CN115048952A - GNSS-IR soil humidity inversion method integrating robust estimation and machine learning - Google Patents

GNSS-IR soil humidity inversion method integrating robust estimation and machine learning Download PDF

Info

Publication number
CN115048952A
CN115048952A CN202210397523.9A CN202210397523A CN115048952A CN 115048952 A CN115048952 A CN 115048952A CN 202210397523 A CN202210397523 A CN 202210397523A CN 115048952 A CN115048952 A CN 115048952A
Authority
CN
China
Prior art keywords
signal
gnss
machine learning
soil humidity
noise ratio
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210397523.9A
Other languages
Chinese (zh)
Inventor
郑南山
丁锐
何佳星
周勋
张秋昭
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China University of Mining and Technology CUMT
Original Assignee
China University of Mining and Technology CUMT
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China University of Mining and Technology CUMT filed Critical China University of Mining and Technology CUMT
Priority to CN202210397523.9A priority Critical patent/CN115048952A/en
Publication of CN115048952A publication Critical patent/CN115048952A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • G06N3/084Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention discloses a GNSS-IR soil humidity inversion method fusing robust estimation and machine learning, relates to the technical field, and particularly relates to a GNSS-IR soil humidity inversion method fusing robust estimation and machine learning, which comprises the following steps: s1, extracting the signal-to-noise ratio and the altitude angle data of the GNSS satellite data; s2, intercepting the signal-to-noise ratio data of a specific altitude angle, fitting the direct signal by adopting low-order polynomial fitting, removing the direct signal, and extracting the signal-to-noise ratio component of the reflected signal; s3, processing the SNR observed value of the reflected signal by a Lomb-Scargle spectrum analysis method, wherein the frequency corresponding to the maximum value of the power spectrum amplitude of the spectrogram is the oscillation frequency of the reflected signal component, and calculating the effective antenna height; and S4, performing cosine fitting through a nonlinear least square algorithm to obtain the amplitude and the phase of the characteristic parameters of the reflected signal. The method can effectively reduce the influence of the quality of signal-to-noise ratio observation data and complex surface environment factors, fully utilize the information of the reflected signals and improve the soil humidity inversion accuracy of the GNSS-IR.

Description

GNSS-IR soil humidity inversion method integrating robust estimation and machine learning
Technical Field
The invention relates to an inversion method, in particular to a GNSS-IR soil humidity inversion method integrating robust estimation and machine learning, and belongs to the technical field of satellite remote sensing.
Background
Soil is an important homeland resource. The soil humidity plays an important role in the aspects of researching ecological water circulation, vegetation water supply, land bearing capacity and the like, so that the timely and accurate monitoring of soil humidity information is very necessary. The traditional soil humidity monitoring means such as a soil hygrometer method has accurate results but consumes a large amount of manpower and material resources, and the measurement period is long; although the satellite remote sensing observation mode can overcome the defect of low spatial resolution of the traditional soil humidity monitoring means, the time resolution is not ideal. Therefore, the GNSS-IR technology based on the satellite interference signal gradually becomes a new technology for monitoring soil humidity by virtue of its high time resolution and high spatial resolution and draws much attention.
GNSS-IR techniques use Right Hand Circular Polarization (RHCP) antennas to receive SNR data to invert surface physical parameters. Bilich and the like research a correlation algorithm and a flow for carrying out soil humidity inversion by utilizing a signal-to-noise ratio observation value by utilizing signal-to-noise ratio observation data and meteorological observation data recorded by a GPS receiver, indicate that the frequency of the signal-to-noise ratio is related to the height change of a relative antenna, and the amplitude and the phase are related to the soil humidity change, and can monitor the soil humidity in a large range through a GPS observation station network. Bilich et al verified that strong correlation exists between multipath interference phase, amplitude, and effective height obtained by frequency calculation and soil humidity by using signal-to-noise ratio observation data of a continental US plate boundary observation network and actually measured soil humidity data. Larson indicates that the amplitude of the multipath signal is related to vegetation moisture content or height. Chew states that GPS multipath signal indicators show good agreement with field observations of vegetation height and vegetation moisture content. The vegetation water content and the soil humidity act on GNSS signals together, and the independent inversion of the vegetation water content or the soil humidity is not strict and inaccurate. However, the existing soil humidity inversion research about GNSS-IR mostly focuses on quantifying the correlation between a soil humidity value and a certain reflection signal characteristic parameter, linear modeling is carried out by combining a reference value of soil humidity, and the influence of observation errors caused by factors such as reflection surface information carried by the other two characteristic parameters and the surrounding environment of a survey station is ignored. In addition, the environment of the surface reflection surface is often very complex, and factors such as the roughness of the soil surface, the vegetation coverage condition, the surface temperature and the like influence each other and are reflected on the change of the soil humidity value. The surface environmental factors are difficult to comprehensively consider, a complex nonlinear problem is formed, and the establishment of a refined soil humidity inversion model through the formula derivation and quantification of the correlation is particularly difficult. In summary, in the aspect of the research of the GNSS-IR soil humidity inversion algorithm, the conventional GNSS-IR soil humidity inversion model is established based on the characteristic parameters of the single reflection signal, and is influenced by the quality of the signal-to-noise ratio observation data and the factors of the complex earth surface environment, the characteristic parameters of the modeling variable reflection signal inevitably have abnormal values, and the linear model is difficult to comprehensively and quantitatively analyze the relationship among the factors, so that the model has poor fitting effect and weak generalization capability
Therefore, in order to more effectively, accurately and timely invert soil humidity information by utilizing a GNSS-IR technology, fully consider the influence of signal-to-noise ratio observation data quality and complex earth surface environment factors, and fully utilize reflecting surface information carried by multi-reflection signal characteristic parameters, a GNSS-IR soil humidity inversion method for controlling data quality and comprehensively utilizing the multi-reflection signal characteristic parameters is urgently needed.
Disclosure of Invention
Aiming at the problems in the prior art, on the basis of removing abnormal values of characteristic parameters of a reflection signal by using a steady estimation MCD, three machine learning algorithms of a back propagation neural network, Gaussian process regression and random forest are introduced, a soil humidity inversion nonlinear model based on GNSS-IR is respectively established by comprehensively considering the frequency, the amplitude and the phase of the reflection signal, and the soil humidity is more effectively and accurately inverted.
In order to achieve the purpose, the invention provides a GNSS-IR soil moisture inversion method integrating robust estimation and machine learning, which comprises the following steps:
s1, extracting the signal-to-noise ratio and the altitude angle data of the GNSS satellite data;
s2, intercepting the signal-to-noise ratio data of a specific altitude angle, fitting the direct signal by adopting low-order polynomial fitting, removing the direct signal, and extracting the signal-to-noise ratio component of the reflected signal;
s3, processing the SNR observed value of the reflected signal by a Lomb-Scargle spectrum analysis method, wherein the frequency corresponding to the maximum value of the power spectrum amplitude of the spectrogram is the oscillation frequency of the reflected signal component, and calculating the effective antenna height;
s4, performing cosine fitting through a nonlinear least square algorithm to obtain the amplitude and the phase of the characteristic parameters of the reflected signal;
s5, respectively adopting MCD robust estimation to the frequency, amplitude and phase time sequence of the generated reflection signal characteristic parameters to carry out robust processing to form a new reflection signal characteristic parameter time sequence, and forming a data set with the soil humidity value;
s6, dividing the data set into a training set and a prediction set, respectively substituting the training set into a BP neural network, a Gaussian process regression and a random forest to model, and comparing the accuracy of the three machine learning algorithms through the prediction set;
and S7, using the accurate model for predicting the soil humidity value.
The invention provides a product quality detection device for radar part production, which has the following beneficial effects:
the method introduces an abnormal value detection method based on MCD steady estimation to preprocess the time sequence of the characteristic parameters of the reflection signals, establishes a nonlinear mapping relation between the characteristic parameters of the GNSS reflection signals and the soil humidity value on the basis, introduces various machine learning methods to fully consider the influence of each characteristic parameter to carry out soil humidity inversion modeling, and effectively improves the accuracy and generalization capability of a soil humidity inversion model.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flow chart of an embodiment of the present invention;
FIG. 3 is a comparison of the inversion results of the present invention and the inversion results of the conventional multiple linear regression model.
Detailed description of the preferred embodiment
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
Example 1
As shown in fig. 1, the GNSS-IR soil humidity inversion method integrating robust estimation and machine learning of the present invention specifically includes the following steps:
s1, extracting the signal-to-noise ratio and the altitude angle data of the GNSS satellite data;
s2, intercepting the signal-to-noise ratio data of a specific altitude angle, fitting the direct signal by adopting low-order polynomial fitting, removing the direct signal, and extracting the signal-to-noise ratio component of the reflected signal;
s3, processing the SNR observed value of the reflected signal by a Lomb-Scargle spectrum analysis method, wherein the frequency corresponding to the maximum value of the power spectrum amplitude of the spectrogram is the oscillation frequency of the reflected signal component, and calculating the effective antenna height;
s4, performing cosine fitting through a nonlinear least square algorithm to obtain the amplitude and the phase of the characteristic parameters of the reflected signal;
s5, respectively adopting MCD robust estimation to the frequency, amplitude and phase time sequence of the generated reflection signal characteristic parameters to carry out robust processing to form a new reflection signal characteristic parameter time sequence, and forming a data set with the soil humidity value;
s6, dividing the data set into a training set and a prediction set, respectively substituting the training set into a BP neural network, a Gaussian process regression and a random forest to model, and comparing the accuracy of the three machine learning algorithms through the prediction set;
and S7, using the accurate model for predicting the soil humidity value.
Example 2
As shown in fig. 2, the method for GNSS-IR soil humidity inversion integrating robust estimation and machine learning of the present invention specifically includes the following steps in soil humidity inversion of a BDS navigation system:
and S1, acquiring BDS satellite data. Importing observation data received by a receiver into TEQC software for data preprocessing, acquiring an SNR (signal to noise ratio) observation value, a satellite altitude angle and the like, and providing a data source for the next characteristic parameter extraction;
s2, intercepting signal-to-noise ratio data of a specific altitude angle (5 degrees to 25 degrees), fitting the direct signal by adopting low-order polynomial fitting, removing the direct signal, and extracting a signal-to-noise ratio component of the reflected signal;
s3, the SNR observed value after the direct signal is removed and sin theta can be expressed by a cosine function with fixed oscillation frequency, resampling is needed, and the relation between the SNR observed value after the direct signal is removed and sin theta is obtained, as shown in formula (1). Processing the SNR observed value of the reflection signal by a Lomb-Scargle spectrum analysis method, wherein the frequency corresponding to the maximum value of the power spectrum amplitude of the spectrogram is the oscillation frequency of the reflection signal component, and calculating the effective antenna height;
Figure RE-GDA0003743708110000061
where is SNR m For the signal-to-noise ratio of the reflected signal, A m As an amplitude parameter of the reflected signal,
Figure RE-GDA0003743708110000062
is the phase parameter of the reflected signal.
S4, performing cosine fitting on the formula (1) through a nonlinear least square algorithm to obtain the amplitude and the phase of the characteristic parameters of the reflected signal;
s5, respectively adopting MCD robust estimation to carry out robust estimation on the frequency, amplitude and phase time sequence of the generated reflection signal characteristic parameters to form a new reflection signal characteristic parameter time sequence, and forming a data set with the soil humidity value;
s6, dividing the data set into a training set and a prediction set, respectively substituting the training set into three machine learning algorithms of BP neural network, Gaussian process regression and random forest for modeling, and comparing the precision of the three machine learning algorithms through the prediction set;
and S7, using the accurate model for predicting the soil humidity value.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and improvements can be made without departing from the principle of the present invention, and these modifications and improvements should also be considered as the protection scope of the present invention.

