The Hyperspectral Inversion Method of Main Ionic Compounds Content in Groundwater based on BP Neural Network

Qing GUO; Lifu ZHANG; Wenchao Qi; Linshan ZHANG

doi:10.11873/j.issn.1004-0323.2024.1.0149

Remote Sensing Technology and Application, Volume. 39, Issue 1, 149(2024)

The Hyperspectral Inversion Method of Main Ionic Compounds Content in Groundwater based on BP Neural Network

Qing GUO^1,2、*, Lifu ZHANG¹, Wenchao Qi¹, and Linshan ZHANG^1,2

Author Affiliations

¹Aerospace Information Research Institute，Chinese Academy of Sciences，Beijing 100101，China

²University of Chinese Academy of Sciences，Beijing 100049，China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(7)

Fig. 1. Average spectra

Download full size

View in Article

Fig. 2. BPNN inversion model fit

Download full size

View in Article

Table 1. Modeling results of KCl， CaCl₂andNaCl based on SPA feature spectrum selection

View table

View in Article

Table 1. Modeling results of KCl， CaCl₂andNaCl based on SPA feature spectrum selection

光谱指标	预处理	个数	谱段位置	训练集			测试集
光谱指标	预处理	个数	谱段位置	R²	RMSE	RPD	R²	RMSE	RPD
KCl	MD	2	238, 1	0.733 5	2 027.55	1.670 6	0.867 9	1 727.43	2.125 1
	MD-SG	15	102、21、19、20、15、18、17、7、34、24、3、5、26、9、8	0.940 9	1 019.18	3.996 3	0.963 0	864.58	5.065 9
	MD-SG-LR	13	102、21、20、19、15、17、7、24、18、8、3、9、26	0.946 7	1 000.69	4.222 6	0.958 0	847.43	4.818 5
	MD-SG-CR	7	219、40、29、20、16、6、30	0.796 7	1 942.95	1.983 4	0.805 4	1 853.38	1.889 6
	MD-SG-NR	15	102、70、19、21、34、17、20、18、3、26、5、7、22、14、24	0.940 4	1 004.17	3.980 9	0.966 1	840.44	5.225 8
CaCl₂	MD	2	291、5	0.551 3	2 236.12	1.110 5	0.622 7	2 660.37	1.028 1
	MD-SG	2	268、34	0.581 2	2 283.67	1.180 1	0.667 3	2 369.60	1.200 1
	MD-SG-LR	2	290、216	0.560 5	2 518.31	1.131 3	0.792 4	1 620.80	1.582 8
	MD-SG-CR	7	117、54、216、30、43、48、282	0.572 8	2 501.15	1.159 8	0.757 2	1 714.94	1.373 4
	MD-SG-NR	2	272、34	0.646 7	2 094.84	1.355 2	0.783 9	1 916.60	1.688 7
NaCl	MD	2	187、10	0.900 6	914.68	3.025 9	0.962 9	7 563.19	4.764 9
	MD-SG	5	289、297、106、56、20	0.941 0	689.03	3.999 3	0.979 0	584.32	6.481 8
	MD-SG-LR	4	289、297、107、14	0.964 5	649.65	5.253 1	0.962 4	545.68	5.085 6
	MD-SG-CR	15	197、60、66、54、15、16、74、32、35、27、21、294、39、25、49	0.851 6	1 393.08	2.399 6	0.596 0	1 307.22	1.642 4
	MD-SG-NR	8	250、9、17、18、2、13、10、4	0.969 0	565.16	5.599 2	0.981 1	470.16	7.172 5

Table 2. Modeling results of KCl， CaCl₂andNaCl based on PCA characteristic spectral bands

View table

View in Article

Table 2. Modeling results of KCl， CaCl₂andNaCl based on PCA characteristic spectral bands

