Landslide monitoring sensor signal calibration method and system

Abstract

The invention relates to the technical field of landslide monitoring, and particularly discloses a landslide monitoring sensor signal calibration method and system, and the method comprises the steps: continuously obtaining the monitoring signals of a plurality of sensors; analyzing and acquiring the statistical characteristics of each sensor, the change consistency with the physically associated sensor and the deviation degree with the historical signal according to the monitoring signal; the credibility score of each sensor is calculated and obtained according to the statistical characteristics, the change consistency and the deviation degree; when the credibility score is lower than a preset score threshold value, analyzing to obtain an error mode; extracting a calibration scheme from a preset calibration scheme library according to the error mode; calibrating a monitoring signal of the corresponding sensor according to the calibration scheme; according to the method, automatic and data-driven calibration of landslide monitoring sensor signals is realized, the limitation of traditional manual calibration in a field environment is overcome, and the accuracy of monitoring data and the reliability of the system are improved.

Description

Technical Field

[0001] The present application relates to the technical field of landslide monitoring, and in particular to a landslide monitoring sensor signal calibration method and system. Background Art

[0002] Landslide monitoring systems are widely deployed in potential landslide areas. Using sensors such as surface displacement, deep displacement, pore water pressure, and rainfall, they continuously collect geological and environmental parameter data, analyze landslide stability, and issue early warnings to protect downstream residents and infrastructure. With technological advancements, these systems are developing towards greater automation, intelligence, and real-time capabilities, improving data collection efficiency and early warning reliability, adapting to the needs of long-term monitoring in complex field environments.

[0003] However, long-term exposure of sensors to harsh field environments can lead to signal errors, such as zero drift and sensitivity loss. This can cause sensor readings to deviate from the actual physical quantity, potentially leading to false alarms or missed alarms. Traditional calibration methods rely on manual labor or external equipment, making them impractical and costly in field scenarios. Furthermore, the diverse error modes, such as zero drift and sensitivity loss, require accurate diagnosis to apply the appropriate calibration solution.

[0004] There is currently no effective technical solution to the above problems. Summary of the Invention

[0005] The purpose of this application is to provide a landslide monitoring sensor signal calibration method and system to achieve automated, data-driven calibration of landslide monitoring sensor signals.

[0006] In a first aspect, the present application provides a landslide monitoring sensor signal calibration method, which is applied to a sensor in a landslide status monitoring system, and the method comprises the following steps: S1. Continuously obtain monitoring signals from multiple sensors; S2. Analyzing the monitoring signals to obtain statistical characteristics of each sensor, consistency of changes with physically associated sensors, and a degree of deviation from historical signals; S3. Calculate and obtain a credibility score for each sensor based on the statistical characteristics, change consistency, and deviation degree; S4. When the credibility score is lower than a preset score threshold, obtain an error pattern based on one or more of the statistical characteristics, the change consistency, and the deviation degree; S5. extracting a calibration solution from a preset calibration solution library according to the error pattern; S6. Calibrate the monitoring signal of the corresponding sensor according to the calibration scheme.

[0007] The method of the present application realizes the automated, data-driven calibration of landslide monitoring sensor signals, overcomes the limitations of traditional manual calibration in field environments, improves the accuracy of monitoring data and the reliability of the system, reduces the demand and cost of manual calibration, and enhances the early warning capability of the landslide monitoring system.

[0008] The landslide monitoring sensor signal calibration method, wherein step S2 comprises: S21. Calculate the mean, variance, and fluctuation range of each sensor according to the monitoring signal as statistical characteristics; S22. Calculate, based on the sensor pairs having physical correlation, a Pearson correlation coefficient between the sensor pairs as the change consistency of the corresponding sensors; S23. Obtain the degree of deviation based on the difference between the historical signal of the same type of sensor under the same environmental conditions or the same time conditions and the monitoring signal.

[0009] The above steps provide a specific and operational method to quantify the statistical characteristics, change consistency and deviation degree of landslide monitoring sensor signals, improve the accuracy of information acquisition, and provide a reliable data basis for subsequent credibility assessment and error pattern recognition.

[0010] The landslide monitoring sensor signal calibration method, wherein step S23 includes: S231. Extract historical signals of similar sensors under the same environmental conditions or the same time conditions, and calculate the mean and variance of the historical signals; S232. Calculate the mean difference and variance difference between the historical signal and the monitoring signal based on the statistical characteristics; S233. Calculate the degree of deviation according to the mean difference and the variance difference.

[0011] The landslide monitoring sensor signal calibration method, wherein step S3 comprises: S31, obtaining the activity level of the landslide at the last moment; S32, extracting a weight group from a preset weight database according to the activity level and the type of sensor; S33. Calculate the statistical characteristics, change consistency, and deviation degree according to the weight group to obtain a credibility score.

[0012] In the landslide monitoring sensor signal calibration method, the process of determining the landslide activity level includes: A1. Extract the displacement, rainfall, and groundwater level change of the corresponding landslide within the preset time window; A2. According to the displacement of the landslide, the rainfall, and the groundwater level change, a preset activity level discrimination rule library is queried to obtain the activity level.

[0013] The landslide monitoring sensor signal calibration method, wherein step S4 comprises: S41. When the credibility score is lower than a preset score threshold, extracting a first deviation state feature, a second deviation state feature, and a third deviation state feature according to the statistical characteristics, change consistency, and deviation degree; S42: Concatenate the first deviation state feature, the second deviation state feature, and the third deviation state feature to obtain a fused deviation feature vector; S43: Input the fused deviation feature vector into a preset error pattern recognition model to generate the error pattern.

[0014] In the landslide monitoring sensor signal calibration method, the error modes include: zero drift, sensitivity attenuation, intermittent failure and periodic interference.

[0015] The landslide monitoring sensor signal calibration method, wherein step S5 comprises: S51 , extracting a corresponding calibration solution from a preset calibration solution library according to the landslide activity level at the last moment, the error pattern, and the type of the corresponding sensor.

[0016] In the landslide monitoring sensor signal calibration method, the calibration scheme includes calibration means and calibration parameter types, and step S6 includes: S61. Based on the calibration parameter type, determine the calibration parameter according to the monitoring signal of the sensor to be calibrated and the historical signal of the same type of sensor under the same environmental conditions or the same time conditions; S62: Update the configuration information of the corresponding sensor based on the calibration means and the calibration parameters to calibrate the monitoring signal of the corresponding sensor.

[0017] In a second aspect, the present application further provides a landslide monitoring sensor signal calibration system, which is applied to sensors in a landslide status monitoring system. The system comprises: An acquisition module, used to continuously acquire monitoring signals from multiple sensors; A characteristic information acquisition module is used to analyze and obtain the statistical characteristics of each sensor, the consistency of changes in the physically associated sensor, and the degree of deviation from historical signals based on the monitoring signal; A scoring module, configured to calculate and obtain a credibility score for each sensor based on the statistical characteristics, change consistency, and deviation degree; an error analysis module, configured to obtain an error pattern based on one or more of the statistical characteristics, the change consistency, and the deviation degree when the credibility score is lower than a preset score threshold; A calibration scheme determination module, configured to extract a calibration scheme from a preset calibration scheme library according to the error pattern; The calibration module is used to calibrate the monitoring signal of the corresponding sensor according to the calibration scheme.

[0018] The system of the present application realizes the automated and data-driven calibration of landslide monitoring sensor signals, overcomes the limitations of traditional manual calibration in field environments, improves the accuracy of monitoring data and the reliability of the system, reduces the demand and cost of manual calibration, and enhances the early warning capability of the landslide monitoring system.

[0019] As can be seen from the above, the present application provides a landslide monitoring sensor signal calibration method and system, wherein the method of the present application calculates the signal credibility score by continuously analyzing the statistical characteristics of the sensor signal, the consistency of changes in the physically associated sensor, and the degree of deviation from the historical signal. When the score is lower than the threshold, the specific error pattern is diagnosed based on these analysis results, thereby extracting and applying the corresponding calibration scheme from the preset calibration scheme library, realizing the automated and data-driven calibration of the landslide monitoring sensor signal, overcoming the limitations of traditional manual calibration in the field environment, improving the accuracy of the monitoring data and the reliability of the system, reducing the demand and cost of manual calibration, and enhancing the early warning capability of the landslide monitoring system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of the landslide monitoring sensor signal calibration method provided in an embodiment of the present application.

[0021] Figure 2 This is a schematic diagram of the structure of the landslide monitoring sensor signal calibration system provided in an embodiment of the present application.

[0022] Reference numerals: 201, acquisition module; 202, characteristic information acquisition module; 203, scoring module; 204, error analysis module; 205, calibration scheme determination module; 206, calibration module. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work fall within the scope of protection of the present application.

[0024] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0025] First, please refer to Figure 1 Some embodiments of the present application provide a landslide monitoring sensor signal calibration method, which is applied to a sensor in a landslide status monitoring system. The method includes the following steps: S1. Continuously obtain monitoring signals from multiple sensors; S2. Analyze the monitoring signals to obtain the statistical characteristics of each sensor, the consistency of changes in the physically associated sensors, and the degree of deviation from historical signals; S3. Calculate and obtain the credibility score of each sensor based on statistical characteristics, change consistency, and deviation degree; S4. when the credibility score is lower than a preset score threshold, analyzing and obtaining an error pattern based on one or more of statistical characteristics, consistency of change, and degree of deviation; S5. extracting a calibration solution from a preset calibration solution library according to the error pattern; S6. Calibrate the monitoring signal of the corresponding sensor according to the calibration scheme.

[0026] Specifically, continuously acquiring monitoring signals from multiple sensors refers to uninterruptedly receiving real-time or quasi-real-time data streams from various monitoring devices such as surface displacement sensors, deep displacement sensors, pore water pressure sensors and rainfall sensors. This can be achieved by wired transmission, wireless transmission or regular collection by data recorders, for example, transmitting data through a fiber optic network or a wireless network.

[0027] More specifically, statistical characteristics refer to the intrinsic numerical distribution characteristics of the sensor monitoring signal over a period of time. They can be achieved by calculating the mean, variance, standard deviation, fluctuation range, skewness, kurtosis, etc. For example, calculating the average value of the signal and the degree of dispersion of the signal value, which is mainly used to reflect the stability, dispersion or distribution form of the signal itself.

[0028] More specifically, change consistency refers to the degree of synchronization or correlation between the signal change trend of a sensor and the signal change trend of other physically associated sensors. It can be achieved by calculating the Pearson correlation coefficient, Spearman rank correlation coefficient, cross-correlation function, etc. For example, calculating the correlation between surface displacement and deep displacement signals is mainly used to determine whether the sensor signal is coordinated with other signals that should change synchronously.

[0029] More specifically, the degree of deviation refers to the degree of difference between the sensor monitoring signal and the historical signal or expected pattern of the sensor under similar environmental or time conditions. It can be achieved by calculating the mean difference, variance difference, root mean square error between the current signal and the historical signal, or by a difference metric based on pattern matching. For example, comparing the statistical distribution difference between the current rainfall signal and the historical rainfall signal for the same period is mainly used to detect whether the signal deviates from its normal behavior range.

[0030] More specifically, the credibility score refers to a numerical value that quantitatively evaluates the reliability of sensor monitoring signals. It can be achieved by weighted summation based on statistical characteristics, consistency of change and degree of deviation, fuzzy logic reasoning or machine learning model output. For example, the various indicators obtained from the analysis are mapped to a score from 0 to 100 through a function. It is mainly used to comprehensively judge whether the sensor signal is reliable and serve as a basis for whether calibration is needed.

[0031] More specifically, the error mode refers to the specific error type or cause that causes the sensor signal to be unreliable, which can include zero drift, sensitivity attenuation, intermittent failure, periodic interference, data jump, data stickiness and other types. It is mainly used to diagnose the root cause of the signal abnormality in order to select a targeted calibration method.

[0032] More specifically, the calibration scheme library refers to a pre-stored set of signal correction methods designed for different error modes and sensor types. It can be implemented using a database table, a lookup table, or a software structure containing different calibration algorithm modules. For example, it stores a bias addition algorithm for zero point drift and a multiplication coefficient algorithm for sensitivity attenuation. It is mainly used to provide a calibration method that matches the diagnosed error mode.

[0033] The method of this application first continuously collects signals from various types of landslide monitoring sensors. These signals are then subjected to a multi-dimensional analysis to extract statistical properties reflecting the signal's inherent characteristics, physical correlation consistency reflecting the relationship between the signals, and the degree of deviation from normal behavior. Based on these analysis results, a credibility score is calculated for each sensor signal, which comprehensively reflects the signal's reliability. When the credibility score falls below a preset threshold, it indicates that the signal may contain errors. The method then further utilizes information such as statistical properties, variation consistency, and degree of deviation obtained from the previous analysis to identify the specific error pattern of the signal through a specific diagnostic process, such as zero drift or sensitivity degradation. Once the error pattern is determined, the method of this application searches and extracts the corresponding calibration scheme from a preset calibration scheme library based on the error pattern. Finally, based on the extracted calibration scheme, the sensor with the error is calibrated to correct the signal value, so that the monitoring signal output by the sensor is closer to the actual physical quantity. This entire process forms a closed loop, achieving automatic detection, diagnosis, and correction of sensor signal errors.

[0034] The method of the present application continuously analyzes the statistical characteristics of sensor signals, the consistency of changes in physically associated sensors, and the degree of deviation from historical signals, calculates the credibility score of the signal, and diagnoses the specific error pattern based on these analysis results when the score is lower than a threshold, thereby extracting and applying the corresponding calibration scheme from a preset calibration scheme library, thereby realizing automated, data-driven calibration of landslide monitoring sensor signals, overcoming the limitations of traditional manual calibration in field environments, improving the accuracy of monitoring data and the reliability of the system, reducing the demand and cost of manual calibration, and enhancing the early warning capability of the landslide monitoring system.

[0035] In some preferred embodiments, step S2 includes: S21. Calculate the mean, variance, and fluctuation range of each sensor according to the monitoring signal as statistical characteristics; S22. Calculate the Pearson correlation coefficient between the sensor pairs based on the sensor pairs having physical correlation, as the change consistency of the corresponding sensors; S23. Obtain the degree of deviation based on the difference between the historical signal and the monitoring signal of the same type of sensor under the same environmental conditions or the same time conditions.

[0036] Specifically, a sensor pair with physical correlation refers to a sensor combination in which the physical quantities monitored by it are intrinsically linked or mutually influenced in a landslide monitoring scenario, such as displacement sensors at different depths, rainfall sensors, and pore water pressure sensors. They can be identified using pre-set sensor association rules or correlation thresholds determined based on historical data analysis.

[0037] More specifically, the same type of sensors refers to sensors that monitor the same type of physical quantity, for example, a plurality of surface displacement sensors or a plurality of pore water pressure sensors.

[0038] More specifically, the same environmental conditions or time conditions refer to situations in which the historical data contains similar external environmental factors (such as rainfall and temperature) or falls within a similar time period (e.g., the same season or time point) as when the current monitoring signal was collected. This can be achieved by associating environmental parameters or time indexes with the historical monitoring data. The difference between the historical signal and the monitoring signal refers to the degree of deviation in numerical or statistical characteristics between the current monitoring signal and the historical signal collected under similar conditions. This can be quantified using various mathematical or statistical methods.

[0039] More specifically, step S21 calculates the mean, variance, and fluctuation range of each sensor's monitoring signal, quantifying the signal's central tendency, dispersion, and amplitude of variation, thereby obtaining the sensor's statistical characteristics. These statistics provide a quantitative description of the fundamental behavior of individual sensor signals. Next, step S22 identifies sensor pairs with physical correlations and calculates the Pearson correlation coefficient between the monitoring signals of these pairs. The Pearson correlation coefficient reflects the degree of linear correlation between two signals, thereby quantifying the consistency of variation between sensors and helping to determine whether the sensor signals conform to the expected physical linkage relationship. Finally, step S23 quantifies the degree of signal deviation by comparing the current monitoring signal with historical signals from similar sensors under similar environmental or temporal conditions. This comparison accounts for the influence of external factors and temporal patterns, making the deviation assessment more accurate. Through these three-dimensional analysis, this solution comprehensively and quantitatively obtains the core information required to assess the reliability of sensor signals, providing a reliable data foundation for subsequent credibility assessment and error pattern identification.

[0040] The above steps provide a specific and operational method to quantify the statistical characteristics, change consistency and deviation degree of landslide monitoring sensor signals, improve the accuracy of information acquisition, and provide a reliable data basis for subsequent credibility assessment and error pattern recognition.

[0041] In some preferred embodiments, step S23 includes: S231. Extract historical signals of similar sensors under the same environmental conditions or the same time conditions, and calculate the mean and variance of the historical signals; S232. Calculate the mean difference and variance difference between the historical signal and the monitoring signal based on statistical characteristics; S233. Calculate the degree of deviation based on the mean difference and the variance difference.

[0042] Specifically, the mean difference refers to the difference between the mean of the monitoring signal and the mean of the historical signal. The variance difference refers to the difference between the variance of the monitoring signal and the variance of the historical signal.

[0043] More specifically, in step S233 , the degree of deviation is positively correlated with the mean difference and the variance difference.

[0044] More specifically, the mean difference reflects the shift in the overall signal level, while the variance difference reflects changes in signal volatility. By calculating these differences, abstract "differences" are converted into concrete numerical values, laying the foundation for subsequent analysis. In this way, the differences in statistical characteristics are combined into a quantitative deviation indicator, providing a more reliable input for sensor credibility scoring and helping to improve the accuracy of sensor signal calibration.

[0045] In some preferred embodiments, in step S233, the degree of deviation is preferably calculated as follows: Deviation degree = α·|(μ c -μ h ) / μ h |+β·|(σ c ²-σ h ²) / σ h ²| (1) Among them, α and β are preset weight coefficients, which can be set according to the use requirements, and are preferably 0.5; μ c 、μ h、 σ c ² and σ h ² are the mean of the monitoring signal, the mean of the historical signal, the variance of the monitoring signal and the variance of the historical signal respectively; where μ h and σ h ² is not 0.

[0046] In some preferred embodiments, step S3 includes: S31, obtaining the activity level of the landslide at the last moment; S32, extracting a weight group from a preset weight database according to the activity level and the type of sensor; S33. Calculate the credibility score based on the statistical characteristics, change consistency and deviation degree of the weight group.

[0047] Specifically, the landslide activity level refers to a classification indicator that reflects the current state of landslide activity. The preset weight database stores a set of weight parameters used to calculate the credibility score. This database pre-configures weight values ​​for different landslide activity levels and different sensor types. A weight set refers to a set of weight values ​​extracted from the weight database that is applicable to a specific landslide activity level and a specific sensor type. This set includes weights for weighting statistical characteristics, consistency of change, and degree of deviation.

[0048] More specifically, the above steps provide a more refined and adaptive method for calculating sensor signal credibility scores. By considering the landslide activity level and sensor type, the weights of evaluation indicators are dynamically adjusted, thereby improving the accuracy of credibility assessment and providing a more reliable foundation for subsequent error pattern recognition and calibration. First, the landslide activity level at the previous moment is obtained. This is to use the previous, reliable landslide state as a state reference. The normal fluctuation range and abnormal performance characteristics of sensor signals vary at different activity levels, so the evaluation criteria need to be adjusted according to the current state of the landslide. Next, based on the obtained activity level and the type of sensor to be evaluated, the corresponding weight set is extracted from a pre-set weight database. This weight database pre-stores the relative importance (i.e., weights) of statistical characteristics, change consistency, and deviation degree in credibility assessment for different sensor types at different landslide activity levels. By extracting weights based on activity level and sensor type, the weights used to calculate the credibility score are optimal for the current context. For example, during periods of landslide activity, the consistency of a displacement sensor's changes may be a more reliable indicator of its credibility than the degree of deviation from historical signals, and in this case, the consistency of change will be given a higher weight. The introduction of these dynamic weights eliminates the static nature of the credibility assessment and allows it to adapt to changes in landslide conditions and sensor type. Finally, based on the extracted weights, the previously acquired statistical characteristics, consistency of change, and degree of deviation are weighted together to produce the final credibility score. Because the weights are dynamically determined based on the landslide activity level and sensor type, the calculated credibility score more accurately reflects the true reliability of the current sensor signal, effectively distinguishing between true geological changes and sensor errors, and providing a more reliable basis for subsequent calibration decisions.

[0049] The method of the present application can dynamically adjust the credibility score calculation process according to the current activity level of the landslide and the type of sensor, so that the credibility assessment process can fully utilize the multi-dimensional signal characteristics and adapt to the complex monitoring environment and landslide state changes by dynamically adjusting the weights, thereby improving the accuracy and adaptability of the credibility assessment, thereby more effectively identifying errors in sensor signals and providing more reliable input for subsequent error pattern recognition and calibration.

[0050] In some preferred embodiments, the process of determining the activity level of a landslide includes: A1. Extract the displacement, rainfall, and groundwater level change of the corresponding landslide within the preset time window; A2. Query the preset activity level discrimination rule library to obtain the activity level based on the landslide displacement, rainfall, and groundwater level change.

[0051] Specifically, the preset time window refers to a continuous time period used to analyze the recent activity status of the landslide, and its length can be set according to the actual monitoring needs and landslide characteristics. The displacement refers to the overall or local movement distance of the landslide body within the preset time window, which can be obtained using data from surface displacement sensors or deep displacement sensors. The rainfall refers to the total amount of rainfall accumulated in the landslide area within the preset time window, which can be obtained using data from rain sensors. The groundwater level change amplitude refers to the maximum change range of the groundwater level in the landslide area within the preset time window, which can be obtained using data from pore water pressure sensors or groundwater level meters. The activity level discrimination rule base refers to a series of rules, thresholds or models used to determine the activity level of the landslide, which can be implemented by using a rule set based on expert experience, a decision tree model trained based on historical monitoring data, or a lookup table. The activity level refers to a classification representation of the current activity intensity of the landslide, preferably divided into three levels: stable, slow-changing, and active, which represent the state of the landslide body being basically inactive, slowly deforming, or rapidly deforming, respectively.

[0052] Specifically, the determination process first extracts the displacement, rainfall and groundwater level change amplitude data related to the landslide to be evaluated within a preset time window from the monitoring system in step A1. These physical quantities are key factors affecting the stability and activity of the landslide, and can directly or indirectly reflect the current state and activity trend of the landslide. By limiting the extraction to a preset time window, it is ensured that the analyzed data can reflect the recent behavior of the landslide, so that the determined activity level is timely and relevant. Then, in step A2, the displacement, rainfall and groundwater level change amplitude obtained in step A1 are used as input to query the preset activity level discrimination rule library. The preset rules or models in the rule library determine the current activity level of the landslide based on these input parameters.

[0053] More specifically, the method of this application achieves objective classification of landslide states by querying a rule base, providing a clear basis for subsequently adjusting sensor credibility assessment weights based on landslide activity levels. This process provides the necessary input for selecting weights based on activity levels, enabling sensor credibility assessment to better adapt to the characteristics of landslides in different activity states, thereby improving assessment accuracy.

[0054] In some preferred embodiments, step S4 includes: S41. When the credibility score is lower than a preset score threshold, extracting a first deviation state feature, a second deviation state feature, and a third deviation state feature according to statistical characteristics, change consistency, and deviation degree; S42, concatenating the first deviation state feature, the second deviation state feature, and the third deviation state feature to obtain a fused deviation feature vector; S43: Input the fused deviation feature vector into a preset error pattern recognition model to generate an error pattern.

[0055] Specifically, the first deviation state feature is a quantitative indicator extracted based on the statistical characteristics of the sensor monitoring signal (such as mean, variance, and fluctuation range) that can reflect the abnormal state of the signal itself. The second deviation state feature is a quantitative indicator extracted based on the consistency of changes between the sensor signal to be diagnosed and the signals of physically associated sensors (such as the Pearson correlation coefficient) that can reflect abnormal coordination between sensors. The third deviation state feature is a quantitative indicator extracted based on the degree of deviation between the sensor signal to be diagnosed and historical signals of similar sensors that can reflect abnormalities in the signal's normal historical behavior. The fused deviation feature vector is used to comprehensively describe the sensor's abnormal state.

[0056] More specifically, the error pattern recognition model is preferably a classification model constructed based on a multi-classifier of a support vector machine. The model can receive a fused deviation feature vector as input and output a corresponding sensor error pattern category. It can adopt a one-vs-Rest or one-vs-One strategy to construct a multi-classifier. The construction and training method of the model can adopt the construction and training method of the classifier, which will not be elaborated here.

[0057] Specifically, when the sensor's credibility score falls below a preset threshold, indicating a possible signal error, the method enters the error pattern diagnosis phase. First, based on the previously calculated statistical characteristics, change consistency, and deviation degree, the first, second, and third deviation state features are extracted. These deviation state features further refine and quantify the original characteristics, aiming to capture the specific patterns or abnormal signals exhibited by different error modes within these characteristics. This separate extraction ensures comprehensive quantification of the signal's abnormal state from different dimensions (the signal's own statistical behavior, its relationship with other related signals, and its comparison with historical behavior). Next, the extracted first, second, and third deviation state features are concatenated to form a fused deviation feature vector. This fused vector integrates abnormal information from different dimensions, providing a more comprehensive and discriminative feature representation that helps more accurately describe the current signal's error state. Finally, this fused deviation feature vector is input into a pre-defined error pattern recognition model based on a multi-class support vector machine. This model is pre-trained to learn and identify the feature vector patterns corresponding to different error modes. By inputting the fused feature vector into this model, the model can automatically classify and diagnose the error type of the current signal, thereby generating a specific error pattern. This machine learning-based classification method utilizes the patterns learned by the model from a large amount of historical data to achieve automated, objective, and highly accurate identification of error patterns, providing a reliable basis for subsequent targeted calibration.

[0058] The above diagnostic process is performed after the credibility score is calculated and before the calibration scheme is extracted. It inherits the judgment results of signal anomalies and provides specific error type information for subsequent calibration steps, making the entire calibration process more complete and effective, thereby improving the data reliability and warning accuracy of the landslide monitoring system.

[0059] In some preferred embodiments, the error patterns include: Zero drift, sensitivity degradation, intermittent failures and periodic interference.

[0060] Specifically, zero drift refers to the phenomenon that the output signal of the sensor deviates from its theoretical zero value when the input physical quantity is zero, which can be manifested as an overall increase or decrease in the output signal baseline. Sensitivity attenuation refers to the phenomenon that the proportional relationship between the sensor output signal and the change of the input physical quantity decreases, which can be manifested as a decrease in the gain of the sensor output signal. Intermittent failure refers to the phenomenon that the sensor outputs abnormally or has no output for a period of time and then returns to normal. It can be manifested as an interruption in the data stream or the appearance of outliers. Among them, periodic interference refers to the phenomenon that a noise signal with a fixed period or a nearly fixed period is superimposed on the sensor output signal, which can be manifested as regular fluctuations in the output signal.

[0061] More specifically, the method of the present application provides a clear classification target for the error pattern recognition model constructed by the multi-classifier based on the support vector machine by clearly defining the type of error pattern. The fused feature vector is input into the preset multi-classifier based on the support vector machine. The classifier maps the feature vector to one of the four error patterns defined in this scheme based on its training results. For example, if the pattern of the feature vector is similar to the feature pattern marked as "zero drift" in the training sample, the classifier outputs "zero drift" as the recognized error pattern.

[0062] The method of the present application clearly defines the types of error patterns, provides a clear classification target for the error pattern recognition model, enables the model to be trained and applied, and improves the accuracy of error pattern recognition. Identifying specific error patterns enables the subsequent extraction of corresponding calibration schemes from the calibration scheme library. Different error patterns require different calibration methods and parameters. Clearly defining the error pattern types allows the calibration process to be calibrated according to the identified error types. This improves the accuracy and effectiveness of calibration, ultimately improving the reliability of data acquisition by the landslide monitoring system and thereby enhancing the accuracy of early warnings.

[0063] In some preferred embodiments, the first deviation state feature includes the mean drift amplitude and variance fluctuation frequency; the second deviation state feature includes the reduction amplitude of the Pearson correlation coefficient; and the third deviation state feature includes the mean difference and variance difference between the monitoring signal and the historical signal.

[0064] Specifically, the mean drift amplitude refers to the degree to which the mean of the sensor signal changes over time, which can be achieved by calculating the amount of change or the rate of change of the signal mean within a time window. The variance fluctuation frequency refers to the speed or periodicity of the fluctuation of the sensor signal variance over a period of time, which can be achieved by performing spectral analysis on the signal variance sequence or calculating the number of times the variance crosses a specific threshold per unit time. The reduction amplitude of the Pearson correlation coefficient refers to the degree to which the Pearson correlation coefficient between the monitoring signal and the signal of the physically associated sensor decreases relative to the normal state or historical state, which can be achieved by calculating the difference between the current correlation coefficient and the historical average correlation coefficient or the normal correlation coefficient. The mean difference between the monitoring signal and the historical signal refers to the difference between the mean of the current monitoring signal and the mean of its historical signal under similar conditions, which can be achieved by directly calculating the difference between the current signal mean and the historical signal mean.

[0065] More specifically, the method of the present application specifically defines the specific contents of the three deviation state features used to construct the fusion deviation feature vector, thereby making the error pattern recognition process more specific and operational. The first deviation state feature is defined as including the mean drift amplitude and the variance fluctuation frequency, which means that the features extracted from the statistical characteristics of the sensor are to quantify the magnitude of the change in the signal mean and the speed of the variance fluctuation frequency. These indicators can reflect the drift or noise characteristics of the signal and provide basic data for identifying error patterns such as zero drift or periodic interference. The second deviation state feature is defined as including the reduction amplitude of the Pearson correlation coefficient, which means that the features extracted from the change consistency of the sensor and the physically associated sensor are to quantify the degree of correlation reduction between it and the physically associated sensor, which helps to identify relative anomalies between the sensor and other normally working sensors, which may indicate problems such as sensitivity attenuation or intermittent failure. The third deviation state feature is defined as including the mean difference and variance difference between the monitoring signal and the historical signal, which means that the features extracted from the degree of deviation between the monitoring signal and the historical signal are to quantify the difference between the mean and variance of the current signal and its historical performance under similar conditions. This can reflect whether the signal has deviated from its normal behavior pattern and provide a historical comparison basis for identifying various types of errors.

[0066] More specifically, the method of the present application provides a clear and specific input feature vector for the error pattern recognition model based on the fusion deviation feature vector by clarifying the specific quantitative indicators of these three deviation state characteristics, thereby improving the accuracy and operability of error pattern recognition, thereby enabling more reliable diagnosis of sensor error types and providing a basis for the subsequent selection of targeted calibration schemes.

[0067] In some preferred embodiments, step S5 includes: S51 , extracting a corresponding calibration solution from a preset calibration solution library according to the landslide activity level, error pattern, and corresponding sensor type at the previous moment.

[0068] Specifically, the sensor type refers to the specific category of the sensor in which the error occurs, which determines the sensor's measurement principle, physical properties, and sensitivity to environmental factors; the preset calibration scheme library refers to a collection that stores a variety of sensor signal calibration strategies, methods, and parameters, which structuredly organizes calibration schemes for different landslide states, error types, and sensor types.

[0069] Specifically, after the sensor signal's credibility score falls below a preset threshold and an error pattern is identified, the method of this application no longer selects a calibration scheme based solely on the error pattern. Instead, it further incorporates the landslide activity level at the previous moment and the type of sensor in which the error occurred. Based on these three pieces of information, a query or match is performed within a pre-set calibration scheme library. This calibration scheme library pre-builds optimized calibration strategies for different combinations of conditions. For example, for an active landslide, even with the same zero-drift error, different calibration algorithms or stricter calibration parameters may be required for the displacement sensor and pore-water pressure sensor. By comprehensively considering these three factors, the system extracts a more targeted calibration scheme from the library that better suits the current monitoring scenario and sensor characteristics. This multi-dimensional matching approach enables the selected calibration scheme to more effectively address specific error issues, while simultaneously taking into account the data accuracy and reliability requirements of the landslide state and the inherent characteristics of different sensor types. This approach makes the calibration process more intelligent and refined, improving the accuracy and reliability of landslide monitoring data and avoiding the mismatch issues that can arise from using a universal calibration scheme.

[0070] In some preferred embodiments, the calibration scheme includes calibration means and calibration parameter types, and step S6 includes: S61. Based on the type of the parameter to be calibrated, determine the calibration parameter according to the monitoring signal of the sensor to be calibrated and the historical signal of the same type of sensor under the same environmental conditions or the same time conditions; S62: Update the configuration information of the corresponding sensor based on the calibration means and calibration parameters to calibrate the monitoring signal of the corresponding sensor.

[0071] Specifically, the calibration schemes in the calibration scheme library include a zero drift calibration scheme, a sensitivity attenuation calibration scheme, and a periodic interference calibration scheme.

[0072] More specifically, a calibration method refers to the method or algorithm used to correct sensor monitoring signals. This can be achieved by adding an offset, multiplying by a scaling factor, applying a filtering algorithm, or performing waveform fitting. A calibration parameter type refers to the type of parameter that must be determined when performing a calibration method. A calibration parameter is a specific value calculated based on the calibration parameter type and actual data, used to quantify sensor errors and guide calibration operations. Updating the corresponding sensor configuration information refers to applying the determined calibration parameters to the sensor data processing flow. This can be achieved by modifying the parameters of the sensor driver, updating the configuration registers of the data acquisition module, or adjusting the algorithm parameters of the back-end data processing software.

[0073] Specifically, this solution addresses the issue of how to effectively perform signal calibration after identifying sensor error patterns by defining the calibration scheme's composition and steps. First, the system extracts a corresponding calibration scheme from a calibration scheme library based on the identified error pattern (e.g., zero drift, sensitivity loss, or periodic interference). Each extracted calibration scheme provides the calibration method (calibration method) and the type of parameters to be determined (calibration parameter type). In step S61, the system uses the calibration parameter type specified in the calibration scheme, combined with the current monitoring signal of the sensor to be calibrated and historical signals from similar sensors under similar conditions, to calculate specific calibration parameter values. For example, for zero drift, the mean difference between the current signal and the historical signal can be calculated as the offset value; for sensitivity loss, the amplitude ratio between the current signal and the historical signal can be calculated as the scaling factor. This dynamic determination of calibration parameters based on actual monitoring data and historical reference data improves the targetedness and accuracy of calibration. Subsequently, in step S62, the system updates the corresponding sensor's configuration information based on the specific calibration parameter values ​​calculated in S61 and the calibration method obtained from the calibration scheme. This means applying the calculated calibration parameters to subsequent signal processing, for example, by using software algorithms to perform real-time corrections on the original monitoring signal, thereby achieving calibration of the sensor monitoring signal. The entire process organically combines error identification, solution selection, parameter determination, and signal correction to form an automated sensor signal calibration process. More specifically, By combining the method of this application with error pattern identification and calibration scheme selection (as implemented in S4 and S5), this solution provides a method that automatically selects an appropriate calibration method and determines calibration parameters based on the actual deviation state of the sensor signal and the identified error pattern, thereby correcting the sensor signal. This method can effectively and accurately determine the required calibration parameters and perform signal calibration operations. This solves the problem of difficult or ineffective calibration processes and improves the reliability of landslide monitoring data.

[0074] Second, please refer to Figure 2 Some embodiments of the present application further provide a landslide monitoring sensor signal calibration system, which is applied to sensors in a landslide status monitoring system. The system includes: An acquisition module 201 is used to continuously acquire monitoring signals from multiple sensors; Characteristic information acquisition module 202, for acquiring statistical characteristics of each sensor, consistency of changes in physically associated sensors, and degree of deviation from historical signals based on monitoring signal analysis; Scoring module 203, used to calculate and obtain the credibility score of each sensor based on statistical characteristics, change consistency and deviation degree; The error analysis module 204 is configured to obtain an error pattern based on one or more of statistical characteristics, change consistency, and deviation degree when the credibility score is lower than a preset score threshold; The calibration scheme determination module 205 is used to extract a calibration scheme from a preset calibration scheme library according to the error pattern; The calibration module 206 is configured to calibrate the monitoring signal of the corresponding sensor according to the calibration scheme.

[0075] The system of the present application continuously analyzes the statistical characteristics of sensor signals, the consistency of changes in physically associated sensors, and the degree of deviation from historical signals, calculates the credibility score of the signal, and diagnoses the specific error pattern based on these analysis results when the score is lower than a threshold, thereby extracting and applying the corresponding calibration scheme from a preset calibration scheme library, thereby realizing automated, data-driven calibration of landslide monitoring sensor signals, overcoming the limitations of traditional manual calibration in field environments, improving the accuracy of monitoring data and the reliability of the system, reducing the need and cost of manual calibration, and enhancing the early warning capability of the landslide monitoring system.

[0076] In addition, the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0077] Furthermore, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0078] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any actual relationship or order between these entities or operations.

[0079] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A landslide monitoring sensor signal calibration method, applied to sensors in a landslide status monitoring system, characterized in that: The method comprises the following steps: S1. Continuously obtain monitoring signals from multiple sensors; S2. Analyzing the monitoring signals to obtain statistical characteristics of each sensor, consistency of changes with physically associated sensors, and a degree of deviation from historical signals; S3. Calculate and obtain a credibility score for each sensor based on the statistical characteristics, change consistency, and deviation degree; S4. When the credibility score is lower than a preset score threshold, obtain an error pattern based on one or more of the statistical characteristics, the change consistency, and the deviation degree; S5. extracting a calibration solution from a preset calibration solution library according to the error pattern; S6. Calibrate the monitoring signal of the corresponding sensor according to the calibration scheme.

2. The landslide monitoring sensor signal calibration method according to claim 1, characterized in that: Step S2 includes: S21. Calculate the mean, variance, and fluctuation range of each sensor according to the monitoring signal as statistical characteristics; S22. Calculate, based on the sensor pairs having physical correlation, a Pearson correlation coefficient between the sensor pairs as the change consistency of the corresponding sensors; S23. Obtain the degree of deviation based on the difference between the historical signal of the same type of sensor under the same environmental conditions or the same time conditions and the monitoring signal.

3. The landslide monitoring sensor signal calibration method according to claim 2, characterized in that: Step S23 includes: S231. Extract historical signals of similar sensors under the same environmental conditions or the same time conditions, and calculate the mean and variance of the historical signals; S232. Calculate the mean difference and variance difference between the historical signal and the monitoring signal based on the statistical characteristics; S233. Calculate the degree of deviation according to the mean difference and the variance difference.

4. The landslide monitoring sensor signal calibration method according to claim 1, characterized in that: Step S3 includes: S31, obtaining the activity level of the landslide at the last moment; S32, extracting a weight group from a preset weight database according to the activity level and the type of sensor; S33. Calculate the statistical characteristics, change consistency, and deviation degree according to the weight group to obtain a credibility score.

5. The landslide monitoring sensor signal calibration method according to claim 4, characterized in that: The process of determining the activity level of the landslide includes: A1. Extract the displacement, rainfall, and groundwater level change of the corresponding landslide within the preset time window; A2. According to the displacement of the landslide, the rainfall, and the groundwater level change, a preset activity level discrimination rule library is queried to obtain the activity level.

6. The landslide monitoring sensor signal calibration method according to claim 1, characterized in that: Step S4 includes: S41. When the credibility score is lower than a preset score threshold, extracting a first deviation state feature, a second deviation state feature, and a third deviation state feature according to the statistical characteristics, change consistency, and deviation degree; S42: Concatenate the first deviation state feature, the second deviation state feature, and the third deviation state feature to obtain a fused deviation feature vector; S43: Input the fused deviation feature vector into a preset error pattern recognition model to generate the error pattern.

7. The landslide monitoring sensor signal calibration method according to claim 6, characterized in that: The error modes include: zero drift, sensitivity attenuation, intermittent failure and periodic interference.

8. The landslide monitoring sensor signal calibration method according to claim 1, characterized in that: Step S5 includes: S51 , extracting a corresponding calibration solution from a preset calibration solution library according to the landslide activity level at the last moment, the error pattern, and the type of the corresponding sensor.

9. The landslide monitoring sensor signal calibration method according to claim 1, characterized in that: The calibration scheme includes calibration means and calibration parameter types, and step S6 includes: S61. Based on the calibration parameter type, determine the calibration parameter according to the monitoring signal of the sensor to be calibrated and the historical signal of the same type of sensor under the same environmental conditions or the same time conditions; S62: Update the configuration information of the corresponding sensor based on the calibration means and the calibration parameters to calibrate the monitoring signal of the corresponding sensor.

10. A landslide monitoring sensor signal calibration system, applied to sensors in a landslide status monitoring system, characterized in that: The system comprises: An acquisition module, used to continuously acquire monitoring signals from multiple sensors; A characteristic information acquisition module is used to analyze and obtain the statistical characteristics of each sensor, the consistency of changes in the physically associated sensor, and the degree of deviation from historical signals based on the monitoring signal; A scoring module, configured to calculate and obtain a credibility score for each sensor based on the statistical characteristics, change consistency, and deviation degree; an error analysis module, configured to obtain an error pattern based on one or more of the statistical characteristics, the change consistency, and the deviation degree when the credibility score is lower than a preset score threshold; A calibration scheme determination module, configured to extract a calibration scheme from a preset calibration scheme library according to the error pattern; The calibration module is used to calibrate the monitoring signal of the corresponding sensor according to the calibration scheme.

Images