A method and system for self-adapting calibration of a level i angle

Abstract

The present application relates to the field of engineering surveying, and discloses a leveling instrument i angle self-adaptive calibration method and system, comprising constructing an i angle calibration dataset fused with environmental parameters; using a machine learning model to intelligently predict i angle deviation; using a multi-station dynamic verification mechanism and an optimization algorithm to cooperatively correct the prediction results; real-time calculation and dynamic modeling to achieve full-automatic calibration; and building a complete self-adaptive calibration system, which can realize real-time and self-adaptive calibration of the i angle in a complex environment, and improve the measurement stability and precision. Compared with the traditional static calibration method, the present application can reduce manual intervention, improve calibration efficiency, and effectively support high-precision engineering surveying and long-term automatic monitoring scenarios.

Description

Technical Field

[0001] This invention relates to the field of engineering surveying technology, and in particular to an adaptive calibration method and system for the i-angle of a level. Background Technology

[0002] In the field of high-precision elevation measurement, the i-angle error of a level (i.e., the angle between the instrument's optical axis and the vertical line of gravity) is widely considered to be one of the key factors affecting measurement accuracy. Traditional i-angle calibration methods typically rely on manual, periodic static calibration under standard environmental conditions. While this technique offers a degree of controllability in laboratory environments, it suffers from significant technical limitations and application bottlenecks in practical field engineering applications. Specifically, this static calibration system faces three main challenges. First, it suffers from severe environmental adaptability. Existing calibration techniques are generally based on a key assumption: that the i-angle remains stable over a certain timescale after calibration. However, in actual engineering measurements, external environmental factors such as temperature, humidity, atmospheric pressure, and altitude fluctuate dynamically, significantly affecting the thermal expansion and contraction and mechanical deformation of the level's structural materials, thus causing i-angle drift. This environmental sensitivity is particularly pronounced in measurement tasks conducted across seasons or multiple climate zones. For example, in high-speed rail track settlement monitoring, when the diurnal temperature difference exceeds 15°C, the level's i-angle can change by milliradians, leading to millimeter-level elevation deviations. This drift exhibits significant temporal correlation and spatial heterogeneity, and traditional calibration methods cannot provide a dynamic compensation mechanism, leading to a continuous amplification of cumulative errors with increasing measurement distance and time, thus posing a challenge to the consistency and stability of the elevation benchmark. Secondly, engineering efficiency is severely constrained. For typical large-scale linear infrastructure projects, such as highways, oil and gas pipelines, and power lines, measurement tasks often span multiple geographical and climatic zones, requiring frequent movement of measuring equipment. Limited by the poor spatial adaptability of traditional static calibration methods, measurement teams often need to frequently interrupt operations for manual calibration, which not only greatly reduces operational efficiency but also introduces systematic errors due to insufficient calibration frequency, especially in high-intensity continuous operation scenarios. Existing technologies generally lack the ability to quantitatively model and dynamically predict the relationship between environmental parameters and i-angle changes, failing to achieve the goal of "one-time modeling, multi-scenario adaptation" or cross-environment error migration correction. This severely restricts the widespread application of intelligent measuring equipment in complex environments. Thirdly, the stability defects of long-term monitoring are prominent. In long-term settlement monitoring and structural health assessment of critical infrastructure such as bridges, dams, and tunnels, the temporal aliasing problem between i-angle error and the actual displacement signal is particularly serious. Traditional adjustment methods typically assume that instrument errors are white noise, which cannot effectively separate the system drift caused by slow changes in the i-angle from the actual structural deformation signal, especially in unattended remote automated monitoring systems. As monitoring time progresses, small shifts in the i-angle can gradually accumulate into statistically significant systematic errors, which in extreme cases may mask the true settlement trend, leading to misjudgments of the structural condition and consequently distorted safety decisions.

[0003] CN111721260A discloses a high-precision beam method for settling measurement based on the correction of the i-angle error of a level instrument. This method is the same as most existing i-angle calibration methods, which correct the i-angle of the level instrument by calculating the height difference. It belongs to a static calibration system that does not consider environmental changes. Summary of the Invention

[0004] To overcome the problems of insufficient environmental adaptability, poor engineering efficiency, and system errors caused by slow changes in the i-angle in existing i-angle calibration methods, this invention provides an adaptive i-angle calibration method and system for a level.

