A method and device for inverting a global navigation satellite system water surface multipath reflection coefficient
By releasing a linear damping medium on the water surface and combining it with machine learning, the water surface reflection coefficient is calculated using a GNSS geodesic receiver, thus solving the multipath error problem of GNSS geodesic receivers in water areas and achieving high-precision GNSS monitoring results.
Patent Information
- Application Number
- CN202610653848.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-13
- Publication Date
- 2026-07-10
AI Technical Summary
Existing GNSS geodetic receivers are susceptible to multipath errors in large water areas, resulting in high false alarm frequencies, system instability, and an inability to effectively correct carrier phase observation equation errors.
By releasing the linear damping medium on the water surface, the observation values are obtained using a GNSS geodesic receiver and the signal incident angle is calculated. The water surface reflection coefficient is inverted using machine learning, and an artificial neural network model is used to correct multipath errors, reduce surface tension, and provide expansion viscosity to weaken the influence of signal reflection.
It effectively suppressed multipath errors on the water surface, improved the stability and positioning accuracy of monitoring points, reduced the frequency of false alarms, and achieved high-precision GNSS monitoring.
Smart Images
Figure CN122361470A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of navigation satellite application technology, and in particular to a method and apparatus for inverting the multipath reflection coefficient of a water surface in a global satellite navigation system. Background Technology
[0002] GNSS geodesic receivers are widely used in deformation monitoring and navigation satellite applications. These devices can monitor coordinate changes at specific points on land. GNSS deformation monitoring technology offers many advantages, including all-weather operation, wide coverage, no line-of-sight between monitoring points, and flexible deployment. However, it is not suitable for large bodies of water. Currently, the product manuals for mature commercial GNSS receivers generally do not recommend use near large bodies of water. This is mainly because they are highly susceptible to multipath errors. Existing equipment cannot adequately correct multipath errors to meet design requirements. Furthermore, there is uncertainty in the carrier phase observation equation error correction. Consequently, early warning and monitoring systems that perform well on land experience high false alarm frequencies and system instability in large bodies of water, requiring urgent improvement. Summary of the Invention
[0003] The features and advantages of the present invention are set forth in part in the description which follows, or may be apparent from the description, or may be learned by practicing the invention.
[0004] To overcome the problems of existing technologies, this invention provides a method for inverting the multipath reflection coefficient of a global satellite navigation system, specifically including the following steps: S1. Release wave-damping material: Release a linear damping medium on the sea surface; S2. Obtaining Observation Parameters: During the effective time when the wave-suppressing effect occurs on the water surface, use a GNSS geodesic receiver to acquire observation data files and calculate the signal incident angle information; receive standard format RENIX files from the satellite using the GNSS geodesic receiver; calculate the satellite elevation angle information, obtain the specific navigation satellite signal incident angle value P, and calculate the reflected signal phase difference. ; S3. Inverting the water surface reflection coefficient α: Place known coordinate control points and several known coordinate monitoring points on the ground, and use the same type of GNSS geodesic receiver to collect observation data files of the sea surface at different time periods to form a long time series observation data file; substitute the multipath error correction term ε into the carrier phase observation equation to calculate the coordinates of the known control points and the points to be measured, and use the machine learning method to invert the water surface reflection coefficient α.
[0005] Preferably, the specific steps for retrieving the water surface reflection coefficient α using machine learning include: S301. Activate the same type of GNSS geodetic receiver installed at the several known coordinate monitoring points to collect a large number of long-time series observation files for different time periods; S302. Divide the long-term observation file into several data segments in an equal proportion. For each data segment, perform an iterative inverse calculation operation to obtain its corresponding real water surface reflection coefficient value α_real. S303. Extract GNSS features from the corresponding data segment for each real water surface reflection coefficient value α_real, and create labels and datasets; S304. Construct a supervised learning dataset and randomly divide it into a training set, a validation set, and a test set according to a preset ratio; S305. Perform normalization processing on all GNSS feature vectors in the divided training set, validation set, and test set; S306. Using an artificial neural network model as a machine learning model, the training set and validation set after normalization are input into the artificial neural network model for training; the network weights and biases inside the model are continuously adjusted iteratively through the backpropagation algorithm and optimizer until the performance of the artificial neural network model on the validation set reaches the convergence of the loss function. After training is completed, the test set is input into the converged artificial neural network model to obtain the water surface reflection coefficient α of the corresponding water area.
[0006] Preferably, the iterative inverse calculation operation includes: S30201, Parameter Initialization and Error Estimation: Set an initial constant value α_init for the water surface reflection coefficient α of the data segment, and calculate the multipath error correction term ε for all epochs of the data segment; S30202, Coordinate Calculation: Substitute the multipath error correction term ε into the carrier phase observation equation to calculate the coordinates of the known control point and the point to be measured; S30203, Inverse calculation of the true water surface reflection coefficient α: Calculate the three-dimensional error of the coordinate system and inversely calculate the true water surface reflection coefficient value α_real.
[0007] Preferably, the multipath error correction term ε is calculated using the following formula: Formula (1).
[0008] Preferably, the artificial neural network model uses GNSS feature vectors as the input layer, outputs a continuous scalar value α as the output layer, and has three layers of neurons as fully connected hidden layers.
[0009] Preferably, the artificial neural network model is: Formula (2), Where X is the input real-time GNSS feature vector; These are the weight coefficients corresponding to the output layer; These are all the weight coefficients within the model; For all bias terms within the model; This represents the activation output value of the j-th neuron in the last hidden layer. For the bias term of the output layer; This involves summing the results over all K neurons in the last hidden layer.
[0010] Preferably, the convergence of the loss function is achieved through a regression loss function, the calculation formula of which is as follows: Formula (3), in, Let be the global regression loss function of the artificial neural network; For all weights and bias parameters in the network; The predicted value of the m-th sample output by the artificial neural network model; This represents the true value of the physical inversion of the m-th sample; The total number of training samples. The regularization coefficient is used. The norm of the weight matrix of the l-th layer is squared. The total number of layers in the artificial neural network.
[0011] Preferably, the predicted value of the reflection coefficient at the current epoch output by the artificial neural network model is... The calculation formula is: Formula (4).
[0012] Preferably, the water surface linear damping medium comprises fatty acid compounds; the fatty acid compounds include straight-chain fatty acids with a carbon chain length of C12-C26; the straight-chain fatty acids include one or more of oleic acid, stearic acid, palmitic acid, lauric acid, and myristic acid.
[0013] Preferably, the present invention also provides an inversion device for the multipath reflection coefficient of a global navigation satellite system (GNSS). The inversion device is used to implement the inversion method for the multipath reflection coefficient of a GNSS as described above. The inversion device includes a storage tank, a drip pipe at the bottom of the storage tank, a floating component at the bottom of the storage tank, an anchor chain at the bottom of the floating component, and an anchor stone at the bottom of the anchor chain. The top of the storage tank is provided with a liquid inlet.
[0014] The beneficial effects of this invention are as follows: When the linear damping medium of this invention comes into contact with the water surface, the molecules can be closely arranged at the interface due to the small steric hindrance of the straight-chain structure, forming a high-density two-dimensional fluid film. This dense arrangement significantly reduces the surface tension of the water surface. According to the principles of fluid mechanics, the reduction in surface tension increases the difficulty of exciting short-wavelength surface waves (i.e., capillary waves), while the expanded viscosity accelerates the attenuation of capillary waves by dissipating the mechanical energy of the waves.
[0015] This invention applies a linear damping medium to key water surface reflection areas via spraying or dripping. This medium forms a Gibbs adsorption layer at the water-air interface, reducing surface tension and increasing surface viscosity. This decreases the amplitude of short-period water surface fluctuations in the key reflection areas, thereby mitigating the impact of specular reflection / scattering components of GNSS signals on received observations. Utilizing the Gibbs adsorption layer formed on the seawater surface by the wave-damping material, and combining this with GNSS satellite signal incident angle information, a model is established to calculate the water surface multipath reflection coefficient. This multipath reflection coefficient is based on the error correction of the subsequent carrier phase observation equation solution, and incorporates machine learning methods to achieve adaptive updates of observation indicators and parameters. This suppresses multipath errors in water surface environments, improves the stability of monitoring point coordinates, and solves the problem of high false alarm frequency in large water areas. Attached Figure Description
[0016] The present invention will be described in detail below with reference to the accompanying drawings and examples. The advantages and implementation methods of the present invention will become more apparent from this description. The accompanying drawings are for illustrative purposes only and do not constitute any limitation on the present invention. In the accompanying drawings: Figure 1 This is a flowchart illustrating a method for inverting the multipath reflection coefficient of a global satellite navigation system in a specific embodiment of the present invention. Figure 2 This is a cross-sectional schematic diagram of a device for inverting the multipath reflection coefficient of a global satellite navigation system in a specific embodiment of the present invention; Figure 3 This is a top view of a device for inverting the multipath reflection coefficient of a global satellite navigation system in a specific embodiment of the present invention. Detailed Implementation
[0017] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.
[0018] In the description of this invention, unless otherwise stated, "a plurality of" or "several" means two or more; the terms "upper," "lower," "inner," "outer," "top," "bottom," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only a simplified description for the convenience of describing this invention, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.
[0019] like Figure 1 As shown, this invention provides a method for inverting the multipath reflection coefficient of a global satellite navigation system, specifically including the following steps: S1. Release of wave-damping material: Release a linear damping medium onto the sea surface; the linear damping medium comprises fatty acid compounds; the fatty acid compounds include straight-chain fatty acids with carbon chain lengths between C12 and C26; the straight-chain fatty acids include one or more of oleic acid, stearic acid, palmitic acid, lauric acid, and myristic acid. Due to the lower steric hindrance of the straight-chain structure, the molecules can be closely arranged at the interface to form a high-density two-dimensional fluid film. This dense arrangement significantly reduces the surface tension of the water surface. According to the principles of fluid dynamics, the reduction in surface tension increases the difficulty of exciting short-wavelength surface waves (i.e., capillary waves), while the expanded viscosity accelerates the attenuation of capillary waves by dissipating the mechanical energy of the waves.
[0020] S2. Obtaining Observation Parameters: During the effective time when the wave-suppressing effect occurs on the water surface, use a GNSS geodesic receiver to acquire observation data files and calculate the signal incident angle information; receive standard format RENIX files from the satellite using the GNSS geodesic receiver; calculate the satellite elevation angle information, obtain the specific navigation satellite signal incident angle value P, and calculate the reflected signal phase difference. The phase difference of the reflected signal The calculation formula is as follows: Formula 5, Where H is the height of the GNSS geodesic receiver above the installation location, and λ is the corresponding navigation satellite carrier wavelength; S3. Inverting the water surface reflection coefficient α: Place known coordinate control points and several known coordinate monitoring points on the ground, and use the same type of GNSS geodesic receiver to collect observation data files of the sea surface at different time periods to form a long time series observation data file; substitute the multipath error correction term ε into the carrier phase observation equation to calculate the coordinates of the known control points and the points to be measured, and use the machine learning method to invert the water surface reflection coefficient α.
[0021] Furthermore, the specific steps for retrieving the water surface reflection coefficient α using machine learning include: S301. After the wave-damping material is evenly distributed on the water surface for 10 minutes, that is, after the water surface is stable, the same type of GNSS geodetic receiver installed at the several known coordinate monitoring points is activated to collect a large number of long-time series observation files of different time periods. S302. Divide the long-term observation file into several data segments in proportion. In this embodiment, the long-term observation file is divided into 30-minute units. Then, for each data segment, perform an iterative back calculation operation to obtain its corresponding real water surface reflection coefficient value α_real. S303. Extract GNSS features from the corresponding data segment for each real water surface reflection coefficient value α_real, and create labels and datasets; S304. Construct a supervised learning dataset and randomly divide it into a training set, a validation set, and a test set according to a preset ratio; in this embodiment, the preset ratio is 70% for the training set, 15% for the validation set, and 15% for the test set. S305. Perform normalization processing on all GNSS feature vectors in the divided training set, validation set, and test set; S306. An artificial neural network model is used as the machine learning model. The normally processed training and validation sets are input into the artificial neural network model for training. The network weights and biases within the model are iteratively adjusted using the backpropagation algorithm and optimizer until the performance of the artificial neural network model on the validation set reaches the convergence of the loss function. After training, the test set is input into the converged artificial neural network model to obtain the water surface reflection coefficient α of the corresponding water area. The optimizer is the Adam optimizer.
[0022] Furthermore, the artificial neural network model uses GNSS feature vectors as the input layer, outputs a continuous scalar value α as the output layer, and has three fully connected hidden layers. The GNSS feature vectors include the original signal-to-noise ratio (SNR), detrended SNR, SNR fluctuation amplitude, SNR peak power, elevation angle sine, antenna vertical height, azimuth angle, carrier phase residual, SNR standard deviation, and carrier wavelength. In this embodiment, the input layer has 10 neurons, and the output layer has 1 neuron, meaning the final goal is to output a continuous scalar value α. The hidden layers are three fully connected layers with 64→128→64 neurons, with densely connected layers in the middle responsible for nonlinear feature extraction.
[0023] Furthermore, the iterative inverse calculation operation includes: S30201, Parameter Initialization and Error Estimation: An initial constant value α_init is set for the water surface reflection coefficient α of the data segment, and the multipath error correction term ε is calculated for all epochs of the data segment; the calculation formula for the multipath error correction term ε is as follows: Formula 1, S30202, Coordinate Calculation: Substitute the multipath error correction term ε into the carrier phase observation equation to calculate the coordinates of the known control point and the point to be measured; S30203, Calculate the real water surface reflection coefficient α_real: Calculate the three-dimensional coordinate error and calculate the real water surface reflection coefficient α_real.
[0024] Furthermore, the artificial neural network model is an inversion model, expressed as: Formula 2, Where X is the input real-time GNSS feature vector; These are the weight coefficients corresponding to the output layer; These are all the weight coefficients within the model; For all bias terms within the model; This represents the activation output value of the j-th neuron in the last hidden layer. For the bias term of the output layer, This is to sum over all K neurons in the last hidden layer (layer L). In this embodiment, this parameter... , These are the parameters that reach their optimal state on the validation set after training and optimization using the aforementioned loss function, thereby enabling the aforementioned artificial neural network model to converge.
[0025] Furthermore, the convergence of the loss function is achieved through a regression loss function, the calculation formula of which is shown below: Formula 3, in, Let be the global regression loss function of the artificial neural network; For all weights and bias parameters in the network; The predicted value of the m-th sample output by the artificial neural network model; This represents the true value of the physical inversion of the m-th sample; The total number of training samples. The regularization coefficient is used. The norm of the weight matrix of the l-th layer is squared. Let be the total number of layers in the artificial neural network. The regression loss function can force the predicted value α̂ to numerically approximate the physical truth α_real, and prevent the artificial neural network model from overfitting noise by constraining the norm of the weights.
[0026] Furthermore, the predicted value of the current epoch reflection coefficient output by the artificial neural network model. The calculation formula is: Formula 4, in, This represents the summation of all K neurons in the last hidden layer (layer L).
[0027] S4. Verify the water surface reflection coefficient α: Fixed GNSS receivers were deployed in a location in Panama, and control points with known three-dimensional high-precision coordinates were selected as observation stations for long-term continuous observation. Multi-epoch data acquisition improved the stability and reliability of parameter estimation.
[0028] (2) Calculation of multipath effect error, which involves inverting the model... Substitute into the following formula to calculate: Formula 6, in, Represents the reflection coefficient. The phase difference introduced for the reflection path.
[0029] (3) Establish a carrier phase model that includes multipath effects: , in: For geometric distance, For integer ambiguity, At the speed of light, For the clock difference between the receiver and the satellite, For the delay of the ionosphere and troposphere, This is the comprehensive error term (including multipath residuals).
[0030] By introducing the aforementioned multi-path model, it is possible to... The systematic errors in the model are modeled and mitigated.
[0031] (4) Based on multi-epoch observation data, the observation equations are solved to obtain the receiver's three-dimensional coordinates. Through multipath correction and error suppression, the stability and accuracy of the solution results are improved.
[0032] (5) Compare the calculated three-dimensional coordinates with the known control point coordinates and calculate the root mean square error (RMSE) as the accuracy evaluation index: when the RMSE is less than 10 mm, the method is considered to meet the requirements of high-precision positioning and monitoring. The results show that under long-term observation conditions, the root mean square error (RMSE) of the three-dimensional positioning method proposed in this invention is better than 10 mm, the plane accuracy reaches the millimeter level, and the elevation accuracy is better than the centimeter level, which meets the requirements of high-precision GNSS static positioning.
[0033] Experimental results under complex experimental conditions demonstrate that the constructed reflection coefficient... The inversion model can effectively reduce the systematic errors caused by multipath effects, improve the consistency and reliability of the positioning solution, and achieve an overall accuracy of millimeter to centimeter level, meeting the requirements of high-precision engineering practice.
[0034] Furthermore, such as Figures 2-3 As shown, the present invention also provides an inversion device for the water surface multipath reflection coefficient of a global navigation satellite system. The inversion device is used to implement the inversion method for the water surface multipath reflection coefficient of a global navigation satellite system as described above. The inversion device includes a storage tank 1, a drip pipe 2 located at the bottom of the storage tank 1, a floating component 3 located at the bottom of the storage tank 1, an anchor chain 4 located at the bottom of the floating component 3, and an anchor stone 5 located at the bottom of the anchor chain 4. The top of the storage tank 1 is provided with a liquid inlet 6, and a sealing cap 7 is provided on the liquid inlet 6. The floating component 3 can be an airbag or a small hull, as long as it can float on the water surface. In use, the water surface linear damping medium is placed in the storage tank 1, and then the entire inversion device is placed at the designated coordinates in the designated water area. The anchor chain 4 is placed in the water, and the float 3 will make the entire device float in the water area. Under the action of gravity, the water surface linear damping medium drips from the drip tube 2 onto the water surface, thereby realizing the formation of a Gibbs adsorption layer through the directional arrangement of the water surface linear damping medium in the water-air interface on the water surface. By reducing the surface tension and providing the surface expansion viscosity, the short-period fluctuation amplitude of the water surface in the key reflection area is reduced, thereby weakening the influence of the specular reflection / scattering components of the GNSS signal on the received observation.
[0035] The preferred embodiments of the present invention have been described above with reference to the accompanying drawings. Those skilled in the art can implement the present invention in various modifications without departing from its scope and spirit. For example, a feature shown or described in one embodiment can be used in another embodiment to obtain yet another embodiment. The above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. All equivalent changes made based on the description and drawings of the present invention are included within the scope of the present invention.
Claims
1. A method for inverting the multipath reflection coefficient of a water surface in a global satellite navigation system, characterized in that, Specifically, it includes the following steps: S1. Release wave-damping material: Release a linear damping medium on the sea surface; S2. Obtain observation parameters: During the effective time when the wave-suppressing effect is generated on the water surface, use a GNSS geodetic receiver to obtain the observation value file and calculate the signal incident angle information; Receive standard format RENIX files from satellites using a GNSS geodetic receiver; Calculate the satellite elevation angle information, obtain the specific incident angle value P of the navigation satellite signal, and calculate the phase difference of the reflected signal. ; S3. Inverting the water surface reflection coefficient α: Place known coordinate control points and several known coordinate monitoring points on the ground, and use the same type of GNSS geodesic receiver to collect observation data files of the sea surface at different time periods to form a long time series observation data file; substitute the multipath error correction term ε into the carrier phase observation equation to calculate the coordinates of the known control points and the points to be measured, and use the machine learning method to invert the water surface reflection coefficient α.
2. The method for inverting the multipath reflection coefficient of a global satellite navigation system according to claim 1, characterized in that, The specific steps for retrieving the water surface reflection coefficient α using machine learning include: S301. Activate the same type of GNSS geodetic receiver installed at the several known coordinate monitoring points to collect a large number of long-time series observation files for different time periods; S302. Divide the long-term observation file into several data segments in an equal proportion. For each data segment, perform an iterative inverse calculation operation to obtain its corresponding real water surface reflection coefficient value α_real. S303. Extract GNSS features from the corresponding data segment for each real water surface reflection coefficient value α_real, and create labels and datasets; S304. Construct a supervised learning dataset and randomly divide it into a training set, a validation set, and a test set according to a preset ratio; S305. Perform normalization processing on all GNSS feature vectors in the divided training set, validation set, and test set; S306. Using an artificial neural network model as a machine learning model, the training set and validation set after normalization are input into the artificial neural network model for training; the network weights and biases inside the model are continuously adjusted iteratively through the backpropagation algorithm and optimizer until the performance of the artificial neural network model on the validation set reaches the convergence of the loss function. After training is completed, the test set is input into the converged artificial neural network model to obtain the water surface reflection coefficient α of the corresponding water area.
3. The method for inverting the multipath reflection coefficient of a global satellite navigation system according to claim 2, characterized in that, The iterative inverse calculation operation includes: S30201, Parameter Initialization and Error Estimation: Set an initial constant value α_init for the water surface reflection coefficient α of the data segment, and calculate the multipath error correction term ε for all epochs of the data segment; S30202, Coordinate Calculation: Substitute the multipath error correction term ε into the carrier phase observation equation to calculate the coordinates of the known control point and the point to be measured; S30203, Inverse calculation of the true water surface reflection coefficient α: Calculate the three-dimensional error of the coordinate system and inversely calculate the true water surface reflection coefficient value α_real.
4. The method for inverting the multipath reflection coefficient of a global satellite navigation system according to claim 3, characterized in that, The formula for calculating the multipath error correction term ε is as follows: Official (1).
5. The method for inverting the multipath reflection coefficient of a global satellite navigation system according to claim 2, characterized in that, The artificial neural network uses GNSS feature vectors as input layers, outputs a continuous scalar value α as output layers, and has three layers of neurons as fully connected hidden layers.
6. The method for inverting the multipath reflection coefficient of a global satellite navigation system according to claim 5, characterized in that, The artificial neural network model is as follows: Official (2) Where X is the input real-time GNSS feature vector; These are the weight coefficients corresponding to the output layer; These are all the weight coefficients within the model; For all bias terms within the model; This represents the activation output value of the j-th neuron in the last hidden layer. For the bias term of the output layer; This involves summing the results over all K neurons in the last hidden layer.
7. The method for inverting the multipath reflection coefficient of a global satellite navigation system according to claim 6, characterized in that, The convergence of the loss function is achieved through a regression loss function, the calculation formula of which is shown below: Official (3) in, Let be the global regression loss function of the artificial neural network; For all weights and bias parameters in the network; The predicted value of the m-th sample output by the artificial neural network model; This represents the true value of the physical inversion of the m-th sample; The total number of training samples. The regularization coefficient is used. The norm of the weight matrix of the l-th layer is squared. The total number of layers in the artificial neural network.
8. The method for inverting the multipath reflection coefficient of a global satellite navigation system according to claim 7, characterized in that, The predicted value of the reflection coefficient at the current epoch output by the artificial neural network model. The calculation formula is: Official (4).
9. The method for inverting the multipath reflection coefficient of a global satellite navigation system according to claim 1, characterized in that, The linear damping medium on the water surface contains fatty acid compounds; the fatty acid compounds include straight-chain fatty acids with a carbon chain length of C12-C26; the straight-chain fatty acids include one or more of oleic acid, stearic acid, palmitic acid, lauric acid, and myristic acid.
10. An inversion device for the multipath reflection coefficient of a water surface in a global satellite navigation system, characterized in that, The inversion device is used to implement the inversion method for the multipath reflection coefficient of a global satellite navigation system as described in any one of claims 1-9. The inversion device includes a liquid storage tank, a drip pipe disposed at the bottom of the liquid storage tank, a floating component disposed at the bottom of the liquid storage tank, an anchor chain disposed at the bottom of the floating component, and an anchor stone disposed at the bottom of the anchor chain; the top of the liquid storage tank is provided with a liquid inlet.