光谱指标	预处理	训练集			测试集
光谱指标	预处理	R²	RMSE	RPD	R²	RMSE	RPD
KCl	MD	0.838 4	1 672.35	2.281 8	0.879 3	1 562.76	2.500 5
	MD-SG	0.892 5	1 302.89	2.886 5	0.952 1	1 030.15	4.093 8
	MD-SG-LR	0.896 7	1 268.46	2.951 3	0.954 6	1 008.30	4.280 0
	MD-SG-CR	0.969 7	736.83	5.667 9	0.973 5	712.97	6.002 3
	MD-SG-NR	0.887 5	1 502.89	2.646 1	0.853 7	1 338.14	3.966 7
CaCl₂	MD	0.577 6	2 420.83	1.171 5	0.720 9	1 979.70	1.448 8
	MD-SG	0.595 7	2 350.49	1.216 0	0.730 2	1 979.12	1.437 5
	MD-SG-LR	0.604 9	2 372.80	1.239 6	0.721 2	1 912.86	1.462 5
	MD-SG-CR	0.606 6	2 410.59	1.244	0.739 5	1 744.42	1.452 8
	MD-SG-NR	0.688 7	2 136.65	1.490 0	0.670 6	1 993.59	1.515 1
NaCl	MD	0.927 2	895.99	3.587 7	0.962 1	618.69	4.771
	MD-SG	0.929 1	899.47	3.625 4	0.957 3	622.09	4.606 4
	MD-SG-LR	0.938 6	842.98	3.924 4	0.961 9	569.53	4.931 4
	MD-SG-CR	0.860 6	1 325.83	2.517 4	0.786 3	1 020.93	2.028 7
	MD-SG-NR	0.967 6	627.35	5.482 3	0.964 3	509.20	5.264 4

Table 3. Modeling results ofKCl， CaCl₂andNaCl based on BPNN full-band

View table

View in Article

Table 3. Modeling results ofKCl， CaCl₂andNaCl based on BPNN full-band

	光谱指标	预处理	训练集			测试集
	光谱指标	预处理	R²	RMSE	RPD	R²	RMSE	RPD
全波段	KCl	MD	0.982 3	521.37	7.378 9	0.970 9	811.37	5.690 4
		MD-SG	0.996 4	248.77	16.706 6	0.998 8	156.89	29.019 6
		MD-SG-LR	0.995 9	279.00	15.589 2	0.997 4	210.31	19.738 0
		MD-SG-CR	0.996 0	271.04	15.491 5	0.990 1	417.46	9.769 8
		MD-SG-NR	0.996 4	247.23	16.691 0	0.998 5	174.43	26.337 0
	CaCl₂	MD	0.828 0	1 384.35	2.387 5	0.658 7	2 530.33	1.380 5
		MD-SG	0.619 5	2 176.94	1.548 0	0.692 9	2 276.43	1.574 0
		MD-SG-LR	0.750 0	1 899.21	1.922 7	0.662 8	2 053.17	1.528 7
		MD-SG-CR	0.680 4	2 163.12	1.575	0.748 6	1 744.91	1.615 9
		MD-SG-NR	0.984 2	443.65	7.942 1	0.919 9	983.25	3.522 4
	NaCl	MD	0.995	203.81	14.135 6	0.991 5	362.6	10.761 1
		MD-SG	0.996 4	168.33	16.712 7	0.992 9	341.13	11.765 5
		MD-SG-LR	0.993 7	273.1	13.116 6	0.984 9	345.46	8.427 5
		MD-SG-CR	0.992 6	310.45	11.670 6	0.944 8	483.26	4.253 1
		MD-SG-NR	0.983 8	408.04	7.601 4	0.998 2	144.34	23.385 3

Table 4. Modeling results of KCl， CaCl₂and NaCl based on PCA-BPNN

View table

View in Article

Table 4. Modeling results of KCl， CaCl₂and NaCl based on PCA-BPNN

	光谱指标	预处理	训练集			测试集
	光谱指标	预处理	R²	RMSE	RPD	R²	RMSE	RPD
PCA	KCl	MD	0.986 7	479.04	8.663 0	0.990 1	447.99	9.986 9
		MD-SG	0.998 1	181.98	22.775 7	0.988 4	488.07	9.380 0
		MD-SG-LR	0.999 1	130.95	32.665 6	0.999 2	120.86	35.748 9
		MD-SG-CR	0.997 0	232.13	18.246 7	0.997 2	232.26	18.912 9
		MD-SG-NR	0.996 7	237.77	17.407 3	0.997 2	240.07	19.070 9
	CaCl₂	MD	0.696 6	2 051.79	1.518 3	0.797 1	1 687.66	1.797 8
		MD-SG	0.697 2	2 034.13	1.491 3	0.750 2	1 904.11	1.460 9
		MD-SG-LR	0.738 8	1 929.35	1.735 8	0.784 7	1 681.03	1.838 6
		MD-SG-CR	0.758 6	1 888.57	1.666 9	0.781 9	1 596.46	1.560 0
		MD-SG-NR	0.895 6	1 237.50	3.033 5	0.951 1	911.95	4.462 5
	NaCl	MD	0.996 5	196.71	16.832 9	0.996 9	177.37	18.023 7
		MD-SG	0.996 6	187.96	17.219 0	0.996 8	189.19	17.752 0
		MD-SG-LR	0.997 6	173.96	20.485 5	0.992 8	199.81	11.739 9
		MD-SG-CR	0.995 2	249.52	14.568 0	0.973 5	335.07	6.171 9
		MD-SG-NR	0.998 8	114.42	28.658 4	0.998 8	112.92	28.973 5

Table 5. Modeling results of KCl， CaCl₂and NaCl based on PLSR full-band

View table

View in Article

Table 5. Modeling results of KCl， CaCl₂and NaCl based on PLSR full-band

	预处理	训练集			测试集
	预处理	R²	RMSE	RPD	R²	RMSE	RPD
KCl	MD	0.948 1	892.82	4.283 6	0.944 4	1 169.83	3.691 1
	MD-SG	0.992 6	429.99	9.644	0.992 6	391.92	11.641 5
	MD-SG-LR	0.990 6	421.17	10.262 3	0.991 7	385.51	10.946 3
	MD-SG-CR	0.992 2	380.06	11.315 1	0.985 5	510.54	8.161 5
	MD-SG-NR	0.989 3	425.73	9.630 3	0.993 9	367.20	12.638 9
CaCl₂	MD	0.832 9	1 364.64	2.236 5	0.613 2	2 699.46	1.235 1
	MD-SG	0.592 2	2 253.46	1.207 2	0.707 5	2 312.62	1.238 8
	MD-SG-LR	0.567	2 499.60	1.146 3	0.833 2	1 651.02	1.521 5
	MD-SG-CR	0.581 2	2 476.29	1.180 1	0.819 4	1 727.21	1.421 3
	MD-SG-NR	0.973 9	569.48	6.117 8	0.958 8	841.66	4.783 2
NaCl	MD	0.987 9	318.01	9.067 8	0.971 9	660.05	5.911 1
	MD-SG	0.962 8	541.82	5.094 2	0.984 2	511.34	7.752 2
	MD-SG-LR	0.979 9	488.10	6.988 3	0.969 2	495.19	5.678 6
	MD-SG-CR	0.931 3	948.05	3.687 1	0.754 9	1 046.03	1.939 6
	MD-SG-NR	0.965 8	593.37	5.324 2	0.979 2	494.97	6.838 2

Tools

Get Citation

Copy Citation Text

Qing GUO, Lifu ZHANG, Wenchao Qi, Linshan ZHANG. The Hyperspectral Inversion Method of Main Ionic Compounds Content in Groundwater based on BP Neural Network[J]. Remote Sensing Technology and Application, 2024, 39(1): 149

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Research Articles

Received: Jul. 15, 2022

Accepted: --

Published Online: Jul. 22, 2024

The Author Email: GUO Qing (1079695784@qq.com)

DOI:10.11873/j.issn.1004-0323.2024.1.0149

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Modeling results of KCl， CaCl2 andNaCl based on SPA feature spectrum selection

Table 1. Modeling results of KCl， CaCl2 andNaCl based on SPA feature spectrum selection

Table 2. Modeling results of KCl， CaCl2 andNaCl based on PCA characteristic spectral bands

Table 2. Modeling results of KCl， CaCl2 andNaCl based on PCA characteristic spectral bands

Table 3. Modeling results ofKCl， CaCl2 andNaCl based on BPNN full-band

Table 3. Modeling results ofKCl， CaCl2 andNaCl based on BPNN full-band

Table 4. Modeling results of KCl， CaCl2 and NaCl based on PCA-BPNN

Table 4. Modeling results of KCl， CaCl2 and NaCl based on PCA-BPNN

Table 5. Modeling results of KCl， CaCl2 and NaCl based on PLSR full-band

Table 5. Modeling results of KCl， CaCl2 and NaCl based on PLSR full-band

Table 1. Modeling results of KCl， CaCl₂andNaCl based on SPA feature spectrum selection

Table 1. Modeling results of KCl， CaCl₂andNaCl based on SPA feature spectrum selection

Table 2. Modeling results of KCl， CaCl₂andNaCl based on PCA characteristic spectral bands

Table 2. Modeling results of KCl， CaCl₂andNaCl based on PCA characteristic spectral bands

Table 3. Modeling results ofKCl， CaCl₂andNaCl based on BPNN full-band

Table 3. Modeling results ofKCl， CaCl₂andNaCl based on BPNN full-band

Table 4. Modeling results of KCl， CaCl₂and NaCl based on PCA-BPNN

Table 4. Modeling results of KCl， CaCl₂and NaCl based on PCA-BPNN

Table 5. Modeling results of KCl， CaCl₂and NaCl based on PLSR full-band

Table 5. Modeling results of KCl， CaCl₂and NaCl based on PLSR full-band