Inversion of Suspended Particulate Matter Concentration in Maozhou River Based on Band Selection of Hyperspectral Data

Zhongkai Chen; Xiaorun Li; Liaoying Zhao

doi:10.3788/LOP202158.2001001

Laser & Optoelectronics Progress, Volume. 58, Issue 20, 2001001(2021)

Inversion of Suspended Particulate Matter Concentration in Maozhou River Based on Band Selection of Hyperspectral Data

Zhongkai Chen¹, Xiaorun Li^1、*, and Liaoying Zhao²

Author Affiliations

¹College of Electrical Engineering, Zhejiang University, Hangzhou, Zhejiang 310027, China

²Institute of Computer Application Technology, Hangzhou Dianzi University, Hangzhou, Zhejiang 310018, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(10)

Fig. 1. Structural diagram of multilayer perceptron with 2 hidden layers

Download full size

Fig. 2. Correlation matrices of remote sensing reflectance data. (a) March 11, 2019; (b) August 5, 2019

Download full size

Fig. 3. Neural network structure in PNN algorithm

Download full size

Fig. 4. Sum of entropy of different algorithms when M=12. (a) March 11, 2019; (b) August 5, 2019

Download full size

Table 1. Statistical results of SPM concentration mg·L^-1
View table
Table 1. Statistical results of SPM concentration mg·L^-1
Date C_min C_max C_mean C_std
2019-03-11 4.00 18.00 6.89 8.10
2019-08-05 4.00 14.00 8.00 10.32

Table 2. Statistical results of hyperspectral data measured by handheld spectrometersnm
View table
Table 2. Statistical results of hyperspectral data measured by handheld spectrometersnm
Date Spectral resolution Minimum wavelength Maximum wavelength
2019-03-11 1.98 399.87 997.58
2019-08-05 1.00 325.00 1075.00

Table 3. Experimental result comparison of SPM concentration inversion using neural network (March 11, 2019)

View table

Table 3. Experimental result comparison of SPM concentration inversion using neural network (March 11, 2019)

Number of bands	MSE					MAE
Number of bands	DTBS	SVM-RFE	PNN-L1	PNN-L2	PNN-ReLU	DTBS	SVM-RFE	PNN-L1	PNN-L2	PNN-ReLU
6	20.13	11.69	6.75	7.10	5.58	2.78	2.10	1.43	1.04	1.15
9	12.77	11.53	4.77	8.40	5.81	2.03	1.46	1.10	1.27	1.14
12	6.26	12.78	5.58	6.31	5.89	1.24	2.22	1.17	1.23	1.18
15	12.99	6.95	9.04	5.11	6.40	2.03	1.60	1.30	1.23	1.18
18	16.26	16.96	4.35	4.41	7.00	2.42	2.05	0.95	1.02	1.02
21	19.32	8.85	3.32	8.53	5.09	2.88	1.54	0.80	1.38	1.21
24	10.11	5.12	10.88	4.04	5.19	1.81	1.08	1.64	0.85	1.19
27	13.18	20.84	4.65	5.79	7.62	2.04	2.74	1.00	1.03	1.13
30	11.91	13.72	5.81	5.16	7.42	1.90	2.39	1.27	0.93	1.39
33	11.74	17.73	7.11	3.88	4.83	1.93	1.99	1.47	0.79	1.04
36	12.84	17.02	7.33	3.98	5.81	1.89	2.10	1.34	0.91	1.22
39	12.54	18.99	3.40	3.55	6.42	2.08	2.25	1.06	0.81	1.34
42	10.79	10.80	8.16	4.69	5.58	1.87	1.72	1.31	1.15	1.13
45	14.83	15.41	4.80	5.19	7.85	2.04	2.57	1.05	1.06	1.28
48	7.60	15.67	8.92	4.23	4.95	1.44	2.42	1.82	0.92	1.16

