Hyperspectral imaging-based classification of rice leaf blast severity over multiple growth stages

Table 2 Model parameters and performances for SVM models throughout multiple growth stages

Year	Bands	Sample	Trainning set	Testing set	Model	Average accuracy	Micro	Macro
Year	Bands	Sample	Trainning set	Testing set	Parameters	Average accuracy	F1 score	F1 score
2019	Full	Level 0	36	16	C = 84.4485	\	\	\
		Level 1	32	14	G = 0.0063457	\	\	\
		Level 2	67	29		\	\	\
		Level 3	50	21		\	\	\
		Level 4	46	19		\	\	\
		Total	231	99		94.75%	0.869	0.883
	491 nm	Level 0	36	16	C = 36.7583	\	\	\
	671 nm^*	Level 1	32	14	G = 8	\	\	\
		Level 2	67	29		\	\	\
		Level 3	50	21		\	\	\
		Level 4	46	19		\	\	\
		Total	231	99		92.73%	0.818	0.833
	491 nm	Level 0	36	16	C = 51.9842	\	\	\
	668 nm^*	Level 1	32	14	G = 3.7321	\	\	\
		Level 2	67	29		\	\	\
		Level 3	50	21		\	\	\
		Level 4	46	19		\	\	\
		Total	231	99		92.32%	0.808	0.816
2021	All	Level 0	30	13	C = 1024	\	\	\
		Level 1	80	34	G = 0.0019531	\	\	\
		Level 2	34	15		\	\	\
		Level 3	50	21		\	\	\
		Level 4	29	13		\	\	\
		Total	223	96		92.92%	0.823	0.808
	491 nm	Level 0	30	13	C = 13.9288	\	\	\
	671 nm	Level 1	80	34	G = 3.0314	\	\	\
		Level 2	34	15		\	\	\
		Level 3	50	21		\	\	\
		Level 4	29	13		\	\	\
		Total	223	96		92.08%	0.802	0.713
	491 nm	Level 0	30	13	C = 238.8564	\	\	\
	668 nm	Level 1	80	34	G = 0.35355	\	\	\
		Level 2	34	15		\	\	\
		Level 3	50	21		\	\	\
		Level 4	29	13		\	\	\
		Total	223	96		92.5%	0.813	0.793

^*491 and 671 nm were selected sensitive wave bands for SRR data acquired in 2019 using SPA
^*491 and 668 nm were selected sensitive wave bands for SRR data acquired in 2021 using SPA

ISSN: 1746-4811