Claims (3)

1. A GNSS-IR soil moisture inversion method integrating robust estimation and machine learning is characterized by comprising the following steps:
s1, extracting the signal-to-noise ratio and the altitude angle data of the GNSS satellite data;
s2, intercepting the signal-to-noise ratio data of a specific altitude angle, fitting the direct signal by adopting low-order polynomial fitting, removing the direct signal, and extracting the signal-to-noise ratio component of the reflected signal;
s3, processing the SNR observed value of the reflected signal by a Lomb-Scargle spectrum analysis method, wherein the frequency corresponding to the maximum value of the power spectrum amplitude of the spectrogram is the oscillation frequency of the reflected signal component, and calculating the effective antenna height;
s4, performing cosine fitting through a nonlinear least square algorithm to obtain the amplitude and the phase of the characteristic parameters of the reflected signal;
s5, respectively adopting MCD robust estimation to carry out robust estimation on the frequency, amplitude and phase time sequence of the generated reflection signal characteristic parameters to form a new reflection signal characteristic parameter time sequence, and forming a data set with the soil humidity value;
s6, dividing the data set into a training set and a prediction set, respectively substituting the training set into a BP neural network, a Gaussian process regression and a random forest to model, and comparing the accuracy of the three machine learning algorithms through the prediction set;
and S7, using the accurate model for predicting the soil humidity value.
2. The GNSS-IR soil moisture inversion method with robust estimation and machine learning combined as claimed in claim 1, wherein the step S5 employs an anti-aliasing method to denoise the time series of 3 feature parameters.
3. The GNSS-IR soil moisture inversion method with robust estimation and machine learning combined as claimed in claim 1 or 2, wherein the step S6 is performed by performing multivariate nonlinear regression modeling by using 3 characteristic parameters as variables and substituting them into machine learning algorithm.
CN202210397523.9A 2022-04-15 2022-04-15 GNSS-IR soil humidity inversion method integrating robust estimation and machine learning Pending CN115048952A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210397523.9A CN115048952A (en) 2022-04-15 2022-04-15 GNSS-IR soil humidity inversion method integrating robust estimation and machine learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210397523.9A CN115048952A (en) 2022-04-15 2022-04-15 GNSS-IR soil humidity inversion method integrating robust estimation and machine learning