Table 4. Experimental result comparison of SPM concentration inversion using neural network model (August 5, 2019)

View table

Table 4. Experimental result comparison of SPM concentration inversion using neural network model (August 5, 2019)

Number of bands	MSE					MAE
Number of bands	DTBS	SVM-RFE	PNN-L1	PNN-L2	PNN-ReLU	DTBS	SVM-RFE	PNN-L1	PNN-L2	PNN-ReLU
6	13.24	30.33	3.92	6.62	5.99	1.91	3.48	1.25	1.40	1.47
9	15.70	29.17	8.28	5.22	5.27	2.17	3.47	1.69	1.38	1.31
12	11.07	11.83	5.71	7.34	9.55	1.61	2.12	1.40	1.38	1.70
15	25.21	14.88	3.18	8.10	7.16	3.02	1.83	1.00	1.58	1.44
18	7.94	20.71	3.26	8.75	7.20	1.78	2.84	1.01	1.40	1.76
21	20.94	17.03	8.85	8.54	7.27	2.70	2.49	1.84	1.88	1.51
24	12.06	7.65	4.08	8.30	6.51	1.79	1.67	1.18	1.64	1.41
27	16.35	14.83	5.72	6.29	4.53	2.29	2.32	1.45	1.42	0.97
30	18.49	15.19	3.08	5.19	13.17	2.46	2.46	1.12	1.22	2.27
33	18.72	17.89	6.15	7.92	8.61	2.55	2.34	1.28	1.53	1.57
36	3.73	17.48	7.45	7.76	3.60	1.06	2.54	1.64	1.66	1.10
39	16.48	12.57	7.23	5.92	6.59	2.39	1.81	1.45	1.41	1.28
42	15.60	40.47	4.83	7.54	4.31	2.15	4.21	1.35	1.68	1.14
45	14.53	20.30	4.02	6.90	5.91	2.36	2.7	1.07	1.50	1.53
48	20.01	15.78	10.00	5.91	6.80	2.56	2.52	1.77	1.43	1.28

Table 5. Experimental result comparison of SPM concentration inversion using random forest model (March 11, 2019)

View table

Table 5. Experimental result comparison of SPM concentration inversion using random forest model (March 11, 2019)

Number of bands	MSE					MAE
Number of bands	DTBS	SVM-RFE	PNN-L1	PNN-L2	PNN-ReLU	DTBS	SVM-RFE	PNN-L1	PNN-L2	PNN-ReLU
6	11.36	10.89	12.89	12.58	10.61	2.40	2.49	2.30	2.36	2.32
9	11.45	12.05	10.54	12.20	11.50	2.29	2.63	2.27	2.27	2.59
12	10.83	11.36	13.63	12.64	12.36	2.17	2.51	2.34	2.33	2.54
15	12.20	11.21	11.34	11.73	13.78	2.37	2.47	2.74	2.27	2.56
18	12.20	11.92	12.19	12.17	12.51	2.35	2.59	2.38	2.46	2.61
21	11.06	11.38	11.46	11.73	12.90	2.35	2.57	2.43	2.33	2.77
24	10.45	11.55	14.87	10.47	16.36	2.24	2.57	2.28	2.31	3.06
27	11.54	11.56	11.60	11.54	13.77	2.45	2.40	2.49	2.45	2.76
30	11.59	10.49	10.76	16.75	11.41	2.25	2.44	2.39	2.88	2.52
33	11.34	10.85	13.36	12.73	10.47	2.29	2.46	2.40	2.59	2.25
36	12.09	11.78	13.73	11.77	11.99	2.34	2.56	2.48	2.42	2.42
39	10.75	10.76	12.46	12.22	11.18	2.34	2.44	2.62	2.44	2.51
42	11.40	12.73	13.64	13.06	12.79	2.36	2.64	2.41	2.62	2.69
45	10.98	11.22	11.62	11.30	10.92	2.29	2.51	2.93	2.32	2.37
48	12.77	11.33	11.41	11.06	11.71	2.41	2.41	2.36	2.42	2.41

Table 6. Experimental result comparison of SPM concentration inversion using random forest model (August 5, 2019)

View table

Table 6. Experimental result comparison of SPM concentration inversion using random forest model (August 5, 2019)

Number of bands	MSE					MAE
Number of bands	DTBS	SVM-RFE	PNN-L1	PNN-L2	PNN-ReLU	DTBS	SVM-RFE	PNN-L1	PNN-L2	PNN-ReLU
6	8.91	8.45	9.46	8.42	10.11	2.40	2.21	2.37	2.18	2.45
9	9.05	8.89	10.11	8.15	8.37	2.29	2.32	2.40	2.19	2.29
12	8.46	8.97	8.68	8.69	8.56	2.17	2.30	2.30	2.24	2.25
15	8.20	8.38	8.85	8.69	7.99	2.37	2.29	2.18	2.19	2.30
18	8.08	8.84	9.72	8.09	8.70	2.35	2.28	2.43	2.20	2.45
21	8.63	8.80	8.58	8.42	7.34	2.35	2.36	2.19	2.19	2.19
24	8.11	8.05	7.80	9.69	7.95	2.24	2.27	2.08	2.45	2.23
27	8.03	8.43	9.22	8.33	8.01	2.45	2.20	2.29	2.14	2.23
30	8.46	8.59	8.81	8.23	9.36	2.25	2.28	2.33	2.17	2.59
33	8.54	8.95	10.48	7.64	9.60	2.29	2.30	2.47	2.06	2.54
36	9.36	8.18	8.15	8.78	7.73	2.34	2.24	2.17	2.30	2.14
39	9.01	8.94	12.37	9.03	6.95	2.34	2.26	2.69	2.29	2.04
42	8.42	8.63	8.77	7.01	7.03	2.36	2.28	2.30	2.03	2.18
45	8.15	8.43	11.38	7.98	7.18	2.29	2.22	2.67	2.12	2.14
48	7.80	8.18	8.36	7.48	7.19	2.41	2.20	2.28	2.00	2.09

Tools

Get Citation

Copy Citation Text

Zhongkai Chen, Xiaorun Li, Liaoying Zhao. Inversion of Suspended Particulate Matter Concentration in Maozhou River Based on Band Selection of Hyperspectral Data[J]. Laser & Optoelectronics Progress, 2021, 58(20): 2001001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Atmospheric Optics and Oceanic Optics

Received: Oct. 26, 2020

Accepted: Jan. 7, 2021

Published Online: Oct. 12, 2021

The Author Email: Li Xiaorun (lxr@zju.edu.cn)

DOI:10.3788/LOP202158.2001001

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Statistical results of SPM concentration mg·L-1

Table 1. Statistical results of SPM concentration mg·L-1

Table 2. Statistical results of hyperspectral data measured by handheld spectrometersnm

Table 2. Statistical results of hyperspectral data measured by handheld spectrometersnm

Table 3. Experimental result comparison of SPM concentration inversion using neural network (March 11, 2019)

Table 3. Experimental result comparison of SPM concentration inversion using neural network (March 11, 2019)

Table 4. Experimental result comparison of SPM concentration inversion using neural network model (August 5, 2019)

Table 4. Experimental result comparison of SPM concentration inversion using neural network model (August 5, 2019)

Table 5. Experimental result comparison of SPM concentration inversion using random forest model (March 11, 2019)

Table 5. Experimental result comparison of SPM concentration inversion using random forest model (March 11, 2019)

Table 6. Experimental result comparison of SPM concentration inversion using random forest model (August 5, 2019)

Table 6. Experimental result comparison of SPM concentration inversion using random forest model (August 5, 2019)

Table 1. Statistical results of SPM concentration mg·L^-1

Table 1. Statistical results of SPM concentration mg·L^-1