Application of Machine Learning in Remote Sensing Retrieval of Typical Water Quality Parameters in Aquatic Environments
Remote sensing technology is a highly effective method for long-term monitoring of large-scale water bodies.The application of machine learning techniques in remote sensing retrieval of typical water quality parameters is reviewed in this study.Initially,the principles,advantages,and disadvantages of several commonly used machine learning algorithms for water quality retrieval are briefly described.Subsequently,the research progress of machine learning models in estimating several key parameters,including chlorophyll-a,suspended matter,dissolved organic matter,phosphorus and nitrogen,is discussed.Additionally,the challenges and issues encountered in this field are analyzed.Based on this review,the following conclusions and prospects are proposed.(1)Machine learning models generally outperform traditional empirical and semi-empirical models in terms of retrieval accuracy.(2)Models capable of quantifying retrieval uncertainty,such as mixture density networks and Bayesian probabilistic neural networks,offer more comprehensive and reliable retrieval.(3)Machine learning models developed from extensive global datasets demonstrate good generalization capabilities and hold potential for productization.(4)Future research should focus on popularizing uncertainty estimation algorithms and transfer learning,evaluating atmospheric correction algorithms,and developing big data applications for remote sensing of aquatic environments.
国家科技期刊平台
NSTL
万方数据
