Improving the SSH Retrieval Precision of Spaceborne GNSS-R Based on a New Grid Search Multihidden Layer Neural Network Feature Optimization Method
The altimetry precision of conventional spaceborne Global Navigation Satellite Systems Reflectometry (GNSS-R) is limited, and the error models are complicated. To compensate for the shortcomings of conventional methods, we present a new grid search multihidden layer neural network feature optimizati...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-07-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/14/13/3161 |
| _version_ | 1827625483461722112 |
|---|---|
| author | Qiang Wang Wei Zheng Fan Wu Huizhong Zhu Aigong Xu Yifan Shen Yelong Zhao |
| author_facet | Qiang Wang Wei Zheng Fan Wu Huizhong Zhu Aigong Xu Yifan Shen Yelong Zhao |
| author_sort | Qiang Wang |
| collection | DOAJ |
| description | The altimetry precision of conventional spaceborne Global Navigation Satellite Systems Reflectometry (GNSS-R) is limited, and the error models are complicated. To compensate for the shortcomings of conventional methods, we present a new grid search multihidden layer neural network feature optimization method (GSMHLFO) for sea surface height (<i>SSH</i>) retrieval. Firstly, the GSMHLFO is constructed by combining the multihidden layer neural network, feature engineering, and a grid search algorithm. Moreover, the retrieval performance of the GSMHLFO and its sensitivity to various features are analyzed. By analyzing 14 feature sets with different information details, we concluded that the elevation, signal-to-noise ratio (<i>SNR</i>), atmospheric delay, and ocean wind speed can provide essential contributions to the <i>SSH</i> retrieval based on GSMHLFO. Secondly, the Technical University of Denmark 18 mean sea surface (DTU18 MSS), which is corrected by the TPXO8 global tide model, was used to verify the GSMHLFO. The number of hidden layers and neurons was optimized using the grid search algorithm. The experimental results show that the proposed GSMHLFO with four hidden layers and 200 neurons per layer has a better retrieval performance. Compared with DTU18, the mean absolute difference (<i>MAD</i>), the root mean square error (<i>RMSE</i>), and the Pearson correlation coefficient (<i>PCC</i>) equal 4.23 m, 5.94 m, and 0.98, respectively. The retrieval precision obtained is significantly improved compared to that reported in the literature for the TDS-1 <i>SSH</i> retrieval. Finally, the retrieval performance of the GSMHLFO and the traditional HALF single-point retracking method were compared. The precision of GSMHLFO is higher than that of traditional retracking method according to <i>MAD</i>, <i>RMSE</i>, and <i>PCC,</i> which are increased by 32.86, 25.00, and 8.99%. The GSMHLFO will provide innovative theoretical and methodological support for the high-precision <i>SSH</i> retrieval of GNSS-R altimetry satellites in the future. |
| first_indexed | 2024-03-09T12:37:19Z |
| format | Article |
| id | doaj.art-d81086dfcc174321833097a3aff2766e |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-09T12:37:19Z |
| publishDate | 2022-07-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-d81086dfcc174321833097a3aff2766e2023-11-30T22:23:29ZengMDPI AGRemote Sensing2072-42922022-07-011413316110.3390/rs14133161Improving the SSH Retrieval Precision of Spaceborne GNSS-R Based on a New Grid Search Multihidden Layer Neural Network Feature Optimization MethodQiang Wang0Wei Zheng1Fan Wu2Huizhong Zhu3Aigong Xu4Yifan Shen5Yelong Zhao6School of Geomatics, Liaoning Technical University, Fuxin 123000, ChinaSchool of Geomatics, Liaoning Technical University, Fuxin 123000, ChinaQian Xuesen Laboratory of Space Technology, China Academy of Space Technology, Beijing 100094, ChinaSchool of Geomatics, Liaoning Technical University, Fuxin 123000, ChinaSchool of Geomatics, Liaoning Technical University, Fuxin 123000, ChinaSchool of Geomatics, Liaoning Technical University, Fuxin 123000, ChinaShenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen 518000, ChinaThe altimetry precision of conventional spaceborne Global Navigation Satellite Systems Reflectometry (GNSS-R) is limited, and the error models are complicated. To compensate for the shortcomings of conventional methods, we present a new grid search multihidden layer neural