Publications (1)

Publication Number Publication Date
CN115048952A true CN115048952A (en) 2022-09-13

Family

ID=83158400

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210397523.9A Pending CN115048952A (en) 2022-04-15 2022-04-15 GNSS-IR soil humidity inversion method integrating robust estimation and machine learning

Country Status (1)

Country Link
CN (1) CN115048952A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115980317A (en) * 2023-03-20 2023-04-18 中国科学院地理科学与资源研究所 Foundation GNSS-R data soil moisture estimation method based on corrected phase
CN116304524A (en) * 2022-12-20 2023-06-23 宁夏回族自治区气象科学研究所 Soil water content monitoring method, equipment, storage medium and device
CN117571968A (en) * 2024-01-12 2024-02-20 山东大学 GNSS-IR-based soil humidity calculation method

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116304524A (en) * 2022-12-20 2023-06-23 宁夏回族自治区气象科学研究所 Soil water content monitoring method, equipment, storage medium and device
CN116304524B (en) * 2022-12-20 2024-04-09 宁夏回族自治区气象科学研究所 Soil water content monitoring method, equipment, storage medium and device
CN115980317A (en) * 2023-03-20 2023-04-18 中国科学院地理科学与资源研究所 Foundation GNSS-R data soil moisture estimation method based on corrected phase
CN117571968A (en) * 2024-01-12 2024-02-20 山东大学 GNSS-IR-based soil humidity calculation method
CN117571968B (en) * 2024-01-12 2024-04-05 山东大学 GNSS-IR-based soil humidity calculation method

Similar Documents

Publication Publication Date Title
CN115048952A (en) GNSS-IR soil humidity inversion method integrating robust estimation and machine learning
CN110609287B (en) Double-frequency radar scatterometer and method for simultaneously measuring sea surface wind field and flow field
CN111337549A (en) GPS multi-satellite fusion soil humidity monitoring method based on fuzzy entropy
CN112782689A (en) Multi-satellite data fusion GNSS-IR soil humidity monitoring method
CN114397425B (en) GNSS-IR soil humidity inversion method based on generalized extension approximation
CN114397636B (en) Method, system and equipment for evaluating uniformity of reflectivity factor of ground-based radar
CN113791091A (en) GNSS-IR-based real-time continuous monitoring method for soil moisture content
Zuffada et al. Global navigation satellite system reflectometry (GNSS-R) algorithms for wetland observations
CN113075706A (en) GNSS-R based snow depth inversion method and application thereof
CN117031509A (en) Soil humidity inversion method and device integrating isolated forest and deep learning
CN110208807A (en) A kind of raininess grade inversion method based on marine radar image detection region otherness parameter
CN111337548A (en) High-precision signal-to-noise ratio fitting model and soil humidity inversion method based on same
CN106767383A (en) The measuring method of the snow depth based on continuous operation GNSS stations signal-to-noise ratio data
Wang et al. A passive technique to monitor evaporation duct height using coastal GNSS-R
CN115494086A (en) GNSS-IR soil humidity monitoring method considering abnormal interference phase
Saponaro et al. Evaluation of aerosol and cloud properties in three climate models using MODIS observations and its corresponding COSP simulator, as well as their application in aerosol–cloud interactions
Nie et al. A C-band scatterometer simultaneous wind/rain retrieval method
Unal High-resolution raindrop size distribution retrieval based on the Doppler spectrum in the case of slant profiling radar
Rao et al. Evaluation of DBS wind measurement technique in different beam configurations for a VHF wind profiler
CN116719062A (en) Satellite signal quality evaluation method and device, equipment and storage medium
CN114488107B (en) Method and device for sea clutter space-time distribution and influence grading product manufacturing
Altuntaş et al. Estimation performance of soil moisture with GPS-IR method
CN105334506A (en) Method and device for estimating sea surface wind speed based on line spectrum intensity in radar echoes
CN101907583A (en) Device and method for monitoring change of soil moisture by utilizing GNSS-R signals
CN113589350B (en) Sea surface wind speed measuring method based on measuring GNSS receiver

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination