.992	[47]
		Rice	Transgenic rice discrimination	TQ Analyst, DA, accuracy = 89.4 %	[48]