[0005] In a first aspect, the present invention provides an adaptive calibration method for the i-angle of a level, the method comprising: The environmental parameters of the target area are obtained, and the measured value of angle i and the observation deviation of angle i are obtained based on the preset monitoring stations in the target area. The measured value of angle i and the environmental parameters are input into the trained prediction model to obtain the predicted value of angle i. The optimal estimation is performed based on the measured value of angle i, the observed deviation of angle i, and the predicted value of angle i to obtain the calibrated angle i value; The preset monitoring stations include at least two.

[0006] According to one specific implementation, in the above calibration method, the optimal estimation adopts the Kalman filter algorithm.

[0007] According to one specific implementation, in the above calibration method, the training process of the prediction model includes: In a stable operating area, the measured value of the i-angle is obtained through monitoring stations as a training label, and the corresponding environmental training parameters are recorded synchronously. A training dataset is constructed based on the training label and the environmental training parameters. The training dataset is input into the LSTM neural network for training to obtain the trained prediction model.

[0008] According to one specific implementation, the calibration method described above includes constructing a training dataset, specifically comprising: After performing a quality check on the environmental training parameters, normalization is then used to eliminate the influence of dimensions, and a multidimensional dataset is generated. The quality inspection includes: Savitzky-Golay filtering is used to reconstruct the time series for the missing time series values ​​in the environmental training parameters. Through 3 The principle is to identify outliers in the corresponding environmental training parameters and correct or remove data points that exceed the mean ± 3 times the standard deviation. The normalization uses the MinMax normalization method to scale the training parameters of each environment after quality check to the [0,1] interval.

[0009] According to one specific implementation, the calibration method further includes: Based on the location information of the target area, the target area is identified to be in different geographical environments. Through transfer learning technology, the parameters of the trained prediction model are adjusted by combining the measured i-angle values ​​collected in the corresponding geographical environment and the environmental training parameters.

[0010] According to one specific implementation, the calibration method further includes: Construct a regional feature knowledge base, and encapsulate the adjusted prediction model and optimal estimation process under different geographical environments in the regional feature knowledge base; Obtain the positioning information of the target area, match the corresponding adjusted prediction model based on the positioning information, and perform i-angle calibration.

[0011] According to one specific implementation, in the above calibration method, the environmental parameters are the same as the environmental training parameters, including temperature, humidity and air pressure.

[0012] According to one specific implementation, the calibration method further includes: The trained prediction model is subjected to lightweight modification; the lightweight modification includes model pruning and weight quantization.

[0013] Secondly, the present invention provides an i-angle adaptive calibration system for a level, the system comprising: A level instrument, set at a pre-set monitoring station, is used to collect the measured value of the i-angle of the target area; Environmental sensing devices are used to collect environmental parameters of the target area; Positioning device, used to collect positioning information of target area; A central processing unit is used to obtain a calibrated i-angle value using an adaptive calibration method for a level as described in any of the above claims, and to calibrate the level based on the calibrated i-angle value. The preset monitoring stations include at least two.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention combines machine learning intelligent prediction with multi-station dynamic verification technology, achieving a technological leap from traditional manual operation to fully automated intelligent leveling angle calibration. By integrating environmental parameters, it can achieve accurate real-time prediction under various complex working conditions. The multi-station collaborative correction mechanism combined with filtering algorithms significantly improves calibration accuracy. Compared to traditional methods, this invention, through real-time calculation of the i-angle and dynamic modeling using machine learning, compresses the entire process to within a few hours, improving overall efficiency by 80-90%. Simultaneously, the model's self-learning capability reduces manual intervention by over 70% and improves accuracy by 20-30%, providing an intelligent solution for modern engineering surveying. Attached Figure Description

[0015] Figure 1 A schematic flowchart of an adaptive calibration method for the i-angle of a level provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an LSTM neural network provided in an embodiment of the present invention; Figure 3 A schematic diagram of the optimal estimation process provided for an embodiment of the present invention. Detailed Implementation

[0016] The present invention will now be described in further detail with reference to specific embodiments. However, this should not be construed as limiting the scope of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0017] Unless otherwise specified, the terms "upper," "lower," "left," "right," "center," "inner," and "outer," etc., used in the description of specific embodiments of the present invention to indicate orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings, or the orientation or positional relationship in which the product / equipment / device is usually placed during use. These terms are merely for the purpose of facilitating the description of the present invention or simplifying the description in specific embodiments, and for enabling those skilled in the art to quickly understand the solution, and do not indicate or imply that a particular device / component / element must have a specific orientation, or be constructed and operated in a specific positional relationship. Therefore, they should not be construed as limitations on the present invention.

