西北工程技术学报2025，Vol.24Issue(3)：247-256,10.
基于深度学习的被动微波遥感土壤水分反演
Soil Moisture Retrieval From Passive Microwave Remote Sensing Based on Deep Learning
摘要
Abstract
To enhance the accuracy of soil moisture retrieval from passive microwave remote sensing data in areas with complex terrain or dense vegetation,an artificial intelligence-based optimization method is proposed.The study area encompasses southern North America(13°-50° N,65°-125° W).Specifically,a fully connected neural network(FCNN)was employed to pre-train AMSR2 brightness temperature data and AMSR2 soil moisture values.An iterative data optimization method inspired by adversarial neural networks was used to enhance model performance by progressively adjusting target values based on feature information from the neural network's hidden layers.Experimental results demonstrate that over the region which exhibits notable heterogeneity in surface cover and meteorological conditions,the optimized model significantly improves all accuracy metrics,with the best-performing model achieving a mean absolute error(MAE)of 0.023 m³/m³,a root mean square error(RMSE)of 0.028 m³/m³,and a correlation coefficient(R)of 0.944.This study offers a novel approach for the application of remote-sensing retrieval techniques in agriculture,water-resource management and climate-change research.关键词
被动微波遥感/深度学习/AMSR2/土壤水分/数据优化Key words
passive microwave remote sensing/deep learning/AMSR2/soil moisture/data optimization分类
农业科技引用本文复制引用
李杰,毛克彪,袁紫晋,李春树,郭中华..基于深度学习的被动微波遥感土壤水分反演[J].西北工程技术学报,2025,24(3):247-256,10.基金项目
宁夏科技厅自然科学基金重点项目(2024AC02032) （2024AC02032）
国家重点研发计划项目(2023YFB3906202-4) （2023YFB3906202-4）
