基于熵值法轨迹聚类多特征参数融合的GNSS-IR土壤湿度反演方法

GNSS-IR Soil Moisture Estimation by Track Clustering and Multi-characteristic Parameter Fusion Using Entropy Method

    摘要
  • 摘要: 全球导航卫星系统干涉反射(global navigation satellite system interferometric reflectometry, GNSS-IR)技术作为一种近地遥感的新兴手段，在土壤湿度监测方面凭借其低成本、高精度等优点成为近年来的研究热点。为了提高GNSS-IR技术反演土壤湿度的精度，采用位于南京市溧水区的自建GNSS测站原始观测数据，提取GPS、BDS、GLONASS、Galileo等系统信噪比（signal-to-noise ratio, SNR）观测数据的多路径干涉相位、振幅和频率等特征参数，分析不同频段、不同轨迹的特征参数随土壤湿度的变化规律，提出一种顾及卫星轨迹差异的多特征数据融合的GNSS-IR土壤湿度反演方法。首先按照不同卫星轨道、不同频段进行轨迹聚类融合，然后采用熵值法进行土壤湿度反演，并将多系统轨迹融合反演结果与传统均权融合方法、多元线性回归方法进行比较。结果表明，SNR的相位、振幅、频率3种特征参数组合进行土壤湿度融合反演比单一相位特征参数及相位、振幅两种特征参数组合的反演结果更优；多系统轨迹融合较单系统轨迹融合反演性能普遍提升，平均相关系数相比单系统提高了4.0%，均方根误差降低了22.8%~39.9%；基于熵值法的多系统轨迹聚类融合土壤湿度反演方法较传统均权融合方法、多元线性回归方法以及赋权融合法反演的均方根误差分别降低34.0%、25.6%和29.5%。所提方法能够提供长期、准确的土壤湿度反演结果。

     

    Abstract:
    Objectives Global navigation satellite system (GNSS) interferometric reflectometry (GNSS-IR) technology, as an emerging approach in near-Earth remote sensing, has become a research hotspot in recent years in the area of soil moisture monitoring owing to its low cost and high precision.
    Methods To enhance the accuracy of GNSS-IR technology in soil moisture retrieval, GNSS observations from a self-built station in Lishui District, Nanjing City, China were used. First, feature parameters including multipath coherent phase, amplitude, and frequency of signal-to-noise ratio (SNR) observations from GPS, BDS, GLONASS and Galileo systems were extracted. Then, based on analysis of characteristics of SNR changes with soil moisture in different systems, frequency bands and orbits, this paper proposed a multi GNSS -system feature-level data fusion inversion method that accounts for differences of satellite trajectory. Trajectory clustering-based fusion was then performed according to orbital types and frequency bands for the observation data from multiple GNSS systems. After determing the weight using entropy method, soil moisture inversion was carried out. The results of multi-system trajectory fusion inversion were compared with those traditional average-weight fusion method and the multiple linear regression method.
    Results The soil moisture fusion inversion result of the combination of feature parameters including phase, amplitude, and frequency of SNR is better than the inversion results of a single phase feature parameter and the combination of two feature parameters including phase and amplitude. Multi-system trajectory fusion generally improves the inversion performance compared to single system trajectory fusion, with an average correlation coefficient increasing by 4.0% and root mean square error （RMSE) decreasing by 22.8% to 39.9%. Compared to traditional average-weight fusion method, multiple linear regression method and weighted fusion method, the multi-system trajectory clustering fusion soil moisture inversion method based on entropy method reduces RMSE by 34.0%, 25.6% and 29.5%, respectively.
    Conclusions The proposed method can provide long-term and accurate soil moisture inversion results.

     

    • HTML全文
    • 参考文献(21)
    • 相关文章
    • 施引文献
  • 资源附件(0)

/

返回文章
返回