Landslide Prediction Method and System Based on GNSS Receiver

Abstract

The present invention relates to the technical fields such as satellite ground equipment, and provides a landslide prediction method and system based on a GNSS receiver. The receiver synchronously acquires triple-frequency carrier signals, constructs a non-combined double-difference model, and hierarchically processes the tropospheric dry component and the zenith wet component; two groups of ultra-wide-lane ambiguities are generated based on the original observations. After meeting the noise threshold, they are cascaded and rounded according to the wavelength and the integer-week results are output. The integer-week results are written into integer linear constraints, and combined with geometric correlation and ionosphere-free equivalent transformation, the narrow-lane ambiguity is reduced to one dimension and the ionospheric term is completely removed. With the baseline displacement, the zenith wet component, and the remaining narrow-lane ambiguity as the states, an extended Kalman filter is implemented in combination with the elevation angle weighted noise matrix to obtain the narrow-lane floating solution. After fixing the integer by the LAMBDA algorithm and ratio test, the displacement is updated. Finally, the displacement rate and the hourly rainfall are synthesized into a hazard index according to the historical weights. When both indicators exceed the threshold, a landslide warning is output to achieve a highly reliable risk determination of the landslide.

Description

Technical Field

[0001] The present invention relates to the technical fields of satellite ground equipment, electronic digital data processing, etc., and in particular to a landslide prediction method and system based on a GNSS receiver. Background Art

[0002] The early prediction of landslides relies on centimeter-level or even millimeter-level monitoring of tiny surface deformations. Traditional single-frequency or dual-frequency GNSS real-time kinematic positioning is easily restricted by factors such as ionospheric delay, tropospheric wet component, and insufficient ambiguity fixing rate, and it is difficult to work stably for a long time under medium-long baseline conditions. With the start of the broadcast of triple-frequency signals by systems such as Beidou and GPS, the multi-frequency ultra-wide lane combination provides an opportunity to improve the reliability of rapid ambiguity fixing; however, existing methods mostly separately process the ultra-wide lane and narrow lane solutions, and do not fully utilize the integer-week result to perform dimensionality reduction constraints on the narrow lane model, resulting in the still limited solution efficiency and success rate. In addition, when the ionospheric activity is strong, traditional geometry-related models need to introduce ionospheric parameters, and it is difficult to accurately match the model errors; although the geometry-independent model can cancel the geometric terms, it loses the ability to directly estimate the baseline vector. For this reason, the prior art proposes to fuse the geometry-related and ionospheric elimination objectives into an equivalent transformation, but there is still a lack of a solution integrated with ultra-wide lane integer-week constraint, Kalman filtering, and rainfall triggering mechanism.

[0003] Aiming at the above deficiencies, there is an urgent need for a multi-frequency GNSS landslide prediction method that can ensure highly reliable ambiguity fixing, can estimate the baseline displacement in real time, suppress ionospheric and tropospheric errors, and give a landslide risk determination in combination with rainfall information. Summary of the Invention

[0004] Aiming at the deficiencies of the above-mentioned prior art, the present invention provides a landslide prediction method and system based on a GNSS receiver to achieve highly reliable ambiguity fixing, can estimate the baseline displacement in real time, suppress ionospheric and tropospheric errors, and give a landslide risk determination in combination with rainfall information.

[0005] In a first aspect, the present invention provides a landslide prediction method based on a GNSS receiver, including:

[0006] A triple-frequency non-combination double-difference observation model is established based on the three carrier frequency signals received by the receiver to obtain an original observation data set including baseline displacement, tropospheric delay, ionospheric delay, and three-dimensional narrow-lane ambiguity; the tropospheric delay includes the dry component model value and the zenith wet component to be estimated; two groups of ultra-wide-lane ambiguities are combined from the original observation data set, and the floating-point estimated values of the two groups of ultra-wide-lane ambiguities and the variances of the floating-point estimated values are calculated; if the equivalent carrier noise corresponding to the variance of the floating-point estimated value is lower than the preset threshold, the ultra-wide-lane ambiguities are cascaded and rounded in descending order of wavelength, and the qualified ultra-wide-lane ambiguity integer results are output to ensure that the overall fixed success rate meets the system requirements;

[0007] The ultra-wide-lane ambiguity integer results are written into integer linear constraints and substituted back into the triple-frequency non-combination double-difference observation model to reduce the three-dimensional narrow-lane ambiguity to be estimated to one dimension; geometrically related and ionosphere-free equivalent transformations are applied to the triple-frequency non-combination double-difference observation model, and the ionospheric delay is eliminated using the integer linear constraints to obtain a simplified observation model containing only baseline displacement, tropospheric zenith wet component, and a single narrow-lane ambiguity; with baseline displacement, zenith wet component, and the remaining narrow-lane ambiguity as state variables, combined with the elevation angle weighted observation noise covariance matrix, the floating-point solution of the narrow-lane ambiguity is recursively obtained using the extended Kalman filter model;

[0008] The floating-point solution of the narrow-lane ambiguity is input into the LAMBDA algorithm for integer fixing of the narrow-lane ambiguity, and is tested according to the fixed strategy based on ratio test. If it passes, the integer solution is accepted and the baseline displacement is updated. Based on the updated baseline displacement, the displacement rate is calculated, and the hourly rainfall provided by the rain gauge is combined with the historical calibration weight to synthesize the hazard index. When the displacement rate and the hazard index both exceed their respective thresholds, a landslide warning flag is generated.

[0009] In a second aspect, the present invention provides a landslide prediction system based on a GNSS receiver, including:

[0010] A rain gauge, arranged at one end of the top crossbar of the installation bracket, for monitoring the hourly rainfall at the landslide monitoring location;

[0011] A receiver, arranged in the middle of the top crossbar, connected and communicating with the rain gauge, and running the landslide prediction method based on a GNSS receiver according to any one of the above claims;

[0012] A lightning rod, arranged at the other end of the top crossbar and higher than the rain gauge and the receiver;

[0013] A GNSS displacement monitoring station, arranged in the middle of the installation bracket, connected and communicating with the receiver;

[0014] A solar power supply system is arranged in the middle of the installation bracket and above the GNSS displacement monitoring station to form a rain shield, and is used to supply power to the GNSS displacement monitoring station, the receiver and the rain gauge.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0016] The present invention provides a landslide prediction method and system based on a GNSS receiver. By establishing a triple-frequency non-combination double-difference observation model according to the three carrier frequency signals received by the receiver, an original observation data set including baseline displacement, tropospheric delay, ionospheric delay and three-dimensional narrow-lane ambiguity is obtained. The tropospheric delay includes a dry component model value and a zenith wet component to be estimated. Two sets of ultra-wide-lane ambiguities are combined from the original observation data set, and the floating-point estimated values of the two sets of ultra-wide-lane ambiguities and the variances of the floating-point estimated values are calculated. If the equivalent carrier noise corresponding to the variance of the floating-point estimated value is lower than a preset threshold, the ultra-wide-lane ambiguities are cascaded and rounded in descending order of wavelength, and the qualified ultra-wide-lane ambiguity integer results are output to ensure that the overall fixed success rate meets the system requirements. The ultra-wide-lane ambiguity integer results are written into integer linear constraints and substituted back into the triple-frequency non-combination double-difference observation model to reduce the three-dimensional narrow-lane ambiguity to be estimated to one dimension. Geometric correlation and ionosphere-free equivalent transformation are applied to the triple-frequency non-combination double-difference observation model, and the ionospheric delay is eliminated by using the integer linear constraints to obtain a simplified observation model containing only baseline displacement, tropospheric zenith wet component and a single narrow-lane ambiguity. With baseline displacement, zenith wet component and the remaining narrow-lane ambiguity as state variables, combined with the elevation angle weighted observation noise covariance matrix, an extended Kalman filter model is used to recursively obtain the floating-point solution of the narrow-lane ambiguity. The floating-point solution of the narrow-lane ambiguity is input into the LAMBDA algorithm for integer fixing of the narrow-lane ambiguity, and a fixed strategy based on ratio test is used for verification. If it passes, the integer solution is accepted and the baseline displacement is updated. The displacement rate is calculated based on the updated baseline displacement and synthesized with the hourly rainfall provided by the rain gauge according to the historical calibration weight to obtain a hazard index. When the displacement rate and the hazard index both exceed their respective thresholds, a landslide warning flag is generated, thereby realizing high-reliability fixing of the ambiguity while estimating the baseline displacement in real time, suppressing ionospheric and tropospheric errors, and giving a landslide risk determination in combination with rainfall information. Description of the Drawings

[0017] The accompanying drawings described herein are used to provide a further understanding of the present invention and form a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. Some specific embodiments of the present invention will be described in detail hereinafter with reference to the accompanying drawings in an exemplary rather than restrictive manner. The same reference numerals in the drawings denote the same or similar components or parts. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0018] Figure 1 is a schematic flowchart of a landslide prediction method based on a GNSS receiver according to an embodiment of the present invention;

[0019] Figure 2 is a schematic architecture diagram of a landslide prediction system based on a GNSS receiver according to an embodiment of the present invention;

[0020] Figure 3 is a schematic structural diagram of a receiver according to an embodiment of the present invention.

[0021] Description of reference numerals:

[0022] 1. Rainfall monitor;

[0023] 2. Mounting bracket; 20. Top crossbar;

[0024] 3. Receiver;

[0025] 4. Lightning rod;

[0026] 5. GNSS displacement monitoring station;

[0027] 6. Solar power supply system. Detailed implementation manners

[0028] In order to enable those skilled in the art of the present technology to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0029] Embodiment 1

[0030] See Figures 1 - 3 , this embodiment provides a landslide prediction method based on a GNSS receiver, including the following steps:

[0031] S101. Establish a triple-frequency non-combination double-difference observation model based on the three carrier frequency signals received by the receiver to obtain an original observation data set including baseline displacement, tropospheric delay, ionospheric delay, and three-dimensional narrow-lane ambiguity; the tropospheric delay includes the dry component model value and the zenith wet component to be estimated; combine two groups of ultra-wide-lane ambiguities from the original observation data set and calculate the floating-point estimated values and variances of the two groups of ultra-wide-lane ambiguities; if the equivalent carrier noise corresponding to the variance of the floating-point estimated value is lower than a preset threshold, perform ultra-wide-lane ambiguity cascade rounding from long to short wavelength and output the qualified ultra-wide-lane ambiguity integer result to ensure that the overall fixed success rate meets the system requirements;

[0032] S102. Write the ultra-wide-lane ambiguity integer result into integer linear constraints and substitute it back into the triple-frequency non-combination double-difference observation model to reduce the three-dimensional narrow-lane ambiguity to be estimated to one dimension; apply geometric correlation and ionosphere-free equivalent transformation to the triple-frequency non-combination double-difference observation model, and use the integer linear constraints to eliminate the ionospheric delay to obtain a simplified observation model containing only baseline displacement, tropospheric zenith wet component, and a single narrow-lane ambiguity; use the baseline displacement, zenith wet component, and remaining narrow-lane ambiguity as state variables, combine with the elevation angle weighted observation noise covariance matrix, and use the extended Kalman filter model to recursively obtain the floating-point solution of the narrow-lane ambiguity;

[0033] S103. Input the floating-point solution of the narrow-lane ambiguity into the LAMBDA algorithm for integer fixing of the narrow-lane ambiguity, and perform verification according to the fixed strategy based on ratio test. If it passes, accept the integer solution and update the baseline displacement, calculate the displacement rate based on the updated baseline displacement, and synthesize the hazard index with the hourly rainfall provided by the rain gauge according to the historical calibration weight. When the displacement rate and the hazard index both exceed their respective thresholds, generate a landslide warning flag.

[0034] It should be noted that in the prior art, for medium- and long-baseline landslide monitoring, if single-frequency or dual-frequency GNSS positioning is used, the insufficient correction of ionospheric delay and tropospheric wet component will amplify the positioning error; the success rate of ambiguity fixing will be reduced due to multipath and noise, resulting in the inability to continuously obtain centimeter-level displacements. In addition, in the prior art, multi-frequency methods often process ultra-wide lane fixing, narrow lane fixing, and rainfall triggering separately, with a long link length and poor real-time performance. In this embodiment, by using a triple-frequency non-combined double-difference observation model to retain the geometric information of satellites and receivers at the data level, and at the same time stratifying and modeling the tropospheric dry component and the zenith wet component, a basis for controllable atmospheric errors is laid; then two groups of ultra-wide lane combinations are used and variance threshold is used as a noise criterion to complete cascaded rounding, and the most easily fixed long-wavelength information is first converted into a highly credible integer prior; then the integer-week result is written into an integer linear constraint and injected into the original model together with geometric correlation and ionosphere-free equivalent transformation, realizing the synchronous completion of three objectives: ambiguity dimension reduction, ionosphere elimination, and geometry preservation, and eliminating the rank deficiency risk of traditional step-by-step operations; the extended Kalman filter jointly recurs the baseline displacement, zenith wet component, and single narrow lane ambiguity, enabling the wet delay to be updated in real time with meteorological changes; then the LAMBDA algorithm is used in conjunction with ratio test to complete narrow lane integer fixing and update the baseline; finally, the displacement rate and hourly rainfall are synthesized into a hazard index according to historical calibration weights, and early warnings are triggered when both indicators exceed the threshold, so as to realize high-reliability fixing of ambiguities while real-time estimating baseline displacements, suppressing ionospheric and tropospheric errors, and combining rainfall information to give landslide risk judgments.

[0035] In some preferred embodiments, the establishment of the triple-frequency non-combination double-difference observation model includes: respectively obtaining the original carrier phase observations and pseudo-range observations at each frequency according to the three-segment carrier frequency signals received by the receiver; performing single-difference calculations between satellites within the station on the original carrier phase observations and pseudo-range observations to eliminate the receiver clock error and obtain single-difference observations; performing inter-station single-difference calculations on the single-difference observations to further eliminate the satellite clock error and obtain double-difference observations. The double-difference observations retain the geometric distance information between the satellite and the receiver and serve as the original observation data set. It should be noted that in this embodiment, the original carrier phase observations and pseudo-range observations are extracted separately by frequency point to ensure that subsequent differential operations have a complete information basis; the intra-station single-difference operation performs differences on the observations of different satellites within the same receiver, instantaneously eliminating the receiver clock error and avoiding the oscillation noise of the receiver itself from contaminating subsequent ambiguity estimation; performing inter-station single-difference on the single-difference observations that have been declocked can synchronously cancel the satellite clock error and achieve strict time synchronization between two stations. The double-difference observations after two-level differencing still retain the geometric distance information between the satellite and the receiver, which can provide a complete geometric column vector for subsequent geometric correlation and ionospheric-free equivalent transformation, enabling the baseline displacement to directly participate in the extended Kalman filter solution. Through this embodiment, the original observation data set completely strips off the clock error noise at the input end, thereby providing high signal-to-noise ratio and geometrically complete basic data for subsequent links (ultra-wide lane rounding, ambiguity dimension reduction, ionospheric elimination), and improving the fast and reliable fixation of ambiguity and baseline accuracy from the source.

[0036] In some preferred embodiments, the tropospheric delay adopts a hierarchical modeling method: the dry component model value is calculated in advance through a standard atmospheric model and on-site measured meteorological data and remains fixed within an epoch, while the zenith wet component is estimated and dynamically updated as a random variable and then added to the extended Kalman filter model to recursively obtain the floating solution of the narrow lane ambiguity. It should be noted that in this embodiment, the dry component is calculated once using a standard atmospheric model combined with on-site meteorological data and remains constant throughout the epoch to ensure that the stable and predictable part of the troposphere is fully compensated; the rapidly changing zenith wet component is separately written as a random variable into the extended Kalman filter model, and it is estimated and dynamically updated in real time through random walk noise, enabling the model to sensitively respond to sudden increases and decreases in wet delay, thereby avoiding the systematic deviation caused by treating the wet component as a constant in traditional methods and avoiding the parameter redundancy caused by repeatedly estimating the dry component, and improving the observability of the model and the filtering convergence speed. The estimated value of the wet component output by the hierarchical modeling is recursively calculated together with the baseline displacement and the narrow lane ambiguity inside the filter, significantly reducing the impact of wet delay error on ambiguity fixation and displacement estimation, and fundamentally improving the accuracy and reliability of early landslide prediction.

[0037] In some preferred embodiments, the two groups of ultra-wide lane ambiguities select combinations with effective wavelengths greater than the original carrier wavelength to improve the noise resistance performance of the observables and reduce the ambiguity estimation error during rounding; the floating-point estimates of the two groups of ultra-wide lane ambiguities are calculated through multi-epoch smoothing averaging for noise suppression. It should be noted that in this embodiment, through the two groups of ultra-wide lane combinations with effective wavelengths greater than the original carrier wavelength, the carrier wavelength can be amplified to the meter level, significantly improving the noise resistance performance of the observables and greatly reducing the impact of the same noise level on the rounding success rate; at the same time, the two groups of combinations are linearly independent of each other, providing redundancy for cascade fixing. In addition, when calculating the floating-point estimate of the ultra-wide lane, multi-epoch smoothing averaging is introduced to expand and cancel the instantaneous observation errors in the time domain, thereby reducing the variance and increasing the ratio test passing rate. The long-wavelength design can solve the problem of noise amplification, and multi-epoch smoothing can solve the problem of instantaneous fluctuations. The superposition of the two makes the ultra-wide lane integer prior reach a high confidence level, laying a reliable foundation for subsequent narrow lane dimension reduction and integer fixing, and effectively solving the problem that the separation of ultra-wide lane and narrow lane leads to limited fixing success rate.

[0038] In some preferred embodiments, when performing cascade rounding of ultra-wide lane ambiguities, the preset threshold of the equivalent carrier noise is a threshold calibrated in advance based on historical landslide monitoring empirical data; when the variance of the floating-point estimate is higher than the preset threshold of the equivalent carrier noise, the delay smoothing strategy of the observables is automatically enabled, increasing the number of historical observation epochs and recalculating the floating-point estimate. It should be noted that the threshold is calibrated offline from long-term landslide monitoring historical data and can reflect the noise upper bound of this site under normal working conditions; once the real-time floating-point variance exceeds the limit, the delay smoothing is automatically started, increasing the number of epochs to re-estimate the floating-point value until the noise is suppressed and then the rounding is performed, so that the integer fixing is always carried out in a controllable noise environment, maintaining a high fixing success rate and avoiding false fixing caused by temporary environmental deterioration.

[0039] In some preferred embodiments, when writing the ultra-wide lane ambiguity integer result into the integer linear constraint, the integer linear constraint uses a transformation matrix that can achieve the reduction of the ambiguity dimension from three-dimensional to one-dimensional, and the transformation matrix is directly constructed from the ultra-wide lane ambiguity integer result. It should be noted that in this embodiment, the ultra-wide lane integer result is directly written into the integer linear constraint to construct a transformation matrix that can reduce the three-dimensional narrow lane ambiguity to one-dimensional, realizing the synchronous injection of dimension reduction and integer prior, reducing the state quantity of the extended Kalman filter, improving the real-time performance, significantly reducing the variance of the remaining narrow lane ambiguity, and increasing the ratio test passing rate, effectively solving the problem of separation and limited efficiency in the solution of ultra-wide lane and narrow lane.

[0040] In some preferred embodiments, the geometric-related and ionospheric-free equivalent transformation is achieved through a one-time overall linear transformation. During the transformation process, two conditions are simultaneously satisfied: completely eliminating the ionospheric delay in the observation model while retaining the geometric relationship between the satellite and the receiver. It should be noted that when the ionospheric activity is strong, if the ionospheric parameters are explicitly estimated in the observation model, new strongly correlated unknowns will be introduced; if geometrically independent combinations are used, the baseline estimation ability will be lost. In this embodiment, by specifying a one-time overall linear transformation to simultaneously meet the two conditions of geometric correlation and ionospheric elimination, the ionospheric delay is completely eliminated, and the geometric relationship between the satellite and the receiver is completely retained, avoiding rank deficiency and numerical noise amplification caused by multiple combinations, enabling the baseline displacement to be directly solved, and the accuracy not to deteriorate due to ionospheric activity, effectively solving the problem that it is difficult to have both ionospheric processing and geometric information.

[0041] In some preferred embodiments, the state variables of the extended Kalman filter model include baseline displacement, tropospheric zenith wet component, and the reduced-dimensional single narrow-lane ambiguity. The observation noise covariance matrix is determined by an adaptive weighting method related to the satellite elevation angle to adjust the weights of different satellite observables in real time. It should be noted that the landslide monitoring area is often blocked, and the multipath and low-elevation signal noises are large. If fixed weights are assigned to all satellite observations, the noise of low-quality observations will be introduced into the state solution. In this embodiment, an adaptive weighting strategy related to the satellite elevation angle is adopted in the extended Kalman filter, with high weights for high-elevation satellites and low weights for low-elevation satellites, reflecting the signal quality in real time; at the same time, the state variables only retain the baseline displacement, zenith wet component, and the reduced-dimensional single narrow-lane ambiguity, ensuring observability and computational efficiency, thereby enhancing the robustness of the filter to environmental changes and ensuring that centimeter-level displacement solutions can still be stably output under complex mountain conditions.

[0042] In some preferred embodiments, when the LAMBDA algorithm performs integer fixing of the narrow-lane ambiguity and conducts the test according to the fixed strategy based on the ratio test, it includes: comparing the residual ratio of the best integer candidate solution and the sub-optimal integer candidate solution with a preset threshold to determine whether the fixing is successful; when the ratio is higher than the preset threshold, the integer solution is confirmed as a reliable solution, otherwise, the floating-point solution recursion of the narrow-lane ambiguity in the next epoch continues. It should be noted that incorrect fixing of the narrow-lane ambiguity will directly contaminate the displacement result and trigger false alarms. In this embodiment, the LAMBDA algorithm is used in combination with the ratio test: by comparing the residual ratio of the best and sub-optimal integer candidates with the threshold to judge the fixing reliability, and the threshold is inversely deduced from the preset failure probability, which not only gives the quantitative upper bound of the risk but also automatically maintains the floating-point state recursion when the ratio is insufficient, thereby strictly controlling the incorrect fixing probability within an acceptable range. At the same time, it can quickly pass the test in the small-variance environment after the ultra-wide-lane dimension reduction, achieving both a high fixing rate and a low false alarm rate, and effectively solving the problem of insufficient ambiguity fixing rate.

[0043] In some preferred embodiments, the displacement rate after the baseline displacement update is calculated by dividing the displacement change by the adjacent epoch time interval, and further processed by multi-epoch sliding window averaging to eliminate the random fluctuations in the calculation results of a single epoch. It should be noted that the displacement rate is an important dynamic indicator for landslide triggering, but single-epoch differencing is vulnerable to occasional noise. In this embodiment, after calculating the displacement rate, multi-epoch sliding window averaging is introduced to smooth out the random spikes and retain the true deformation trend, thereby reducing the false alarm rate and avoiding misjudgment of the risk index due to instantaneous observation anomalies. At the same time, the window length can be adaptively set according to the sampling rate, which not only maintains timeliness but also improves stability, meeting the requirements of early warning for continuous and reliable speed thresholds.

[0044] In some preferred embodiments, the hourly rainfall data of the rain gauge is sent to the receiver in real time through a wired or wireless communication link. The receiver is equipped with an abnormal data detection module, and the abnormal data detection module automatically eliminates or interpolates and supplements the significantly abnormal rainfall data. It should be noted that the rainfall data is an important component of the risk index, but rainfall calculation is vulnerable to faults or communication interference. In this embodiment, an abnormal data detection module is configured at the receiver end to automatically eliminate or interpolate and complete the significantly abnormal or missing rainfall data, ensuring the continuity and reasonableness of the rainfall sequence, thereby avoiding misreporting caused by extreme error data amplifying the risk index, and also ensuring that the risk assessment does not interrupt during short-term communication interruptions, effectively solving the problem of lack of quality control in the rainfall trigger link.

[0045] In some preferred embodiments, in the synthesis of the risk index, the historical calibration weight is an empirical weight obtained through statistical regression analysis of long-term monitoring data. When the long-term historical data of the monitoring point changes, the empirical weight is automatically adjusted to adapt to the change of the site conditions and reflect the true landslide risk. It should be noted that the landslide triggering mechanism may change over time due to changes in vegetation, drainage engineering, etc., and the fixed weight will gradually become inaccurate. In this embodiment, the historical calibration weight is automatically updated through statistical regression of long-term monitoring data, making the risk index dynamically fit the actual situation of the site. When the long-term characteristics of the monitoring point change, the weight can be adaptively adjusted to avoid risk assessment deviation caused by model aging and ensure the accuracy of early warning.

[0046] In some preferred embodiments, when the displacement rate and the risk index both exceed their respective thresholds, the generated landslide warning sign is uploaded to the remote cloud platform through the GNSS displacement monitoring station connected and communicating with the receiver. The remote cloud platform has a multi-level warning push function, and the multi-level warning push function pushes the warning message to different levels of management terminals according to the degree of danger. It should be noted that the warning information needs to be transmitted to the appropriate level within the shortest time. In this embodiment, after the displacement rate and the risk index both exceed the limit, the warning sign is uploaded to the cloud platform through the GNSS displacement monitoring station, and the cloud platform classifies and pushes it to different management terminals according to the risk level, so as to avoid information overflow, ensure that key decision-makers receive high-risk alarms in the first time, and at the same time, general management terminals only receive necessary information to improve the response efficiency.

[0047] Embodiment 2

[0048] See Figures 1 - 3 , this embodiment provides a landslide prediction system based on a GNSS receiver, including:

[0049] A rainfall monitor 1 is arranged at one end of the top crossbar 20 of the installation bracket 2, and is used to monitor the hourly rainfall at the landslide monitoring position;

[0050] A receiver 3 is arranged in the middle of the top crossbar, is connected and communicates with the rainfall monitor, and runs the above-mentioned landslide prediction method based on the GNSS receiver;

[0051] A lightning rod 4 is arranged at the other end of the top crossbar and is higher than the rainfall monitor and the receiver;

[0052] A GNSS displacement monitoring station 5 is arranged in the middle of the installation bracket and is connected and communicates with the receiver;

[0053] A solar power supply system 6 is arranged in the middle of the installation bracket and forms a rain shield above the GNSS displacement monitoring station, and is used to supply power to the GNSS displacement monitoring station, the receiver and the rainfall monitor.

[0054] It should be noted that the rainfall monitor is installed at one end of the top crossbar, which is at the highest point of the bracket and far from the metal antenna, so as to avoid shielding and splash error; the hourly rainfall is collected in real time as an external factor triggering landslides, providing reliable input for the risk index, and can solve the problem of false alarms in the traditional single deformation criterion. The lightning rod is placed at the other end of the crossbar and is higher than the rainfall monitor and the receiver, which can provide a lightning drainage channel for the whole machine, reduce the damage risk of direct lightning and induced lightning to the receiver circuit, and ensure that the monitoring link is not interrupted during heavy rain. The GNSS displacement monitoring station is arranged at the low center-of-gravity position in the middle of the bracket, reducing the interference of wind load and vibration on centimeter-level displacement solution; at the same time, the GNSS displacement monitoring station can communicate bidirectionally with the receiver, upload the warning sign generated by the receiver to the cloud through the cellular or NB-IoT network, and realize remote multi-station collaboration. The solar power supply system is installed above the monitoring station to form a rain shield, which not only ensures the battery life on rainy days, but also enables the rainfall sensor, the receiver, and the displacement monitoring station to share an independent power supply circuit.

[0055] It also needs to be noted that when the receiver runs the above-mentioned landslide prediction method based on the GNSS receiver, the geometric information of the satellite and the receiver is retained at the data level by using the triple-frequency non-combination double-difference observation model, and at the same time, the tropospheric dry component and the zenith wet component are modeled layer by layer, laying the foundation for controllable atmospheric error; then two groups of ultra-wide lane combinations are used and the variance threshold is used as the noise criterion to complete cascade rounding, and the longest wavelength information that is most easily fixed is first converted into a highly credible integer prior; then the integer result is written into the integer linear constraint and injected into the original model together with the geometric correlation and the ionospheric-free equivalent transformation, realizing the synchronous completion of three objectives: ambiguity dimension reduction, ionospheric elimination, and geometric preservation, and eliminating the rank deficiency risk of traditional step-by-step operations; the extended Kalman filter jointly recurs the baseline displacement, the zenith wet component, and the single narrow lane ambiguity, so that the wet delay can be updated in real time with the change of meteorology; then the LAMBDA algorithm is used in combination with the ratio test to complete the narrow lane integer fixation and update the baseline; finally, the displacement rate and the hourly rainfall are synthesized into a risk index according to the historical calibration weight, and the double-index simultaneous over-threshold triggers a warning, so as to realize high-reliability fixation of the ambiguity while estimating the baseline displacement in real time, suppressing ionospheric and tropospheric errors, and giving a landslide risk determination in combination with rainfall information.

[0056] It should be pointed out that the above embodiments are only preferred specific embodiments of the present invention, and the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention. The protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A landslide prediction method based on a GNSS receiver, characterized in that, Including: Establish a triple-frequency non-combination double-difference observation model based on the three carrier frequency signals received by the receiver to obtain an original observation data set including baseline displacement, tropospheric delay, ionospheric delay, and three-dimensional narrow-lane ambiguity; the tropospheric delay includes the dry component model value and the zenith wet component to be estimated; use the original observation data set to combine two sets of ultra-wide-lane ambiguities and calculate the floating-point estimated values of the two sets of ultra-wide-lane ambiguities and the variances of the floating-point estimated values; if the equivalent carrier noise corresponding to the variance of the floating-point estimated value is lower than the preset threshold, perform ultra-wide-lane ambiguity cascade rounding from long to short wavelength and output the qualified ultra-wide-lane ambiguity integer result to ensure that the overall fixed success rate meets the system requirements; Write the ultra-wide-lane ambiguity integer result into the integer linear constraint and substitute it back into the triple-frequency non-combination double-difference observation model to reduce the three-dimensional narrow-lane ambiguity to be estimated to one dimension; Apply geometric correlation and ionosphere-free equivalent transformation to the triple-frequency non-combination double-difference observation model, and use the integer linear constraint to eliminate the ionospheric delay to obtain a simplified observation model containing only baseline displacement, tropospheric zenith wet component, and a single narrow-lane ambiguity; take baseline displacement, zenith wet component, and the remaining narrow-lane ambiguity as state variables, combine with the altitude angle weighted observation noise covariance matrix, and use the extended Kalman filter model to recursively obtain the floating-point solution of the narrow-lane ambiguity; Input the floating-point solution of the narrow-lane ambiguity into the LAMBDA algorithm for integer fixing of the narrow-lane ambiguity, and perform verification according to the fixed strategy based on ratio test. If it passes, accept the integer solution and update the baseline displacement, calculate the displacement rate based on the updated baseline displacement, and synthesize the hazard index with the hourly rainfall provided by the rain gauge according to the historical calibration weight. When the displacement rate and the hazard index both exceed their respective thresholds, generate a landslide warning sign; When the LAMBDA algorithm performs integer fixing of the narrow-lane ambiguity and performs verification according to the fixed strategy based on ratio test, it includes: judging whether the fixing is successful by comparing the residual ratio of the optimal integer candidate solution and the sub-optimal integer candidate solution with the preset threshold; when the ratio is higher than the preset threshold, confirm that the integer solution is a reliable solution, otherwise continue to recursively calculate the floating-point solution of the narrow-lane ambiguity in the next epoch; The calculation of the displacement rate after updating the baseline displacement is carried out by dividing the displacement change amount by the time interval between adjacent epochs, and further processed by multi-epoch sliding window averaging to eliminate the random fluctuation of the calculation result of a single epoch.

2. The landslide prediction method according to claim 1, wherein The establishment of the triple-frequency non-combination double-difference observation model includes: respectively obtaining the original carrier phase observations and pseudorange observations at each frequency according to the three carrier frequency signals received by the receiver; performing single-difference calculation between satellites within the station on the original carrier phase observations and pseudorange observations to eliminate the receiver clock error and obtain single-difference observations; performing inter-station single-difference calculation on the single-difference observations to further eliminate the satellite clock error and obtain double-difference observations, and the double-difference observations retain the geometric distance information between the satellite and the receiver and are used as the original observation data set.

3. The landslide prediction method according to claim 1, wherein The tropospheric delay is modeled in a layered manner: the dry component model value is calculated in advance through a standard atmospheric model and on-site measured meteorological data, and remains fixed within an epoch. The zenith wet component, as a random variable, is estimated in real time and dynamically updated and then added to the extended Kalman filter model to recursively obtain the float solution of the narrow-lane ambiguity.

4. The landslide prediction method according to claim 1, characterized in that, The two groups of ultra-wide-lane ambiguities select combinations with effective wavelengths greater than the original carrier wavelength to improve the noise resistance performance of the observables and reduce the ambiguity estimation error during rounding; the float estimates of the two groups of ultra-wide-lane ambiguities are calculated through multi-epoch smoothing averaging to suppress noise.

5. The landslide prediction method according to claim 1, characterized in that, When performing cascaded rounding of the ultra-wide-lane ambiguity, the preset threshold of the equivalent carrier noise is a threshold calibrated in advance based on historical landslide monitoring empirical data; when the variance of the float estimate is higher than the preset threshold of the equivalent carrier noise, the delay smoothing strategy of the observables is automatically enabled, increasing the number of historical observation epochs and recalculating the float estimate.

6. The landslide prediction method according to claim 1, wherein When writing the integer result of the ultra-wide-lane ambiguity into the integer linear constraint, the integer linear constraint uses a transformation matrix that can reduce the ambiguity dimension from three-dimensional to one-dimensional, and the transformation matrix is directly constructed from the integer result of the ultra-wide-lane ambiguity.

7. The landslide prediction method according to claim 1, wherein The geometric correlation and ionosphere-free equivalence transformation are realized through a one-time overall linear transformation, and two conditions are simultaneously satisfied during the transformation process: completely eliminating the ionospheric delay in the observation model and retaining the geometric relationship between the satellite and the receiver.

8. The landslide prediction method according to claim 1, characterized in that, The state variables of the extended Kalman filter model include baseline displacement, tropospheric zenith wet component, and the reduced-dimensional single narrow-lane ambiguity, and the observation noise covariance matrix is determined by an adaptive weighting method related to the satellite elevation angle to adjust the weights of different satellite observables in real time.

9. The landslide prediction method according to claim 1, wherein When the LAMBDA algorithm performs integer fixing of the narrow-lane ambiguity and conducts the test according to the fixed strategy based on the ratio test, it includes: comparing the residual ratio of the optimal integer candidate solution and the sub-optimal integer candidate solution with a preset threshold to determine whether the fixing is successful; when the ratio is higher than the preset threshold, the integer solution is confirmed as a reliable solution, otherwise, continue to recursively calculate the float solution of the narrow-lane ambiguity in the next epoch.

10. A landslide prediction system based on a GNSS receiver, characterized in that, Including: A rain gauge, arranged at one end of the top crossbar of the installation bracket, for monitoring the hourly rainfall at the landslide monitoring location; A receiver, arranged in the middle of the top crossbar, connected and communicating with the rain gauge, and running the landslide prediction method based on a GNSS receiver according to any one of claims 1 - 9; A lightning rod, arranged at the other end of the top crossbar and higher than the rain gauge and the receiver; A GNSS displacement monitoring station, arranged in the middle of the installation bracket, connected and communicating with the receiver; A solar power supply system, arranged in the middle of the installation bracket and above the GNSS displacement monitoring station to form a rain shield, for supplying power to the GNSS displacement monitoring station, the receiver, and the rain gauge.

Images