학술논문
结合经验小波变换与BP神经网络的GNSS-IR土壤湿度反演 / GNSS-IR soil moisture inversion via combining empirical wavelet transform and BP neural network
Document Type
Academic Journal
Author
Source
中国科技论文 / China Sciencepaper. 19(1):92-98
Subject
Language
Chinese
ISSN
2095-2783
Abstract
全球导航卫星系统(global navigation satellite system,GNSS)接收机接收到的多路径信号可用于测量土壤湿度.针对目前全球卫星导航系统多径干涉遥感技术(global navigation satellite system interferometry and reflectometry,GNSS-IR)在土壤湿度反演领域精度较低的问题,实验利用美国板块边界观测(plate boundary observation,PBO)网的P040测站作为研究对象,选取该测站的2颗卫星PRN10和PRN14分别进行信噪比数据分析,将L2波段反射信号的延迟相位作为输入、PBO中 H2O的土壤湿度值作为输出,建立基于经验小波变换神经网络(empirical wavelet transform-back propagation,EWT-BP)模型的土壤湿度反演模型,并对提出的土壤湿度反演方法进行精度评定.结果表明,EWT-BP模型反演的土壤湿度与实际土壤湿度之间具有较强相关性,其决定系数(R2)分别为0.820和0.844,相比传统低阶多项式拟合法,R2分别提高了21.66%和28.66%,该结果验证了EWT-BP模型能够有效提升GNSS-IR土壤湿度反演的精确度和可靠性.
The multipath signals received by global navigation satellite system(GNSS)receivers can be used to measure soil mois-ture.In order to solve the current problem of low accuracy of GNSS multipath interferometric remote sensing(GNSS-IR)in the field of soil moisture inversion,the P040 station of the U.S.plate boundary observation(PBO)network was employed as the research object herein,and the two satellites of this station,PRN10 and PRN14 were selected for the analysis of the signal-to-noise ratio data respectively.Moreover,the delayed phases of the reflected signals of the L2-band were taken as the input and the soil moisture value of H2O in PBO was taken as output.A soil moisture inversion model based on empirical wavelet transform neural net-work model(EWT-BP)was established,and the accuracy of the proposed soil moisture inversion method was evaluated.The experimental results show that there is a strong correlation between the soil moisture inverted by the EWT-BP model and the actual soil moisture,with the coefficient of determination(R2)of 0.820 and 0.844,respectively.Additionally the coefficient of determina-tion(R2)is increased by 21.66%and 28.66%compared with that of the traditional low-order polynomial fitting method,which verifies that the EWT-BP model is able to improve the accuracy of the GNSS-IR soil moisture inversion method effectively.IR soil moisture inversion accuracy and reliability.
The multipath signals received by global navigation satellite system(GNSS)receivers can be used to measure soil mois-ture.In order to solve the current problem of low accuracy of GNSS multipath interferometric remote sensing(GNSS-IR)in the field of soil moisture inversion,the P040 station of the U.S.plate boundary observation(PBO)network was employed as the research object herein,and the two satellites of this station,PRN10 and PRN14 were selected for the analysis of the signal-to-noise ratio data respectively.Moreover,the delayed phases of the reflected signals of the L2-band were taken as the input and the soil moisture value of H2O in PBO was taken as output.A soil moisture inversion model based on empirical wavelet transform neural net-work model(EWT-BP)was established,and the accuracy of the proposed soil moisture inversion method was evaluated.The experimental results show that there is a strong correlation between the soil moisture inverted by the EWT-BP model and the actual soil moisture,with the coefficient of determination(R2)of 0.820 and 0.844,respectively.Additionally the coefficient of determina-tion(R2)is increased by 21.66%and 28.66%compared with that of the traditional low-order polynomial fitting method,which verifies that the EWT-BP model is able to improve the accuracy of the GNSS-IR soil moisture inversion method effectively.IR soil moisture inversion accuracy and reliability.