[0018] Furthermore, the use of terms such as "horizontal," "vertical," "suspended," "parallel," and "coaxial" does not imply that the corresponding device / component / element must be absolutely horizontal, vertical, suspended, parallel, or coaxial. Slight tilt or deviation is permissible, as long as it does not affect the normal function of the relevant component. For example, "horizontal" simply means that its direction is more horizontal relative to "vertical," not that the structure must be perfectly horizontal; a slight tilt is acceptable. "Coaxial" means that two components are arranged as coaxially as possible, allowing them to move coaxially or approximately coaxially when their relative positions change. Alternatively, it can be simplified to mean that the corresponding device / component / element, when arranged in "horizontal," "vertical," "suspended," "parallel," or "coaxial" directions, can have an error / deviation of ±10% relative to the corresponding direction, more preferably within ±8%, more preferably within ±6%, more preferably within ±5%, and more preferably within ±4%. For example, the deviation in the "coaxial" direction is controlled within 0.2-1mm, preferably within 0.2-0.5mm. As long as the corresponding device / component / element is within the error / deviation range, it can still achieve its function in the solution of the present invention.

[0019] Furthermore, in the description of the embodiments of the present invention, "several", "more than", and "a number of" represent at least two. The number can be any number, such as two, three, four, five, six, seven, eight, or nine, and can even exceed nine.

[0020] Furthermore, in the description of the technical solution of this invention, unless otherwise explicitly specified / limited / restricted, the terms "set up," "install," "connect," "link," "provided with," "laid out," and "arranged" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to connection methods commonly used in the art, such as welding, riveting, bolting, and threaded connections. Such connections can be mechanical, electrical, or communication connections; they can be direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components.

[0021] In the field of engineering surveying, the stability of the i-angle of a level directly affects the reliability of settlement observation data. Traditional i-angle calibration methods rely on manual periodic calibration, which is not only inefficient but also unable to cope with real-time drift problems caused by complex environments. Especially in long-term monitoring of large infrastructure, dynamic changes in the i-angle can accumulate significant errors, seriously affecting measurement accuracy. Although existing technologies attempt to correct through environmental compensation or fixed formulas, they lack the ability to model the nonlinear effects of multiple coupled factors and cannot achieve intelligent adaptation across regions. In current practical engineering cases, i-angle drift has been repeatedly found to threaten the stability of long-term monitoring data, urgently requiring the construction of an intelligent calibration mechanism that can perceive environmental changes in real time and dynamically correct i-angle errors. Traditional static i-angle calibration systems show significant technological lag when facing complex application scenarios such as dynamic environments, cross-regional measurements, and long-term unattended monitoring. The key to solving these challenges lies in building an intelligent calibration system based on environmental perception, with machine learning as its core, and possessing online learning and dynamic compensation capabilities, thereby achieving full lifecycle management and real-time adaptive control of i-angle errors.

[0022] The core of this invention lies in establishing an i-angle dynamic closed-loop calibration system based on machine learning and environmental perception. Through multi-station collaborative observation and adaptive learning mechanism, it realizes intelligent compensation and accuracy improvement of level measurement error.

[0023] This invention relates to an adaptive calibration system for the i-angle of a level instrument based on machine learning and multi-station dynamic correction. It integrates multiple sensors to collect environmental parameters and observation data in real time, establishing a dynamic correlation model between the i-angle and environmental factors. After training an i-angle change prediction model using machine learning algorithms, it is deployed to a mobile device for real-time prediction and compensation. To improve system reliability, differential verification and data fusion are performed using observation data from multiple stations, effectively eliminating single-station system errors. The system possesses adaptive learning capabilities to dynamically adjust model parameters, catering to the environmental characteristics of different regions. Furthermore, an intelligent feedback mechanism continuously optimizes the i-angle error.

[0024] The purpose of this invention is to achieve adaptive adjustment of the i-angle according to environmental parameters, reduce manual intervention, and improve the measurement efficiency and accuracy of cross-regional engineering. It proposes a dynamic i-angle calibration method based on machine learning modeling and multi-station collaborative correction.

[0025] Specifically, firstly, a nonlinear mapping model between environmental parameters and i-angle drift is constructed, and the LSTM algorithm is used to dynamically predict i-angle changes, breaking through the limitations of traditional static calibration and significantly improving environmental adaptability.

