Remote Sensing Inversion of Total Nitrogen in Hyperspectral Water Bodies Based on Random Frog Dimensionality Reduction and PSO-BPNN

Ziyue Zhang; Xiao Liu; Lili Du; Shun Yao; Xiaobing Sun; Wei Xiong; Di Cao

doi:10.3788/LOP250625

Laser & Optoelectronics Progress, Volume. 62, Issue 17, 1730001(2025)

Remote Sensing Inversion of Total Nitrogen in Hyperspectral Water Bodies Based on Random Frog Dimensionality Reduction and PSO-BPNN

Ziyue Zhang^1,2, Xiao Liu^2、**, Lili Du^2、*, Shun Yao², Xiaobing Sun², Wei Xiong^1,2, and Di Cao³

Author Affiliations

¹College of Environmental Science and Optoelectronic Technology, University of Science and Technology of China, Hefei 230026, Anhui , China

²Key Laboratory of Optical Calibration and Characterization, Anhui Institute of Optics and Fine Mechanics, Hefei Institutes of Physical Science, Chinese Academy of Sciences, Hefei 230031, Anhui , China

³Anhui Hefei Ecological Environment Monitoring Center, Hefei 230071, Anhui , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(19)

Fig. 1. Distribution of sampling points

Download full size

Fig. 2. Distribution of mass concentration for total nitrogen

Download full size

Fig. 3. Spectral reflectance curves. (a) Raw spectral reflectance curves; (b) SG-treated spectral reflectance curves

Download full size

Fig. 4. Hyperspectral camera onboard the unmanned aerial vehicle and its route planning. (a) Hyperspectral camera onboard the unmanned aerial vehicle; (b) route planning

Download full size

Fig. 5. Preprocessing flowchart of unmanned aerial vehicle hyperspectral image

Download full size

Fig. 6. Unmanned aerial vehicle hyperspectral image and spectra. (a) Pre-processed image; (b) spectral curve of vegetation; (c) spectral curve of water body

Download full size

Fig. 7. Flowchart of the RF dimensionality reduction algorithm

Download full size

Fig. 8. Flowchart of PSO-BPNN

Download full size

Fig. 9. RF dimensionality reduction result

Download full size

Fig. 10. Training results of different models. (a) PSO-BPNN; (b) BPNN; (c) PLS

Download full size

Fig. 11. Comparison of mass concentration for total nitrogen between the inverted and measured results by different models. (a) PSO-BPNN; (b) BPNN; (c) PLS

Download full size

Fig. 12. Prediction results of the test set by different models. (a) PSO-BPNN; (b) BPNN; (c) PLS

Download full size

Fig. 13. Comparison of mass concentrations for total nitrogen between predicted and measured results by different models. (a) PSO-BPNN; (b) BPNN; (c) PLS

Download full size

Fig. 14. Spatial distribution of mass concentration for total nitrogen

Download full size

Table 1. Dataset segmentation
View table
Table 1. Dataset segmentation
Dataset Number of sample Mass concentration /（mg∙L^-1）
Maximum Minimum Mean Standard deviation
Total set 71 5.8610 1.2948 4.0750 1.1084
Training set 53 5.8610 1.2948 4.0679 1.0834
Test set 18 5.6955 1.3975 3.9291 1.1712

Table 2. Parameter settings for hyperspectral camera onboard the unmanned aerial vehicle

View table

Table 2. Parameter settings for hyperspectral camera onboard the unmanned aerial vehicle

Parameter	Value
Flight altitude /m	200
Frame rate /（frame·s^-1）	200
Lens focal length /mm	25
Number of spectral channel	128
Spectral resolution /nm	3.5
Spatial resolution /m	0.014
Line field of view width /m	31.1
Spectral range /nm	350‒1000

Table 3. Evaluation indicators of the training set by different models
View table
Table 3. Evaluation indicators of the training set by different models
Model R² MAE /（mg∙L^-1） MAPE /% RMSE /（mg∙L^-1）
PSO-BPNN 0.862 0.336 8.2 0.405
BPNN 0.827 0.341 8.4 0.451
PLS 0.609 0.534 16.0 0.677

Table 4. Evaluation indicators of the test set by different models
View table
Table 4. Evaluation indicators of the test set by different models
Model R² MAE /（mg∙L^-1） MAPE /% RMSE /（mg∙L^-1）
PSO-BPNN 0.711 0.520 14.0 0.640
BPNN 0.603 0.544 16.8 0.738
PLS 0.567 0.642 19.4 0.770

Table 5. Comparison between inverted and measured values for six verification points

View table

Table 5. Comparison between inverted and measured values for six verification points

Sampling point	Longitude /（°）	Latitude /（°）	Mass concentration /（mg∙L^-1）		Relative error /%
Sampling point	Longitude /（°）	Latitude /（°）	Inverted value	Measured value	Relative error /%
1	117.36555	31.74683	5.82	5.53	5.24
2	117.36576	31.74710	5.61	5.62	0.18
3	117.36606	31.34763	5.51	5.36	2.80
4	117.36649	31.74775	5.88	5.84	0.68
5	117.36636	31.74757	4.95	5.22	5.17
6	117.36548	31.74668	3.26	3.44	5.23

Tools

Get Citation

Copy Citation Text

Ziyue Zhang, Xiao Liu, Lili Du, Shun Yao, Xiaobing Sun, Wei Xiong, Di Cao. Remote Sensing Inversion of Total Nitrogen in Hyperspectral Water Bodies Based on Random Frog Dimensionality Reduction and PSO-BPNN[J]. Laser & Optoelectronics Progress, 2025, 62(17): 1730001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites