基于半监督光学分类的白洋淀叶绿素a浓度反演方法

doi:10.3964/j.issn.1000-0593(2025)07-2068-10

摘要
参考文献
相关文章 (15)

全文: PDF (12834 KB)
输出: BibTeX | EndNote (RIS)

摘要：利用遥感技术反演叶绿素a浓度(Chlorophyll-a，Chl-a)被认为是实现水体富营养化监测的有效途径。由于内陆水体光学性质复杂且存在较大时空分异性，单一的反演模型难以准确反演Chl-a浓度，基于水体光学分类结果进行针对性建模反演是内陆水体Chl-a反演的重要技术方法之一。以白洋淀为研究区，以实测反射光谱和Chl-a浓度为数据源，提出一种基于半监督光学分类的反演方法。首先，为确保分类后各类别样本数量足以支持反演模型的构建，按照7∶1的比例对样本进行划分建模集和验证集。通过多种光谱指数及摩尔投票算法选取出白洋淀具有代表性的标记样本。其次，选取模糊C均值聚类算法和随机森林算法构建半监督分类器，基于获取的代表性标记样本，进一步挖掘未标记样本中的潜在信息，提高未标记样本的分类精度。最后，针对各水体类型分别建立反演模型，计算各水体类型的质心光谱，利用距离权重进行混合加权获取最终反演结果。结果表明，白洋淀水体可根据光谱特征分为浮游植物主导型、中间型和悬浮物主导型，不同类型水体具有较为明显的光学性质差异，根据光学分类结果可针对性选择最优反演模型，提升反演精度。与传统的光学分类策略相比，本方法表现最佳，其平均相对误差、均方根误差和平均绝对误差均有所下降，能更加准确的反演Chl-a浓度(MRE=0.10，RMSE=0.126 μg·L^-1，MAE=0.106 μg·L^-1)。此外，应用ZY01-02E影像数据进行Chl-a反演，能够有效揭示Chl-a浓度的空间分布状况。该方法展现出在水体富营养化监测中的应用潜力，也为内陆水体Chl-a浓度的遥感反演提供了新思路。

关键词：半监督分类；水体光学类型；叶绿素a；白洋淀

Abstract：The estimation of Chlorophyll-a concentration (Chl-a) using remote sensing technology is considered an effective way to monitor eutrophication in water bodies. Due to the complexity of the optical properties of inland water bodies and the existence of large spatial and temporal variability, it is difficult for a single estimation model to accurately estimate Chl-a concentration, and targeted modelling estimation based on the results of the optical classification of water bodies is one of the most important technological approaches for inland water body Chl-a inversion. In this study, Baiyangdian is taken as the study area, and a semi-supervised optical classification-based estimation method is proposed using the measured reflectance spectra and Chl-a concentration as the data source. First, to ensure that the number of samples in each category after classification is sufficient to support the construction of the estimation model, this study divides the samples into modelling set and validation set according to the ratio of 7∶1. Representative labelled samples of Baiyangdian were selected by multiple spectral indices and Moore's voting algorithm. Secondly, the fuzzy C-mean clustering algorithm and random forest algorithm are selected to construct a semi-supervised classifier, based on the representative labelled samples obtained, to further explore the potential information in the unlabelled samples and improve the classification accuracy of the unlabelled samples. Finally, the estimation models were established for each water type, the centre-of-mass spectra of each water type were calculated, and the final estimation results were obtained by hybrid-weighting using distance weights. The results show that Baiyangdian water bodies can be classified into phytoplankton-dominated, intermediate and suspended matter-dominated according to the spectral characteristics, and different types of water bodies have obvious differences in optical properties, which can be used to select the optimal estimation model and improve the estimation accuracy according to the optical classification results. Compared with the traditional optical classification strategy, the method proposed in this study performed the best, with a decrease in the mean relative error, root mean square error and mean absolute error, and was able to estimate Chl-a concentration more accurately (MRE=0.10, RMSE=0.126 μg·L^-1, MAE=0.106 μg·L^-1). In addition, applying ZY01-02E image data for Chl-a estimation in this study can effectively reveal the spatial distribution of Chl-a concentration. This method demonstrates the potential for application in eutrophication monitoring of water bodies, and also provides a new idea for remote sensing estimation of Chl-a concentration in inland water bodies.