[0026] Secondly, by fusing multi-station differential analysis and Kalman filter data, real-time error correction and system stability enhancement were achieved, solving the problem that single-station observations are easily affected by random errors.

[0027] Finally, a cross-regional transfer learning mechanism is introduced to enable the model to have geographical generalization ability, and a complete closed-loop control system from environmental perception to intelligent prediction to automatic calibration is constructed, thereby reducing the need for manual intervention and improving the automation level and engineering efficiency of settlement monitoring.

[0028] This invention achieves a technological leap from passive manual adjustment to active intelligent compensation in i-angle calibration of level instruments through the deep integration of machine learning and environmental perception, combined with multi-station dynamic verification and adaptive optimization. It has significant advantages in terms of measurement accuracy, automation level and engineering applicability.

[0029] The technical solution provided by the present invention will be further described and explained below with reference to specific embodiments.

[0030] Please refer to Figure 1 The diagram illustrates a flowchart of an adaptive calibration method for the i-angle of a level instrument provided by an embodiment of the present invention. The method includes: Step 1: Obtain the environmental parameters of the target area, and obtain the measured value of angle i and the observation deviation of angle i based on the preset monitoring stations in the target area.

[0031] In this embodiment of the invention, the preset monitoring stations include at least two. The target area is the area where key measurement points are set during the actual operation, and the preset monitoring stations are set at the key measurement points for observation. The multi-station system can obtain the measured i-angle value of the target area and the respective i-angle observation deviation through synchronous observation.

[0032] Step 2: Input the measured value of angle i and the environmental parameters into the trained prediction model to obtain the predicted value of angle i.

[0033] The training process of the prediction model includes: In a stable operating area, the measured value of the i-angle is obtained through monitoring stations as a training label, and the corresponding environmental training parameters are recorded synchronously. A training dataset is constructed based on the training label and the environmental training parameters. The training dataset is input into the LSTM neural network for training to obtain the trained prediction model.

[0034] For example, two stations, A and B, can be set up in a stable environment, 40-60 meters apart, with a leveling rod placed between them for forward and backward measurements. Stations A and B observe the leveling rod readings, then exchange positions and repeat the measurements to eliminate residual errors, obtaining multiple sets of elevation difference data. Simultaneously, environmental parameters such as temperature, humidity, and air pressure are recorded to construct a multidimensional dataset containing spatiotemporal correlations. Then, using the multi-station observation data, the initial value of angle i is calculated using a traditional formula. The calculation formula is:

[0035] in, The value is 206265″, where D is the distance between the two stations. The difference is in elevation.

[0036] Furthermore, based on the collected multidimensional dataset, feature engineering was performed, including time-series analysis of environmental parameters and spatial correlation modeling. An LSTM neural network was selected to construct the prediction model, using environmental parameters such as temperature, humidity, and air pressure as input features, and the measured i-angle value as the output label. Cross-validation was used to optimize the model hyperparameters, and key environmental influencing factors were identified through feature importance analysis. Please refer to [reference needed]. Figure 2 This illustrates a schematic diagram of the LSTM neural network provided in an embodiment of the present invention.

[0037] Specifically, each time step of the model includes the following gating mechanisms: a forget gate, an input gate, a candidate state gate, and an output gate. The forget gate determines how much information from the previous time step's memory state should be discarded, and its formula is:

[0038] in, It is the sigmoid activation function; Here is the weight matrix for the forget gate; This is the hidden state from the previous time step; The input vector is such as the current temperature, humidity, and air pressure. This is the forget gate bias vector.

[0039] The input gate determines which information from the current input is added to the memory unit, and the formula is:

[0040]

[0041] in, For input gate output; The input gate weight matrix; The input gate bias vector; represents the candidate memory information for the current time step; tanh is the hyperbolic tangent function; , For memory cell weights and biases.

[0042] The formula for updating the memory state is:

[0043] in, The state of the memory cell at the current time step; This is for Hadamard element-wise multiplication.

[0044] The formulas for output gate and hidden state are:

[0045]

[0046] in, For output gate; , The output gate weights and biases. This represents the hidden state at the current time step, which is the output of the LSTM.

[0047] The final formula for the predicted value of angle i is:

[0048] in, To predict the i-angle value of the level instrument; , The hidden states are mapped to actual predicted values ​​to form the output layer weights and biases. This invention uses environmental parameters such as temperature, humidity, and air pressure as input features, and measured i-angle values ​​as output labels. Cross-validation is used to optimize the model hyperparameters, and key environmental influencing factors are identified through feature importance analysis to establish an environmental i-angle mapping model. The collected environmental parameter data (temperature, humidity, air pressure, etc.) undergo quality checks, and Savitzky-Golay filtering is used to reconstruct the time series for missing values ​​to ensure data continuity. Through 3... Outliers are identified in principle, and data points exceeding the mean ± 3 standard deviations are corrected or removed. Simultaneously, the MinMax normalization method is used to scale each feature to the [0,1] interval to eliminate the influence of dimensions.

[0049] Furthermore, this embodiment of the invention employs an LSTM neural network structure, combined with Bayesian hyperparameter optimization and time-series cross-validation, to establish a nonlinear mapping relationship between environmental parameters and the i-angle value. Bayesian optimization constructs a probabilistic surrogate model of the objective function and dynamically selects the next set of hyperparameter combinations to be tested, using the acquisition function to approximate the global optimum with the fewest iterations. Its mathematical essence is:

[0050] in, Predict the mean for the surrogate model. For uncertainty estimation; k To explore and develop a trade-off coefficient.

[0051] Further, model validation. In this embodiment, the training set and validation set are divided according to spatial location for training. The root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R-Squared, R²) are used to test the generalization ability of the prediction model.

[0052] RMSE measures the difference between predicted and measured values, amplifying the contribution of larger errors through the sum of squares and square root calculations. The calculation formula is as follows:

[0053] MAE calculates the average of the absolute differences between predicted and measured values, applies a linear penalty to the error, is unaffected by outliers, and provides intuitive results. The calculation formula is as follows:

[0054] R² (R-Squared) represents the proportion of the variance of the target variable explained by the model, and measures the goodness of fit. The calculation formula is as follows:

[0055] in, The average of the measured values. These are measured values; is the predicted value; n is the number of samples.

[0056] R² reveals a model’s ability to explain the fluctuations in the target variable. Its standardized properties (ranging from -∞ to 1) allow for comparisons across datasets. 1 indicates a perfect fit, 0 is equivalent to mean prediction, and a negative value indicates that the model is worse than the benchmark. In other words, the closer R² is to 1, the better the fit.

[0057] It is understood that the stable operating area in the embodiments of the present invention refers to the area where the measured values ​​of the i-angle observed by each monitoring station have been calibrated and the terrain is stable and unchanged, so as not to cause deviation in the i-angle observation.

[0058] Step 3: Based on the measured value of angle i, the observed deviation of angle i, and the predicted value of angle i, perform optimal estimation to obtain the calibrated angle i value.

[0059] Specifically, in actual operation, a multi-station system is deployed at key measurement points for synchronous observation. By comparing the differences in measurement results from multiple stations, the actual deviation of the i-angle is accurately calculated. Using a Kalman filter algorithm, the predicted values ​​from the machine learning model are optimized with the measured data from multiple stations, and the i-angle parameters are updated and dynamically adjusted in real time to obtain the optimized i-angle value. This i-angle value can be used as a calibration parameter for the leveling instrument to adjust the instrument's zero point or correct historical measurement data, or it can be directly used as a value for subsequent measurements to improve the accuracy of operations such as elevation measurement. Through this method, the measurement accuracy and reliability of the leveling instrument are significantly improved, achieving closed-loop optimization from intelligent prediction to practical application.

[0060] In this embodiment of the invention, the optimal estimation employs the Kalman filter algorithm. The Kalman filter algorithm is an optimal recursive algorithm for state estimation of dynamic systems. Its core idea is to progressively optimize the state estimation through two steps: prediction and update, combining the system model and observation data. Please refer to [reference needed]. Figure 3 This illustrates a flowchart of the optimal estimation process provided by an embodiment of the present invention.

[0061] Specifically, the calculation process of Kalman filtering is as follows: 1. Initialization Before starting filtering, initial values ​​need to be set: Initial state estimation vector Let : be the measured value of the initial i-angle, representing the state of the system at the initial moment.

[0062] Initial covariance matrix : Represents the uncertainty of the initial state estimation, i.e., the error covariance of the measured value of angle i.

[0063] 2. Loop processing (time step) ） For each time step Perform the following steps: (1) Prediction: Based on the state at the previous time step, predict the state and covariance at the current time step.

[0064] State prediction:

[0065] in, Let f be the state transition matrix, which describes the evolution of angle i from the previous time f1 to the current time f; It is the measured or optimized estimate of angle i at the previous moment.

[0066] Covariance prediction:

[0067] in, Let be the covariance matrix of the previous time step; The process noise covariance matrix represents the uncertainty of the system model (such as external disturbances).

[0068] (2) Calculate the Kalman gain: determine the weights of the predicted and measured values ​​to minimize the estimation error.

[0069]

[0070] in, The observation matrix maps the i-angle state to the observation space; The observation noise covariance matrix represents the measurement error of the measured values ​​from multiple stations. If observation noise If it is larger, then The smaller the value, the more the filtering result depends on the predicted value of the i-angle; If observation noise Smaller, then The value is relatively large, and the filtering result depends more on the measured value of the i-angle.

[0071] (3) State update: Combine the predicted value of i angle with the measured value to correct the current state estimate.

[0072]

[0073] in, The measured value of angle i at the current moment; This is called innovation, representing the observation bias at angle i. The updated i-angle estimate can be used as a correction parameter or directly.

[0074] (4) Covariance update: update the uncertainty of the state estimate, i.e. the covariance matrix.

[0075]

[0076] Where I is the identity matrix.

[0077] The uncertainty of the updated state estimate makes the optimized i-angle estimate more reliable.

[0078] 3. Termination Conditions The filtering process ends if the set number of time steps N is reached; it can also terminate early if the state estimation converges. Understandably, in the Kalman filter algorithm, the optimal estimation is always performed using the measured i-angle value, the i-angle observation bias, and the predicted i-angle value. Before the final data is output, the estimation result is represented by the aforementioned i-angle estimate. After the filtering process ends, the output is the final optimal estimation result, i.e., the calibrated i-angle value.

[0079] In one possible implementation, to achieve the continuous evolution and adaptive optimization capabilities of the method provided by the present invention, the calibration method further includes: Based on the location information of the target area, the target area is identified to be in different geographical environments. Through transfer learning technology, the parameters of the trained prediction model are adjusted by combining the measured i-angle values ​​collected in the corresponding geographical environment and the environmental training parameters.

[0080] For example, once a new geographical environment is identified, the above calibration method automatically triggers an incremental learning process. The core of this process lies in utilizing transfer learning technology to transfer the general knowledge accumulated by the pre-trained model to the new scene. Simultaneously, it combines environmental data collected from the new area with measured i-angle values ​​for online fine-tuning, dynamically adjusting model parameters to quickly adapt to the specific environmental characteristics of the new area, such as differences in terrain undulations, vegetation cover, or climate conditions. This online learning mechanism not only avoids the huge computational overhead of retraining required by traditional methods but also ensures continuous optimization of model performance through incremental updates, effectively solving the problem of insufficient model generalization ability when operating across regions.

[0081] To build the system's long-term memory capability, this invention simultaneously establishes a regional feature knowledge base during use, archiving and storing specific sub-models and their feature data for different geographical environments. Each sub-model encapsulates the environmental patterns and calibration experience of a specific region, forming a reusable knowledge unit. When re-entering a historical work area, the corresponding sub-model can be intelligently matched for rapid deployment, significantly improving response efficiency. This design gives the invention learning characteristics similar to human experience accumulation, gradually improving the cognitive system of environmental diversity by continuously expanding the coverage of the knowledge base. The knowledge base adopts a hierarchical storage architecture, preserving both regional specificity and maintaining globally shared features, achieving a balance between storage efficiency and model accuracy.

[0082] As the work area expands and data accumulates, this invention demonstrates significant continuous evolutionary capabilities. The synergistic effect of transfer learning and incremental learning allows the learning process in new areas to inherit existing knowledge rather than starting from scratch, greatly reducing sample requirements. Simultaneously, this invention can evaluate model performance in real time through an online feedback mechanism, enabling targeted adjustments to the feature extraction network and regression prediction module. This adaptive optimization mechanism is not only reflected in immediate improvements for single tasks but also in long-term evolution through iterative updates of the knowledge base. The resulting calibration method possesses intelligent characteristics of environmental self-awareness, model self-adjustment, and knowledge self-organization, maintaining stable measurement accuracy in complex and ever-changing field operations.

[0083] Furthermore, in order to enable the prediction model provided in the embodiments of the present invention to be deployed to mobile terminal systems, the calibration method provided in the embodiments of the present invention further includes: The trained prediction model is then subjected to lightweight modification; the lightweight modification includes model pruning and weight quantization.