network feature optimization method (GSMHLFO) for sea surface height (<i>SSH</i>) retrieval. Firstly, the GSMHLFO is constructed by combining the multihidden layer neural network, feature engineering, and a grid search algorithm. Moreover, the retrieval performance of the GSMHLFO and its sensitivity to various features are analyzed. By analyzing 14 feature sets with different information details, we concluded that the elevation, signal-to-noise ratio (<i>SNR</i>), atmospheric delay, and ocean wind speed can provide essential contributions to the <i>SSH</i> retrieval based on GSMHLFO. Secondly, the Technical University of Denmark 18 mean sea surface (DTU18 MSS), which is corrected by the TPXO8 global tide model, was used to verify the GSMHLFO. The number of hidden layers and neurons was optimized using the grid search algorithm. The experimental results show that the proposed GSMHLFO with four hidden layers and 200 neurons per layer has a better retrieval performance. Compared with DTU18, the mean absolute difference (<i>MAD</i>), the root mean square error (<i>RMSE</i>), and the Pearson correlation coefficient (<i>PCC</i>) equal 4.23 m, 5.94 m, and 0.98, respectively. The retrieval precision obtained is significantly improved compared to that reported in the literature for the TDS-1 <i>SSH</i> retrieval. Finally, the retrieval performance of the GSMHLFO and the traditional HALF single-point retracking method were compared. The precision of GSMHLFO is higher than that of traditional retracking method according to <i>MAD</i>, <i>RMSE</i>, and <i>PCC,</i> which are increased by 32.86, 25.00, and 8.99%. The GSMHLFO will provide innovative theoretical and methodological support for the high-precision <i>SSH</i> retrieval of GNSS-R altimetry satellites in the future.https://www.mdpi.com/2072-4292/14/13/3161GSMHLFOfeature engineeringTDS-1integral delay waveformDTU18 MSS |
| spellingShingle | Qiang Wang Wei Zheng Fan Wu Huizhong Zhu Aigong Xu Yifan Shen Yelong Zhao Improving the SSH Retrieval Precision of Spaceborne GNSS-R Based on a New Grid Search Multihidden Layer Neural Network Feature Optimization Method Remote Sensing GSMHLFO feature engineering TDS-1 integral delay waveform DTU18 MSS |
| title | Improving the SSH Retrieval Precision of Spaceborne GNSS-R Based on a New Grid Search Multihidden Layer Neural Network Feature Optimization Method |
| title_full | Improving the SSH Retrieval Precision of Spaceborne GNSS-R Based on a New Grid Search Multihidden Layer Neural Network Feature Optimization Method |
| title_fullStr | Improving the SSH Retrieval Precision of Spaceborne GNSS-R Based on a New Grid Search Multihidden Layer Neural Network Feature Optimization Method |
| title_full_unstemmed | Improving the SSH Retrieval Precision of Spaceborne GNSS-R Based on a New Grid Search Multihidden Layer Neural Network Feature Optimization Method |
| title_short | Improving the SSH Retrieval Precision of Spaceborne GNSS-R Based on a New Grid Search Multihidden Layer Neural Network Feature Optimization Method |
| title_sort | improving the ssh retrieval precision of spaceborne gnss r based on a new grid search multihidden layer neural network feature optimization method |
| topic | GSMHLFO feature engineering TDS-1 integral delay waveform DTU18 MSS |
| url | https://www.mdpi.com/2072-4292/14/13/3161 |
| work_keys_str_mv | AT qiangwang improvingthesshretrievalprecisionofspacebornegnssrbasedonanewgridsearchmultihiddenlayerneuralnetworkfeatureoptimizationmethod AT weizheng improvingthesshretrievalprecisionofspacebornegnssrbasedonanewgridsearchmultihiddenlayerneuralnetworkfeatureoptimizationmethod AT fanwu improvingthesshretrievalprecisionofspacebornegnssrbasedonanewgridsearchmultihiddenlayerneuralnetworkfeatureoptimizationmethod AT huizhongzhu improvingthesshretrievalprecisionofspacebornegnssrbasedonanewgridsearchmultihiddenlayerneuralnetworkfeatureoptimizationmethod AT aigongxu improvingthesshretrievalprecisionofspacebornegnssrbasedonanewgridsearchmultihiddenlayerneuralnetworkfeatureoptimizationmethod AT yifanshen improvingthesshretrievalprecisionofspacebornegnssrbasedonanewgridsearchmultihiddenlayerneuralnetworkfeatureoptimizationmethod AT yelongzhao improvingthesshretrievalprecisionofspacebornegnssrbasedonanewgridsearchmultihiddenlayerneuralnetworkfeatureoptimizationmethod |