Key words：Semi-supervised classification;Optical type of water;Chlorophyll-a;Baiyangdian

收稿日期: 2024-10-12 修订日期: 2025-02-10

中图分类号:

TP79

基金资助: 高分辨率对地观测系统国家科技重大专项（67-Y50G05-9001-22/23），河北省中央引导地方科技发展资金项目(246Z7602G)，河北省高等学校科学技术研究项目(CXY2023011)资助

通讯作者: 赵起超 E-mail: rs_zhao@nciae.edu.cn

作者简介: 陈雯玥，女，2000年生，北华航天工业学院遥感信息工程学院硕士研究生 e-mail: cwy0102@stumail.nciae.edu.cn

引用本文:

陈雯玥，赵起超，杨秀峰，韩宝辉，张雨晴. 基于半监督光学分类的白洋淀叶绿素a浓度反演方法[J]. 光谱学与光谱分析, 2025, 45(07): 2068-2077.
CHEN Wen-yue, ZHAO Qi-chao, YANG Xiu-feng, HAN Bao-hui, ZHANG Yu-qing. Estimation Approach of Chlorophyll-a Concentration in Baiyangdian Based on Semi-Supervised Optical Classification. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(07): 2068-2077.

链接本文:

https://www.gpxygpfx.com/CN/10.3964/j.issn.1000-0593(2025)07-2068-10 或 https://www.gpxygpfx.com/CN/Y2025/V45/I07/2068

[1] LI Na，LI Jia-xi，LI Guo-wen，et al(李娜，黎佳茜，李国文，等). Acta Hydrobiologica Sinica(水生生物学报)，2018，42(4)：854.
[2] BAO Ying，TIAN Qing-jiu，CHEN Min，et al(包颖，田庆久，陈旻，等). Spectroscopy and Spectral Analysis(光谱学与光谱分析)，2016，36(8)：2562.
[3] Cui T W，Zhang J，Wang K，et al. ISPRS Journal of Photogrammetry and Remote Sensing，2020，163: 187.
[4] LUO Jie-chun-yi，QIN Long-jun，MAO Peng，et al(罗婕纯一，秦龙君，毛鹏，等). Remote Sensing Technology and Application(遥感技术与应用)，2021，36(3)：473.
[5] KUANG Run-yuan，ZHAO Yan-fu，LUO Wei，et al(况润元，赵艳福，罗卫，等). Journal of China Hydrology(水文)，2017，37(6)：23.
[6] LI Yun-mei，ZHAO Huan，BI Shun，et al(李云梅，赵焕，毕顺，等). National Remote Sensing Bulletin(遥感学报)，2022，26(1)：19.
[7] FENG Chi，JIN Qi，WANG Yan-nan，et al(冯驰，金琦，王艳楠，等). Environmental Science(环境科学)，2015，36(5)：1557.
[8] SONG Zi-hao，KUANG Run-yuan(宋子豪，况润元). Jiangxi Hydraulic Science & Technology(江西水利科技)，2021，47(1)：70.
[9] Zhang F，Li J，Shen Q，et al. International Journal of Applied Earth Observation and Geoinformation，2019，74: 138.
[10] Shi K，Li Y，Li L，et al. Science of the Total Environment，2013，444: 1.
[11] ZHANG Fang-fang，LI Jun-sheng，WANG Chao，et al(张方方，李俊生，王超，等). National Remote Sensing Bulletin(遥感学报)，2023，27(3)：769.
[12] Dang X，Du J，Wang C，et al. Remote Sensing，2023，15(8): 2209.
[13] Jiang G，Loiselle S A，Yang D，et al. Remote Sensing of Environment，2020，241：111735.
[14] Le C，Li Y，Zha Y，et al. Remote Sensing of Environment: An Interdisciplinary Journal，2011，115(2)：725.
[15] Sun D，Hu C，Qiu Z，et al. Remote Sensing of Environment，2014，155: 289.
[16] HAN Bao-hui，ZHAO Qi-chao，CHANG Rong，et al(韩宝辉，赵起超，常荣，等). Journal of Geo-Information Science(地球信息科学学报)，2023，25(9)：1882.
[17] Lichtenthaler H K, Buschmann C. Current Protocols in Food Analytical Chemistry，John Wiley & Sons, Inc., 2001.
[18] Gan H，Sang N，Huang R，et al. Neurocomputing，2013，101: 290.
[19] KUANG Run-yuan，LUO Wei，ZHANG Meng(况润元，罗卫，张萌). Resources and Environment in the Yangtze Basin(长江流域资源与环境)，2015，24(5)：773.