[0084] Specifically, the trained prediction model is lightweighted by applying techniques such as model pruning and weight quantization to compress the model size. Model pruning reduces the model parameter size by removing redundant connections or neurons: assuming the parameter matrix in LSTM... Introducing a pruning mask matrix :

[0085] in, Indicates the first Connections are pruned to improve model sparsity. A weight-threshold-based pruning strategy can be used:

[0086] in It is the pruning threshold.

[0087] Weight quantization converts floating-point weights to fixed-point representations, reducing storage and computational load: weights... Quantified as :

[0088] in: This is the scaling factor (scale); This is the zero-point offset. Quantization significantly reduces the model's size, making it suitable for low-power devices.

[0089] Based on the above technical solutions, this invention provides an adaptive calibration method for the i-angle of a level instrument based on machine learning and multi-station dynamic correction. This method combines intelligent prediction via machine learning with dynamic verification technology from multiple stations, achieving a technological leap from traditional manual operation to fully automated and intelligent i-angle calibration. The system, through a multi-sensor fusion environmental perception network and a lightweight model, can achieve accurate real-time prediction under various complex working conditions. The multi-station collaborative correction mechanism, combined with filtering algorithms, significantly improves calibration accuracy. The regional adaptive learning function endows the system with continuous optimization capabilities, adapting to different geographical environmental characteristics. Simultaneously, the introduction of augmented reality interactive interfaces and blockchain technology significantly improves user experience and data reliability. Compared to traditional methods, this invention, through real-time sensor calculation of the i-angle, dynamic modeling via machine learning, lightweight deployment on mobile devices, and multi-station collaborative correction fusion, compresses the entire process to within a few hours, improving overall efficiency by 80-90%. At the same time, the model's self-learning capability reduces manual intervention by more than 70% and improves accuracy by 20-30%, providing an intelligent solution for the modern engineering surveying field.

[0090] On the other hand, the present invention also provides an i-angle adaptive calibration system for a level, the system comprising: A level instrument, set at a pre-set monitoring station, is used to collect the measured value of the i-angle of the target area; Environmental sensing devices are used to collect environmental parameters of the target area; Positioning device, used to collect positioning information of target area; A central processing unit is used to obtain a calibrated i-angle value using an adaptive i-angle calibration method for a level as described in any of the preceding claims, and to calibrate the level based on the calibrated i-angle value.

[0091] It is understood that the central processing unit can be a central processing unit (CPU) or a microcontroller (MCU).

[0092] The central processing unit may further include hardware chips. These hardware chips may be application-specific integrated circuits (ASICs), programmable logic devices (PLDs), or combinations thereof. The PLDs may be complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), generic array logic (GALs), or any combination thereof.

[0093] The central processing unit may further include memory for storing the aforementioned information, the program of the central processing unit, etc. The memory may include volatile memory, such as random-access memory (RAM); the memory may also include non-volatile memory, such as flash memory, hard disk drive (HDD), or solid-state drive (SSD); the memory may also include combinations of the above types of memory.

[0094] In one possible implementation, the i-angle adaptive calibration system for a level provided in this embodiment of the invention can also be integrated into a system with multiple functional modules. For example, it can be a mobile terminal system including an environmental sensor module, a machine learning modeling module, a mobile lightweight model deployment module, a multi-station collaborative correction module, and a cross-regional adaptive learning module. The mobile terminal system collects data such as temperature, humidity, and air pressure in real time through the environmental sensor module, predicts i-angle changes in conjunction with the machine learning modeling module, and can deploy a lightweight model to achieve mobile intelligent analysis. The multi-station collaborative correction module uses Kalman filtering to fuse measured and predicted data, dynamically calibrating the i-angle error. Simultaneously, relying on the cross-regional adaptive learning module, a transfer learning model is used to optimize the adaptability to different geographical environments, enabling the system to have cross-regional adaptive capabilities and automatically identify new environmental characteristics and update model parameters. The system integrates visualization, anomaly detection, automatic calibration, and blockchain evidence storage functions, constructing a complete closed-loop control system from environmental perception to intelligent prediction, which, through dynamic verification, ultimately achieves autonomous evolution.

[0095] Understandably, the integrated mobile terminal system possesses functions such as real-time environmental data reception, online model inference, and GPS location services, enabling real-time prediction of the i-angle and identification of regional features on-site. The real-time data acquisition module receives environmental sensor data in real time via an integrated Bluetooth / Wi-Fi interface.

[0096] The sampling frequency per second meets the accuracy requirements of on-site leveling measurements. The online model inference module runs a lightweight model on a mobile device, based on the current time-series window input. Perform online predictions:

[0097] in It is a pruned / quantized model. The predicted i-angle value.

[0098] The region identification and adaptation module uses the GPS positioning module to obtain the current coordinates:

[0099] Associate the location with the historical environment-i-angle mapping model to assist in dynamically switching or adjusting the model parameters:

[0100] The visualization and prompting module displays the current environmental status, the predicted i-angle value, and historical trends. It also uses anomaly thresholds. Provide early warning for deviations of the i-angle, when The system will prompt the operator or trigger automatic calibration to ensure measurement accuracy and system reliability.

[0101] It should be understood that the system disclosed in the embodiments of the present invention can be implemented in other ways. For example, the division of modules is merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the communication connection between units may be through some interfaces, servers, or indirect coupling or communication connections, and may be electrical or other forms.

[0102] Furthermore, in the embodiments of the present invention, the functional modules can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0103] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0104] Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner for ease of understanding.

[0105] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for adaptive calibration of the i-angle of a level, characterized in that, The method includes: The environmental parameters of the target area are obtained, and the measured value of angle i and the observation deviation of angle i are obtained based on the preset monitoring stations in the target area. The measured value of angle i and the environmental parameters are input into the trained prediction model to obtain the predicted value of angle i. The optimal estimation is performed based on the measured value of angle i, the observed deviation of angle i, and the predicted value of angle i to obtain the calibrated angle i value; The preset monitoring stations include at least two.

2. The adaptive calibration method for the i-angle of a level instrument according to claim 1, characterized in that, The optimal estimation uses the Kalman filter algorithm.

3. The adaptive calibration method for the i-angle of a level instrument according to claim 1, characterized in that, The training process of the prediction model includes: In a stable operating area, the measured value of the i-angle is obtained through monitoring stations as a training label, and the corresponding environmental training parameters are recorded synchronously. A training dataset is constructed based on the training label and the environmental training parameters. The training dataset is input into the LSTM neural network for training to obtain the trained prediction model.

4. The adaptive calibration method for the i-angle of a level instrument according to claim 3, characterized in that, Constructing the training dataset specifically includes: After performing a quality check on the environmental training parameters, normalization is then used to eliminate the influence of dimensions, and a multidimensional dataset is generated. The quality inspection includes: Savitzky-Golay filtering is used to reconstruct the time series for the missing time series values ​​in the environmental training parameters. Through 3 The principle is to identify outliers in the corresponding environmental training parameters and correct or remove data points that exceed the mean ± 3 times the standard deviation. The normalization uses the MinMax normalization method to scale the training parameters of each environment after quality check to the [0,1] interval.

5. The adaptive calibration method for the i-angle of a level instrument according to claim 4, characterized in that, The method further includes: Based on the location information of the target area, the target area is identified to be in different geographical environments. Through transfer learning technology, the parameters of the trained prediction model are adjusted by combining the measured i-angle values ​​collected in the corresponding geographical environment and the environmental training parameters.

6. The adaptive calibration method for the i-angle of a level instrument according to claim 5, characterized in that, The method further includes: Construct a regional feature knowledge base, and encapsulate the adjusted prediction model and optimal estimation process under different geographical environments in the regional feature knowledge base; Obtain the positioning information of the target area, match the corresponding adjusted prediction model based on the positioning information, and perform i-angle calibration.

7. The adaptive calibration method for the i-angle of a level instrument according to claim 5, characterized in that, The environmental parameters are the same as those used in environmental training, including temperature, humidity, and air pressure.

8. The adaptive calibration method for the i-angle of a level instrument according to claim 5, characterized in that, The method further includes: The trained prediction model is then subjected to lightweight modification; the lightweight modification includes model pruning and weight quantization.

9. A leveling instrument i-angle adaptive calibration system, characterized in that, The system includes: A level instrument, set at a pre-set monitoring station, is used to collect the measured value of the i-angle of the target area; Environmental sensing devices are used to collect environmental parameters of the target area; Positioning device, used to collect positioning information of target area; A central processing unit is used to obtain a calibrated i-angle value using an adaptive calibration method for a level as described in any one of claims 1 to 8, and to calibrate the level based on the calibrated i-angle value. The preset monitoring stations include at least two.

Images