Multi-evidence fusion gnss data quality control method, system and device

Abstract

The application relates to a multi-evidence fusion GNSS data quality control method, system and equipment, and belongs to the technical field of satellite navigation and engineering safety monitoring. Data acquisition and feature extraction are carried out first, then dynamic weight calculation and occultation screening are carried out, multi-evidence fusion cycle slip detection is carried out, cycle slip repair and state machine updating are carried out, atmospheric delay correction and geometric observation construction are carried out, time sequence anomaly removal and adaptive noise estimation are carried out, the quality indexes of GNSS monitoring data are calculated based on the previous steps, and a structured quality label is generated. The application can significantly improve the reliability, continuity and traceability of GNSS monitoring data, and provides stable input for millimeter-level displacement monitoring and intelligent alarm.

Description

Technical Field

[0001] This invention relates to a GNSS (Global Navigation Satellite System) data quality control method, system, and equipment that integrates multiple evidences, and is particularly suitable for high-precision displacement monitoring scenarios of dams, slopes, bridges, landslides, and other major engineering structures, belonging to the field of satellite navigation and engineering safety monitoring technology. Background Technology

[0002] BeiDou / GNSS technology has been widely applied in the field of engineering structure safety monitoring, providing continuous, all-weather, and high-precision three-dimensional displacement information for structures such as dams, slopes, and bridges. In long-term monitoring scenarios, GNSS observation data is often affected by factors such as antenna obstruction, multipath effects, ionospheric and tropospheric disturbances, carrier cycle slips, receiver noise, and communication interruptions, resulting in problems such as noise abrupt changes, abnormal jumps, short-term missing measurements, and unstable quality in the observation sequence.

[0003] Existing GNSS data preprocessing and quality control methods have the following shortcomings: First, cycle slip detection often relies on a single combination of quantities or a fixed threshold; second, existing methods provide a relatively coarse description of observation quality, making it difficult to form structured quality labels that can be used for subsequent filtering, change detection, and alarms; third, existing quality control methods and back-end filters are usually independent of each other, lacking the ability to perform online feedback adjustment based on the filter innovation sequence.

[0004] Therefore, it is necessary to develop a GNSS monitoring data preprocessing and quality control method that combines multi-evidence fusion, state machine constraints, adaptive noise estimation, and quality label output to improve the reliability, continuity, and interpretability of millimeter-level displacement monitoring data in complex environments. Summary of the Invention

[0005] The purpose of this invention is to overcome the above-mentioned shortcomings and provide a GNSS data quality control method that integrates multiple evidences, which can significantly improve the reliability, continuity and traceability of GNSS monitoring data, and provide stable input for millimeter-level displacement monitoring and intelligent alarm.

[0006] The technical solution adopted in this invention is as follows: A method for quality control of GNSS data integrating multiple pieces of evidence includes the following steps: S1. Data Acquisition and Feature Extraction: Acquire the raw observation data and / or positioning solution results output by the GNSS monitoring station, and construct the Melbourne-Wübbena (MW) combination, Geometry-Free (GF) combination, and pseudorange and carrier phase differential features based on pseudorange and carrier phase observations; S2. Dynamic Weight Calculation and Occultation Screening: Based on signal-to-noise ratio, satellite elevation angle, and multipath impact index, dynamic weight calculation is performed on the observations of each satellite and frequency point; for observations with low signal-to-noise ratio, low satellite elevation angle, or strong multipath impact, the system performs weight reduction processing; for observations below the quality requirements, they are eliminated or marked. S3. Multi-evidence Fusion Cycle Slip Detection: For observation data filtered by occultation, multiple cycle slip discrimination evidences between adjacent epochs are calculated, including Melbourne-Wübbena combined difference, Geometry-Free combined difference, pseudorange and carrier phase difference variation, Doppler variation, residual energy, and wavelet energy. A multi-evidence vector corresponding to the satellite-frequency point is constructed. After normalizing the evidence vector, it is input into a confidence mapping function to obtain the cycle slip occurrence confidence. Combined with a dynamic threshold and a satellite-frequency point state machine, the occurrence and type of cycle slip are determined. S4. Cycle slip repair and state machine update: For satellite-frequency points that are determined to have cycle slips or are suspected of having cycle slips, based on cycle slip confidence, number of consecutive abnormal epochs, combined observation change characteristics and current state machine state, perform integer cycle repair, ambiguity reinitialization, observation weight clearing or temporary removal operations. S5. Perform atmospheric delay correction and geometric observation construction on the observation data after filtering and cycle slip processing; S6. Time series anomaly removal and adaptive noise estimation: The observation data obtained in step S5 is processed by the positioning engine to obtain the ENU displacement time series, and anomaly point identification, short-term missing measurement compensation and smoothing are performed. S7. Based on the previous steps, calculate the quality indicators of the GNSS monitoring data and generate structured quality labels. The quality indicators include one or more of the following: DOP value, ambiguity ratio, residual RMS, effective satellite number, cycle slip count, missing measurement duration, observation weights, signal-to-noise ratio statistics, number of outliers, and positioning status stability indicators.

[0007] The data mentioned in step S1 of the above method includes, but is not limited to, epoch timestamp, satellite number, frequency number, pseudorange observation, carrier phase observation, Doppler observation, signal-to-noise ratio, satellite elevation angle, positioning mode, DOP value (Dilution of Precision), ambiguity ratio, root mean square residual value, and receiver status information. For the Each era, the first satellite, the first Extract pseudorange observations from each frequency point. Carrier phase observations Signal-to-noise ratio, satellite elevation angle For dual-frequency observations, and using Doppler observations, construct the Melbourne-Wübbena combination: , in, For wide-lane wavelength, Indicates the first The satellite in the Melbourne-Wübbena combined observations for each epoch; and They represent the first and the Pseudorange observations at frequency points; and They represent the first and the Carrier phase observations at frequency points; Constructing Geometry-Free Combinations: ; in, Indicates the first The satellite in the Geometry-Free combined observations corresponding to each epoch; Indicates the first The satellite in the Each era, the first Carrier phase observations at frequency points; Indicates the first The satellite in the Each era, the first Carrier phase observations at frequency points; and They represent and The carrier wavelength corresponding to the frequency point; Construct pseudorange and carrier phase difference features: , in, Indicates the first The satellite in the Each era, the first Pseudorange and carrier phase difference characteristics at each frequency point; This represents the corresponding pseudorange observation; This represents the corresponding carrier phase observation value; Indicates the first The carrier wavelength corresponding to each frequency point.

[0008] The calculation process for the dynamic weights in step S2 is as follows: First, calculate the signal-to-noise ratio weights as follows: , in, SNR The current observed signal-to-noise ratio, This is the lower limit threshold for the signal-to-noise ratio. This is the signal-to-noise ratio reference threshold; Next, calculate the satellite elevation angle weights, using a sine function or a piecewise function: , The overall observation weight is calculated and expressed as follows: , in, For the first Era 1 The first satellite The observation weight of each frequency point Weights are determined by the impact of multiple paths.

[0009] The weighted observation set includes the satellite number, frequency point number, pseudorange observation, carrier phase observation, Doppler observation, overall observation weight, quality label, and validity mask of the observations retained in each epoch.

[0010] The evidence vector mentioned in step S3 is: , in, For the Melbourne-Wübbena combinatorial difference, For Geometry-Free Combinatorial Difference, This is the difference between the pseudorange and the carrier phase. For Doppler variation, For residual energy, Wavelet energy; The evidence vector is normalized to obtain And the confidence level of cycle slip occurrence is calculated using the confidence mapping function: , in, For the first The satellite in the Confidence level of cycle slip at epoch, This is the normalized evidence vector. For the evidence weight vector, For bias terms, It can be a logistic regression function, an exponential function, or another nonlinear mapping function.

[0011] In step S4, the system maintains an independent state machine for each satellite-frequency point. The state machine includes at least a normal state, a suspected state, a repair / reinitialization state, and a cooldown state. In the normal state, when the confidence level of a cycle slip exceeds a first threshold, the system enters the suspected state. In the suspected state, when the confidence level of cycle slips in multiple consecutive epochs exceeds a second threshold, a cycle slip is confirmed, and the system enters the repair or reinitialization state. After the repair is completed, the system enters the cooldown state. During the cooldown period, the system raises the cycle slip judgment threshold or restricts repeated ambiguity reinitialization to avoid frequent state jumps due to noise fluctuations or critical observation conditions. After the cooldown period ends, if the observation quality returns to normal, the state machine returns to the normal state. This step realizes the confirmation, repair, suppression of repeated triggering, and continuous state management of cycle slip events.

[0012] In step S5, a free combination of ionospheric components is constructed based on dual-frequency or multi-frequency observation data to reduce the impact of the first-order ionospheric delay; simultaneously, the tropospheric delay is estimated and corrected by combining the tropospheric model and mapping function. Tropospheric delay is expressed as: , in, For the zenith tropospheric delay, A mapping function related to satellite elevation angle; T s,k The delay is the result of the correction. For dual-frequency carrier phase observations, construct a free combination of ionospheric components: , in, and Let be the carrier frequencies of the two frequency points, and . For the corresponding carrier phase observations, and For the corresponding wavelength, L IF This represents the carrier phase observation value after the ionosphere has freely combined.

[0013] In step S6, each component in the ENU displacement sequence Perform Hampel filtering and calculate within a sliding window. Median: , Calculate the median absolute deviation: , in, Indicates the first Each epoch corresponds to the median absolute deviation within the sliding window; Represents the median function; Indicates the first [number]th ... Each sample value; This represents the median within the sliding window; It represents the absolute value of the deviation between the sample value and the median; If the following conditions are met: , in, Indicates the first Displacement sample values ​​for each epoch; This represents the median within the corresponding sliding window; This indicates the absolute deviation of the current sample value from the median; This indicates the threshold for anomaly detection. The median absolute deviation is expressed as follows: when the above inequality holds, it indicates that the current sample may be an outlier. If a value is identified as an outlier, the median or interpolated value will be used instead. For filter state estimation, let the innovation vector be: , in, Indicates the first The filter innovation vector of each epoch; Indicates the first The observation vector of each epoch; Represents the observation matrix; Indicates the first Prior state estimation vector for each epoch; superscript This indicates that the state estimate is a predicted value or a priori estimate; The observation noise covariance matrix is ​​then updated using the following formula: , in, For the first epoch observation noise covariance matrix, As a smoothing factor, The system can adjust the observation weight parameters and cycle slip determination threshold in reverse based on the updated noise level, using the filter's innovative vector.

[0014] The observation weights mentioned in step S7 are expressed as follows: , For the first Era 1 The first satellite The observation weights of each frequency point.

[0015] Another objective of this invention is to provide a GNSS data quality control system that integrates multiple pieces of evidence, including a data acquisition and feature extraction module, a dynamic weight calculation and occultation screening module, a multi-evidence fusion cycle slip detection module, a cycle slip repair and state machine module, an atmospheric correction and geometric observation construction module, a time series anomaly removal and adaptive noise estimation module, and a quality index and quality label generation module. The data acquisition and feature extraction module is used to acquire the raw observation data and / or positioning solution results output by the GNSS monitoring station, and construct the Melbourne-Wübbena (MW) combination, Geometry-Free (GF) combination, and pseudorange and carrier phase differential features based on pseudorange observations and carrier phase observations; The dynamic weight calculation and occultation screening module is used to perform dynamic weight calculation on the observations of each satellite and frequency point based on the signal-to-noise ratio, satellite elevation angle and multipath impact index, to form a weighted observation set, and to reduce the weight or remove the observations with low elevation angle and / or low signal-to-noise ratio. The multi-evidence fusion cycle slip detection module is used to calculate the MW combination difference, GF combination difference, pseudorange and carrier phase difference change, Doppler change, residual energy and / or wavelet energy of adjacent epochs, construct the evidence vector corresponding to the satellite-frequency point, normalize the evidence vector and input it into the confidence mapping function to obtain the cycle slip occurrence confidence, and combine the dynamic threshold and the satellite-frequency point state machine to determine whether a cycle slip has occurred and the type of cycle slip; The cycle slip repair and state machine module performs cycle slip repair, ambiguity re-initialization, observation weight clearing, or temporary removal operations on satellite-frequency points that are determined to have cycle slips or are suspected of having cycle slips, based on cycle slip confidence, number of consecutive abnormal epochs, combined observation change characteristics, and current state machine status. The Atmospheric Correction and Geometric Observation Construction Module performs atmospheric delay correction and geometric observation construction on the observation data after filtering and cycle slip processing (constructing ionospheric free combination and / or Geometry-Free geometric observations based on dual-frequency or multi-frequency observations). The time series anomaly removal and adaptive noise estimation module performs anomaly identification, short-term missing measurement compensation, and smoothing on the ENU displacement time series obtained by atmospheric correction and geometric observation construction through the positioning engine calculation and conversion. The quality index and quality label generation module is used to calculate one or more quality indices, including DOP value, ambiguity ratio, residual RMS, effective satellite number, cycle slip count, missing measurement duration, observation weight sum, signal-to-noise ratio statistics, number of outliers, and positioning status stability index, and generate structured quality labels based on the quality indices.

[0016] The multi-evidence fusion cycle slip detection module includes an evidence vector construction unit, an evidence normalization unit, a confidence calculation unit, and a dynamic threshold determination unit.

[0017] An electronic device for GNSS data quality control that integrates multiple pieces of evidence includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the GNSS data quality control method that integrates multiple pieces of evidence as described above.

[0018] A GNSS data quality control storage device that integrates multiple evidences, wherein the storage device stores a program that, when executed by a processor, implements the steps of the GNSS data quality control method integrating multiple evidences as described above.

[0019] The beneficial effects of this invention are: (1) Cycle slip detection is performed by multi-evidence fusion, which makes full use of multi-source information such as MW, GF, pseudorange-carrier phase, Doppler, residual energy and wavelet energy, thereby reducing the risk of false detection and false negative detection of cycle slip caused by a single criterion; (2) The introduction of satellite-frequency state machine and cooling mechanism can suppress repeated judgments and frequent re-initialization of ambiguity under critical conditions, thereby improving the stability of data processing; (3) Establish a closed-loop control mechanism between QC (quality control) and the filter innovation sequence, which can dynamically adjust the weight and judgment threshold according to the observed noise changes, thereby enhancing the long-term operational adaptability; (4) Output quality indicators and quality labels, which can provide interpretable data for subsequent filtering, change detection, intelligent alarm and operation audit; (5) This invention is applicable to the deployment of edge computing devices and can meet the needs of real-time, continuous and traceable millimeter-level displacement monitoring in engineering monitoring scenarios.

[0020] This invention constructs an evidence vector containing multiple observation features, uses a confidence mapping function to achieve probabilistic cycle slip determination, and combines a satellite-frequency state machine to control the cycle slip confirmation, repair, and cooling process. Simultaneously, it utilizes an innovative filtering sequence to perform online feedback adjustment of the observation noise covariance and QC threshold, thereby forming a closed-loop mechanism between quality control and state estimation. This significantly improves the reliability, continuity, and traceability of GNSS monitoring data, providing stable input for millimeter-level displacement monitoring and intelligent alarms. Attached Figure Description

[0021] Figure 1 A schematic diagram of the overall process of the method of this invention; Figure 2 A schematic diagram of the multi-evidence fusion cycle slip detection module of this invention; Figure 3 Schematic diagram of the satellite-frequency state machine of this invention; Figure 4 Schematic diagram of the innovative sequence closed-loop control of QC-filtering in this invention; Figure 5 System structure block diagram of the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be understood that the following embodiments are only used to illustrate the present invention and are not intended to limit the scope of protection of the present invention.

[0023] Example 1: A GNSS data quality control method integrating multiple pieces of evidence, including steps (such as...) Figure 1 )as follows: S1. Data Acquisition and Feature Extraction: Acquire the raw observation data and / or positioning solution results output by the GNSS monitoring station. The data includes, but is not limited to, epoch timestamps, satellite numbers, frequency numbers, pseudorange observations, carrier phase observations, Doppler observations, signal-to-noise ratio, satellite elevation angle, positioning mode, DOP value (Dilution of Precision), ambiguity ratio, root mean square residual value, and receiver status information.

[0024] The edge monitoring terminal receives raw observation data and / or real-time positioning results output by the GNSS receiver. The raw observation data may include RINEX, RTCM, or receiver-specific binary format data, and the real-time positioning results may include NMEA, JSON, Protobuf, or other data formats.

[0025] For the Each era, the first satellite, the first Extract pseudorange observations from each frequency point. Carrier phase observations Signal-to-noise ratio Satellite elevation angle Basic features such as Doppler observations are used. Furthermore, based on pseudorange and carrier phase observations, Melbourne-Wübbena combinatorial and Geometry-Free combinatorial features, as well as pseudorange and carrier phase difference features, are constructed to provide multi-dimensional inputs for subsequent dynamic weighting, cycle slip detection, and quality assessment.

[0026] For dual-frequency observations, construct the Melbourne-Wübbena combination: , in, For wide-lane wavelength, Indicates the first The satellite in the Melbourne-Wübbena combined observations for each epoch; and They represent the first and the Pseudorange observations at frequency points; and They represent the first and the Carrier phase observations at frequency points; and These represent the corresponding carrier wavelengths.

[0027] Constructing Geometry-Free Combinations: ; in, Indicates the first The satellite in the Geometry-Free combined observations corresponding to each epoch; Indicates the satellite number; Indicates the epoch number; and Indicates two different carrier frequencies; Indicates the first The satellite in the Each era, the first Carrier phase observations at frequency points; Indicates the first The satellite in the Each era, the first Carrier phase observations at frequency points; and They represent and The carrier wavelength corresponding to the frequency point.

[0028] Construct pseudorange and carrier phase difference features: , in, Indicates the first The satellite in the Each era, the first Pseudorange and carrier phase difference characteristics at each frequency point; This represents the corresponding pseudorange observation; This represents the corresponding carrier phase observation value; Indicates the first The carrier wavelength corresponding to each frequency point; , , These represent the satellite number, frequency point number, and epoch number, respectively.

[0029] Through the above processing, the raw GNSS observation data is transformed into a feature set containing signal quality, geometric quality, combined observation and residual statistics information.

[0030] S2. Dynamic Weight Calculation and Occultation Screening: Based on signal-to-noise ratio, satellite elevation angle, and multipath impact indicators, the system dynamically calculates the weights of observations from each satellite and at each frequency. Observations with low signal-to-noise ratio, low satellite elevation angle, or strong multipath impact are downweighted; observations with severe anomalies or that do not meet the minimum quality requirements are removed or marked.

[0031] The signal-to-noise ratio weight can be calculated as follows: , in, The current observed signal-to-noise ratio, This is the lower limit threshold for the signal-to-noise ratio. This is the reference threshold for the signal-to-noise ratio.

[0032] Satellite elevation angle weights can be represented using a sine function or a piecewise function, for example: , in, Indicates the satellite elevation angle weight. Indicates the first In the first observation epoch The elevation angle of a satellite relative to the receiver's ground plane. This formula indicates that the weight of satellite observations increases with increasing elevation angle. When the satellite elevation angle is high, its signal propagation path is short, and it is less affected by obstruction, multipath effects, and atmospheric delay residual errors, thus it is assigned a larger weight; when the satellite elevation angle is low, its signal is more susceptible to obstruction, multipath reflections, and atmospheric delay errors, thus it is assigned a smaller weight; when the satellite is below the horizon or the elevation angle is unavailable, the weight is determined by... The weight is limited to 0 to avoid negative weights. This elevation angle weight can increase the contribution of high-elevation satellite observations when forming a weighted observation set, and reduce the impact of low-elevation satellite anomalies on the displacement calculation results.

[0033] The comprehensive observation weight can be expressed as: , in, For the first Era 1 The first satellite The observation weight of each frequency point Weights are determined by the impact of multiple paths.

[0034] This step generates a weighted observation set, providing reliable input for subsequent cycle slip detection, state estimation, and quality assessment.

[0035] When the satellite elevation angle is lower than a preset threshold, the signal-to-noise ratio is lower than a preset threshold, or the multipath effect exceeds a preset threshold, the system can downweight, remove, or mark the observation as a low-quality observation.

[0036] In this embodiment, the weighted observation set specifically includes the satellite number, frequency point number, pseudorange observation, carrier phase observation, Doppler observation, integrated observation weight, quality label, and validity mask of the observations retained in each epoch. For example, the 1st... The weighted set of observations for each epoch can be denoted as: If a certain observation and If so, the observation is considered a valid weighted observation and enters subsequent cycle slip detection and solution; if If the observation is not included in subsequent calculations, it will only be retained for quality traceability.

[0037] S3. Multi-evidence Fusion Cycle Slip Detection: For observation data filtered by occultation, multiple cycle slip discrimination evidences between adjacent epochs are calculated, including Melbourne-Wübbena combined difference, Geometry-Free combined difference, pseudorange and carrier phase difference variation, Doppler variation, residual energy, and wavelet energy. Multi-evidence vectors corresponding to satellite-frequency points are constructed. After normalizing the evidence vectors, they are input into a confidence mapping function to obtain the cycle slip occurrence confidence. Combined with a dynamic threshold and a satellite-frequency state machine, the occurrence and type of cycle slip are determined.

[0038] like Figure 2 As shown, the multi-evidence cycle slip detection module receives multiple observed features, including: , in, Indicates the first The satellite in the The change in the Melbourne-Wübbena combination relative to the previous epoch; Indicates the first The satellite in the Melbourne-Wübbena combined observations for each epoch; Indicates the first The satellite in the Melbourne-Wübbena combined observations for each epoch.

[0039] , in, Indicates the first The satellite in the Geometry-Free combinational change of an epoch relative to the previous epoch; Indicates the first The satellite in the Geometry-Free Combined Observations for Each Ephemeris; Indicates the first The satellite in the Geometry-Free Combined Observations for Each Ephemeris.

[0040] , in, Indicates the first The satellite in the Each era, the first The difference between pseudorange and carrier phase at each frequency point; Indicates the first Pseudorange and carrier phase difference characteristics of each epoch; Indicates the first The pseudorange and carrier phase difference characteristics of each epoch.

[0041] , in, Indicates the first The satellite in the Each era, the first Doppler observations of change at each frequency point; Indicates the first Doppler observations for each epoch; Indicates the first Doppler observations for each epoch.

[0042] Simultaneously calculate residual energy and wavelet energy In this embodiment, the residual energy can be calculated using the weighted sum of squares of the posterior residuals or the filtered innovative residuals, for example: , in, Indicates the first The residual energy of each epoch; This represents the residual vector composed of pseudorange residuals, carrier phase residuals, or filter innovation residuals. Indicates the number of residual samples; This represents the residual weighted matrix.

[0043] Wavelet energy can be calculated using the sum of squares of the high-frequency wavelet detail coefficients within a sliding window, for example: , in, Indicates the first Wavelet energy of each epoch; to Indicates the range of wavelet decomposition levels involved in the statistics; Indicates the first Number of layer wavelet detail coefficients; This represents the corresponding wavelet detail coefficients. Therefore, residual energy reflects the degree of model residual anomaly, while wavelet energy reflects the degree of short-term abrupt changes and high-frequency anomalies; both serve as auxiliary evidence for cycle slip detection.

[0044] Constructing the evidence vector: , in, Indicates the first The satellite in the Multiple evidence vectors corresponding to each epoch; This represents the Melbourne-Wübbena combinatorial difference; This represents a geometry-free combinatorial difference; This represents the difference between the pseudorange and the carrier phase. Indicates the change in Doppler; Represents residual energy; Indicates wavelet energy; superscript This represents the transpose of a vector.

[0045] The evidence vector is normalized to obtain And calculate the cycle slip confidence using the confidence mapping function: , Logical functions can be used: , in, For the first The first satellite Epoch cycle slip confidence level, It is a nonlinear mapping function. For the evidence weight vector, As a bias term, when When the dynamic threshold corresponding to the current state is exceeded, the satellite-frequency point is determined to have a suspected cycle slip or a confirmed cycle slip.

[0046] By combining dynamic thresholds and satellite-frequency state machines, the occurrence and type of cycle slips can be determined, thereby reducing false positives and false negatives caused by a single criterion in complex environments.

[0047] S4. Cycle slip repair and state machine update: For satellite-frequency points that are determined to have cycle slips or are suspected of having cycle slips, based on the cycle slip confidence, the number of consecutive abnormal epochs, the combined observation change characteristics, and the current state machine state, perform integer cycle repair, ambiguity re-initialization, observation weight clearing, or temporary removal operations.

[0048] like Figure 3 The diagram shows the maintenance state machine for each satellite-frequency point. The states include at least a normal state, a suspected state, and a cooling-off period state. In the normal state, if... If the threshold is exceeded, the system enters a suspected state; in the suspected state, if multiple consecutive epochs occur... If the second threshold is exceeded, a cycle slip is confirmed. After confirmation, depending on the cycle slip type, integer cycle repair, ambiguity re-initialization, or observation weight reset is performed.

[0049] In this embodiment, cycle slip types can be categorized according to their impact range and processing method into single-frequency cycle slips, multi-frequency synchronous cycle slips, single-epoch sudden cycle slips, continuous cycle slips, small-amplitude repairable cycle slips, large-amplitude cycle slips, and unrepairable cycle slips. For single-frequency small-amplitude cycle slips, integer cycle repair can be prioritized; for multi-frequency synchronous cycle slips or continuous cycle slips, ambiguity re-initialization can be performed; for unrepairable cycle slips, the satellite-frequency observation is temporarily removed or its weight is reduced.

[0050] After cycle jump confirmation, the state machine enters a cooldown period. During the cooldown period, the dynamic threshold can be increased or repeated reinitialization can be prohibited to avoid frequent state transitions caused by critical noise. After the cooldown period ends, the state machine returns to its normal state.

[0051] This step enables the confirmation, repair, suppression of repeated triggering, and continuous state management of cycle slip events.

[0052] S5. For the observation data that has passed screening and cycle slip processing (quality control), atmospheric delay correction and geometric observation construction are performed. In this embodiment, observations that have passed quality control refer to those that are retained in the weighted observation set and whose overall observation weight is greater than or equal to... Effectiveness mask is There are no unrepaired cycle slips, and observations are not marked as DATA_GAP, HIGH_NOISE, or unrepairable cycle slips. Ionospheric free combinations can be constructed based on dual-frequency or multi-frequency observation data to mitigate the effects of first-order ionospheric delay; simultaneously, tropospheric delay can be estimated and corrected by combining tropospheric models and mapping functions.

[0053] Tropospheric delay is estimated from observations that have passed quality control. The tropospheric delay can be expressed as: , in, For zenith tropospheric delay, This is a mapping function.

[0054] For dual-frequency observations, free combinations of ionospheric components can be constructed: , in, This represents the carrier phase observation value after the ionosphere has freely combined; and These represent the carrier frequencies of the first and second frequency points, respectively. and These represent the carrier phase observations at the first and second frequency points, respectively. and These represent the carrier wavelengths corresponding to the first and second frequency points, respectively. ,in It is the speed of light. Because... and It is usually expressed in carrier cycles, so by multiplying by... and This is converted into a length quantity. This combination utilizes the property that ionospheric delay is inversely proportional to the square of the carrier frequency to linearly combine the dual-frequency carrier phase observations, thereby eliminating or significantly reducing the first-order ionospheric delay error, improving the reliability of subsequent high-precision positioning and displacement calculation, and reducing the influence of the first-order ionospheric delay.

[0055] This yields geometric observations suitable for subsequent high-precision positioning, filtering, and smoothing. These geometric observations are then incorporated into the subsequent positioning solution or filtering state estimation process, outputting ENU displacement and covariance information. Therefore, the processing result of step S5 is directly connected to the displacement time series processing in step S6.

[0056] S6. Time Series Anomaly Removal and Adaptive Noise Estimation: The ENU displacement time series obtained from the observation data in step S5, processed and transformed by the positioning engine, is subjected to Hampel filtering to remove isolated outliers. Short-term missing data are interpolated to fill in the gaps. The observation noise covariance matrix is ​​updated online based on the filter's innovative sequence, and the observation weight parameters and cycle slip threshold are adjusted inversely. Here, the ENU displacement sequence refers to the displacement sequence formed after the aforementioned filtered and atmospherically corrected observation data is processed by the positioning engine and transformed to the monitoring station's local East-North-Sky coordinate system, including three components: E, N, and U.

[0057] For each part of the ENU displacement sequence The median is then subjected to Hampel filtering. The median is calculated within the sliding window. , Calculate the median absolute deviation: , in, Indicates the first Each epoch corresponds to the median absolute deviation within the sliding window; Represents the median function; Indicates the first [number]th ... Each sample value; This represents the median within the sliding window; It represents the absolute value of the deviation between the sample value and the median.

[0058] If the following conditions are met: , in, Indicates the first Displacement sample values ​​for each epoch; This represents the median within the corresponding sliding window; This indicates the absolute deviation of the current sample value from the median; This indicates the threshold for anomaly detection. This represents the absolute deviation of the median. When the above inequality holds, it indicates that the current sample may be an outlier.

[0059] Then If a value is identified as an outlier, the median or interpolation result will be used instead.

[0060] For filter state estimation, let the innovation vector be: , in, Indicates the first The filter innovation vector of each epoch; Indicates the first The observation vector of each epoch; Represents the observation matrix; Indicates the first Prior state estimation vector for each epoch; superscript This indicates that the state estimate is a predicted value or a priori estimate.

[0061] The observation noise covariance matrix can then be updated using the following formula: , in, For the first epoch observation noise covariance matrix, As a smoothing factor, This is the innovation vector for the filter. Based on the updated noise level, the system can adjust the observation weight parameters and cycle slip threshold in reverse. When the innovation variance continues to increase, the system can adjust the observation weight parameters and the dynamic threshold in S3 in reverse, forming a vector as shown in Figure 3. Figure 4 The closed-loop control shown.

[0062] S7. Quality Indicator and Quality Label Generation: Calculate at least one of the following quality indicators: DOP (Dilution of Precision), ambiguity ratio, root mean square (RMS) residuals, number of effective satellites, cycle slips per unit time, missing measurement duration, and weighted sum. Generate quality labels based on the quality indicators and output preprocessed results containing ENU displacement results, covariance information, quality indicators, and quality labels.

[0063] The system generates quality labels based on various quality indicators. These labels include, but are not limited to, LOW_SNR, LOW_ELEV, HIGH_MP, SLIP_DETECTED, SLIP_REPAIRED, TROPO_ANOMALY, HIGH_NOISE, DATA_GAP, LOW_WEIGHT, and NORMAL. Specifically, LOW_SNR indicates a low signal-to-noise ratio, LOW_ELEV indicates a low satellite elevation angle, HIGH_MP indicates strong multipath interference, SLIP_DETECTED indicates a cycle slip detected, SLIP_REPAIRED indicates a cycle slip has been corrected, TROPO_ANOMALY indicates a tropospheric anomaly, HIGH_NOISE indicates a high noise level, DATA_GAP indicates missing or interrupted data, LOW_WEIGHT indicates low observation weight, and NORMAL indicates normal data quality. These quality labels provide a structured basis for subsequent filtering, common mode error suppression, change detection, alarm analysis, and data traceability.

[0064] The final output includes preprocessed results such as ENU displacement results, covariance information, observation quality indicators, anomaly records, and quality labels. This result can serve as input for subsequent displacement filtering, common mode error suppression, change detection, intelligent alarms, and data storage, enabling a complete quality control process for GNSS monitoring data from raw observations to high-reliability displacement sequences.

[0065] DOP can be calculated based on the weighted geometric matrix of the current epoch, for example: , in, Indicates the first Precision factor for each epoch; Indicates the first The observation geometry matrix for each epoch; Represents the observation weight matrix; This represents the matrix trace operation.

[0066] The residual RMS can be expressed as: , in, Indicates the first The root mean square of the residuals of each epoch; Indicates the first The first calendar year One residual sample; This represents the number of residual samples. The number of effective satellites can be expressed as satisfying the validity mask. And weight The number of satellites; the duration of missing data can be obtained by multiplying the number of consecutive missing epochs by the sampling interval; covariance information can be output after transforming the positioning solution or filter state covariance to the ENU coordinate system, for example... ,in This is the coordinate transformation matrix. It is the covariance matrix in the geocentric geofixed coordinate system.

[0067] The observation weights can be expressed as: , in, Indicates the first The summation is the total weight of all valid satellite and frequency observations for each epoch; the summation range includes the satellite numbers involved in the calculation. and frequency point number ; Indicates the first The satellite in the Each era, the first The comprehensive observation weights at each frequency point.

[0068] Engineering application examples: In a GNSS displacement monitoring scenario for a dam, the system receives multi-frequency, multi-constellation GNSS observation data at a sampling rate of 1 Hz. The satellite elevation angle threshold is set to... to The lower limit of the signal-to-noise ratio is 30 dB-Hz, and the reference signal-to-noise ratio is 45 dB-Hz.

[0069] The system performs data parsing, weight calculation, and multi-evidence cycle slip detection for each epoch. When a satellite's... , When both pseudorange and carrier phase differential are abnormal, cycle slip confidence If the dynamic threshold is exceeded, the state machine enters a suspected state; if the threshold condition is still met for multiple consecutive epochs, the cycle slip is confirmed and ambiguity is reinitialized. The system then enters a cooling-off period to avoid repeated triggering.

[0070] The preprocessed ENU displacement sequence enters the filtering and monitoring module. The system estimates the observation noise covariance online based on the innovative sequence and adaptively adjusts the quality control threshold. Finally, it outputs millimeter-level displacement sequences, covariance information, and quality labels, providing a data foundation for subsequent change detection and intelligent alarms.

[0071] Example 2: A GNSS data quality control system integrating multiple evidences, including a data acquisition and feature extraction module, a dynamic weight calculation and occultation screening module, a multi-evidence fusion cycle slip detection module, a cycle slip repair and state machine module, an atmospheric correction and geometric observation construction module, a time series anomaly removal and adaptive noise estimation module, and a quality index and quality label generation module. The data acquisition and feature extraction module is used to receive and acquire raw observation data and / or positioning solution results output by the GNSS monitoring station. The data includes, but is not limited to, epoch timestamps, satellite numbers, frequency numbers, pseudorange observations, carrier phase observations, Doppler observations, signal-to-noise ratio, satellite elevation angle, positioning mode, DOP value (Dilution of Precision), ambiguity ratio, root mean square residual value, and receiver status information. Based on the pseudorange observations and carrier phase observations, the module constructs Melbourne-Wübbena (MW) combinations, Geometry-Free (GF) combinations, and pseudorange and carrier phase differential features. The dynamic weight calculation and occultation screening module is used to perform dynamic weight calculation on the observations of each satellite and frequency point based on the signal-to-noise ratio, satellite elevation angle and multipath impact index, to form a weighted observation set, and to reduce the weight or remove the observations with low elevation angle and / or low signal-to-noise ratio. The multi-evidence fusion cycle slip detection module is used to calculate the MW combination difference, GF combination difference, pseudorange and carrier phase difference change, Doppler change, residual energy and / or wavelet energy of adjacent epochs, construct the evidence vector corresponding to the satellite-frequency point, normalize the evidence vector and input it into the confidence mapping function to obtain the cycle slip occurrence confidence, and combine the dynamic threshold and the satellite-frequency point state machine to determine whether a cycle slip has occurred and the type of cycle slip; The cycle slip repair and state machine module performs cycle slip repair, ambiguity re-initialization, observation weight clearing, or temporary removal operations on satellite-frequency points that are determined to have cycle slips or are suspected of having cycle slips, based on cycle slip confidence, number of consecutive abnormal epochs, combined observation change characteristics, and current state machine status. The Atmospheric Correction and Geometric Observation Construction Module performs atmospheric delay correction and geometric observation construction on the observation data after filtering and cycle slip processing (constructing ionospheric free combination and / or Geometry-Free geometric observations based on dual-frequency or multi-frequency observations). The time series anomaly removal and adaptive noise estimation module performs anomaly identification, short-term missing measurement compensation, and smoothing on the ENU displacement time series obtained by atmospheric correction and geometric observation construction through the positioning engine calculation and conversion. The quality index and quality label generation module is used to calculate one or more quality indices, including DOP value, ambiguity ratio, residual RMS, effective satellite number, cycle slip count, missing measurement duration, observation weight sum, signal-to-noise ratio statistics, number of outliers, and positioning status stability index, and generate structured quality labels based on the quality indices.

[0072] The multi-evidence fusion cycle slip detection module includes an evidence vector construction unit, an evidence normalization unit, a confidence calculation unit, and a dynamic threshold determination unit.

[0073] Each module can be deployed in an edge computing terminal, a monitoring receiver terminal, or a server, and can be implemented through software processes, embedded programs, or executable services.

[0074] An electronic device for GNSS data quality control that integrates multiple pieces of evidence includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the GNSS data quality control method for integrating multiple pieces of evidence as described in Embodiment 1 above.

[0075] A GNSS data quality control storage device that integrates multiple evidences, wherein the storage device stores a program, and when the program is executed by a processor, it implements the steps of the GNSS data quality control method that integrates multiple evidences as described in Embodiment 1 above.

[0076] The above is a detailed description of the present invention in conjunction with specific embodiments, and the scope of protection of the present invention is not limited thereto.

Claims

1. A GNSS data quality control method integrating multiple pieces of evidence, characterized by: The steps include the following: S1. Data Acquisition and Feature Extraction: Acquire the raw observation data and / or positioning solution results output by the GNSS monitoring station, and construct the Melbourne-Wübbena combination, Geometry-Free combination, and pseudorange and carrier phase differential features based on pseudorange and carrier phase observations; S2. Dynamic Weight Calculation and Occultation Screening: Based on signal-to-noise ratio, satellite elevation angle, and multipath impact index, dynamic weight calculation is performed on the observations of each satellite and frequency point; for observations with low signal-to-noise ratio, low satellite elevation angle, or strong multipath impact, the system performs weight reduction processing; for observations below the quality requirements, they are eliminated or marked. S3. Multi-evidence Fusion Cycle Slip Detection: For observation data filtered by occultation, multiple cycle slip discrimination evidences between adjacent epochs are calculated, including Melbourne-Wübbena combined difference, Geometry-Free combined difference, pseudorange and carrier phase difference variation, Doppler variation, residual energy, and wavelet energy. A multi-evidence vector corresponding to the satellite-frequency point is constructed. After normalizing the evidence vector, it is input into a confidence mapping function to obtain the cycle slip occurrence confidence. Combined with a dynamic threshold and a satellite-frequency point state machine, the occurrence and type of cycle slip are determined. S4. Cycle slip repair and state machine update: For satellite-frequency points that are determined to have cycle slips or are suspected of having cycle slips, based on cycle slip confidence, number of consecutive abnormal epochs, combined observation change characteristics and current state machine state, perform integer cycle repair, ambiguity reinitialization, observation weight clearing or temporary removal operations. S5. Perform atmospheric delay correction and geometric observation construction on the observation data after filtering and cycle slip processing; S6. Time series anomaly removal and adaptive noise estimation: The observation data obtained in step S5 is processed by the positioning engine to obtain the ENU displacement time series, and anomaly point identification, short-term missing measurement compensation and smoothing are performed. S7. Based on the previous steps, calculate the quality indicators of the GNSS monitoring data and generate structured quality labels. The quality indicators include one or more of the following: DOP value, ambiguity ratio, residual RMS, effective satellite number, cycle slip count, missing measurement duration, observation weights, signal-to-noise ratio statistics, number of outliers, and positioning status stability indicators.

2. The GNSS data quality control method based on fusion of multiple evidence as described in claim 1, characterized in that, The data mentioned in step S1 includes, but is not limited to, epoch timestamp, satellite number, frequency point number, pseudorange observation, carrier phase observation, Doppler observation, signal-to-noise ratio, satellite elevation angle, positioning mode, DOP value, ambiguity ratio, root mean square residual value, and receiver status information.

3. The GNSS data quality control method based on fusion of multiple evidence as described in claim 1, characterized in that, In step S1, for dual-frequency observations, construct the Melbourne-Wübbena combination: ， in, For wide-lane wavelength, Indicates the first The satellite in the Melbourne-Wübbena combined observations for each epoch; and They represent the first and the Pseudorange observations at frequency points; and They represent the first and second, respectively. Carrier phase observations at frequency points; Constructing Geometry-Free Combinations: ， in, Indicates the first The satellite in the Geometry-Free combined observations corresponding to each epoch; Indicates the first The satellite in the Each era, the first Carrier phase observations at frequency points; Indicates the first The satellite in the Each era, the first Carrier phase observations at frequency points; and They represent and The carrier wavelength corresponding to the frequency point; Construct pseudorange and carrier phase difference features: ， in, Indicates the first The satellite in the Each era, the first Pseudorange and carrier phase difference characteristics at each frequency point; This represents the corresponding pseudorange observation; This represents the corresponding carrier phase observation value; Indicates the first The carrier wavelength corresponding to each frequency point.

4. The GNSS data quality control method based on fusion of multiple evidence as described in claim 1, characterized in that, The calculation process of the dynamic weights mentioned in step S2 is as follows: First, calculate the signal-to-noise ratio weights as follows: ， in, The current observed signal-to-noise ratio, This is the lower limit threshold for the signal-to-noise ratio. This is the signal-to-noise ratio reference threshold; Next, calculate the satellite elevation angle weights, using a sine function or a piecewise function: ， The overall observation weight is calculated and expressed as follows: ， in, For the first Era 1 The first satellite The observation weight of each frequency point Weights are influenced by multiple paths.

5. The GNSS data quality control method based on fusion of multiple evidences according to claim 1, characterized in that, The evidence vector mentioned in step S3 is: ， in, For the Melbourne-Wübbena combinatorial difference, For Geometry-Free Combinatorial Difference, This is the difference between the pseudorange and the carrier phase. For Doppler variation, For residual energy, Wavelet energy; The evidence vector is normalized to obtain And the confidence level of cycle slip occurrence is calculated using the confidence mapping function: ， in, For the first The satellite in the Confidence level of cycle slip at epoch, This is the normalized evidence vector. For the evidence weight vector, For bias terms, It can be a logistic regression function, an exponential function, or another nonlinear mapping function.

6. The GNSS data quality control method based on fusion of multiple evidences according to claim 1, characterized in that, step... In S4, an independent state machine is maintained for each satellite-frequency point. The state machine includes at least a normal state, a suspected state, a repair / reinitialization state, and a cooling-off period state. In the normal state, when the confidence of a cycle slip exceeds the first threshold, it enters the suspected state. In the suspected state, when the confidence level of cycle slips in multiple consecutive epochs exceeds the second threshold, the cycle slip is confirmed to have occurred, and the process enters the repair or reinitialization state. After the repair is completed, the system enters a cooling-off period. During the cooling-off period, the system raises the cycle slip judgment threshold or restricts the re-initialization of repeated ambiguity to avoid frequent state jumps due to noise fluctuations or critical observation conditions. After the cooling-off period ends, if the observation quality returns to normal, the state machine returns to the normal state. This step realizes the confirmation, repair, suppression of repeated triggering, and continuous state management of cycle slip events.

7. The GNSS data quality control method based on fusion of multiple evidences according to claim 1, characterized in that, In step S6, each component in the ENU displacement sequence Perform Hampel filtering and calculate within a sliding window. Median: ， Calculate the median absolute deviation: ， in, Indicates the first Each epoch corresponds to the median absolute deviation within the sliding window; Represents the median function; Indicates the first [number]th ... Each sample value; This represents the median within the sliding window; It represents the absolute value of the deviation between the sample value and the median; If the following conditions are met: ， in, Indicates the first Displacement sample values ​​for each epoch; This represents the median within the corresponding sliding window; This indicates the absolute deviation of the current sample value from the median; This indicates the threshold for anomaly detection. The median absolute deviation is expressed as follows: when the above inequality holds, it indicates that the current sample may be an outlier. If a value is identified as an outlier, the median or interpolated value will be used instead. For filter state estimation, let the innovation vector be: ， in, Indicates the first The filter innovation vector of each epoch; Indicates the first The observation vector of each epoch; Represents the observation matrix; Indicates the first Prior state estimation vector for each epoch; superscript This indicates that the state estimate is a predicted value or a priori estimate; The observation noise covariance matrix is ​​then updated using the following formula: ， in, For the first epoch observation noise covariance matrix, As a smoothing factor, The system can adjust the observation weight parameters and cycle slip determination threshold in reverse based on the updated noise level, using the filter's innovative vector.

8. A GNSS data quality control system integrating multiple pieces of evidence, characterized in that: It includes modules for data acquisition and feature extraction, dynamic weight calculation and occultation screening, multi-evidence fusion and cycle slip detection, cycle slip repair and state machine, atmospheric correction and geometric observation construction, time series anomaly removal and adaptive noise estimation, and quality index and quality label generation. The data acquisition and feature extraction module is used to acquire the raw observation data and / or positioning solution results output by the GNSS monitoring station, and construct the Melbourne-Wübbena combination, Geometry-Free combination, and pseudorange and carrier phase difference features based on pseudorange observations and carrier phase observations. The dynamic weight calculation and occultation screening module is used to perform dynamic weight calculation on the observations of each satellite and frequency point based on the signal-to-noise ratio, satellite elevation angle and multipath impact index, to form a weighted observation set, and to reduce the weight or remove the observations with low elevation angle and / or low signal-to-noise ratio. The multi-evidence fusion cycle slip detection module is used to calculate the MW combination difference, GF combination difference, pseudorange and carrier phase difference change, Doppler change, residual energy and / or wavelet energy of adjacent epochs, construct the evidence vector corresponding to the satellite-frequency point, normalize the evidence vector and input it into the confidence mapping function to obtain the cycle slip occurrence confidence, and combine the dynamic threshold and satellite-frequency point state machine to determine whether a cycle slip has occurred and the type of cycle slip; The cycle slip repair and state machine module performs cycle slip repair, ambiguity re-initialization, observation weight clearing, or temporary removal operations on satellite-frequency points that are determined to have cycle slips or are suspected of having cycle slips, based on cycle slip confidence, number of consecutive abnormal epochs, combined observation change characteristics, and current state machine status. The Atmospheric Correction and Geometric Observation Construction Module performs atmospheric delay correction and geometric observation construction on the observation data after filtering and cycle slip processing. The time series anomaly removal and adaptive noise estimation module performs anomaly identification, short-term missing measurement compensation, and smoothing on the ENU displacement time series obtained by atmospheric correction and geometric observation construction through the positioning engine calculation and conversion. The quality index and quality label generation module is used to calculate one or more quality indices, including DOP value, ambiguity ratio, residual RMS, effective satellite number, cycle slip count, missing measurement duration, observation weight sum, signal-to-noise ratio statistics, number of outliers, and positioning status stability index, and generate structured quality labels based on the quality indices.

9. An electronic device for GNSS data quality control integrating multiple pieces of evidence, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the GNSS data quality control method for fusing multiple evidences as described in any one of claims 1-7.

10. A GNSS data quality control storage device integrating multiple pieces of evidence, wherein the storage device stores a program, characterized in that, When the program is executed by the processor, it implements the steps of the GNSS data quality control method for fusing multiple evidences as described in any one of claims 1-7.

Images