Calibration method and system of high-precision inclination angle measurement sensor

Abstract

The invention relates to the technical field of inclination angle measurement sensors, and provides a calibration method and system for a high-precision inclination angle measurement sensor, and the method comprises the steps: obtaining the original output data of the high-precision inclination angle measurement sensor at a plurality of known attitude angles, so as to construct a multi-dimensional attitude mapping table, carrying out the error gravity center vector analysis of the multi-dimensional attitude mapping table, and carrying out the calibration of the high-precision inclination angle measurement sensor. After an error mapping graph is generated, boundary deviation suppression is carried out, after a correction graph is obtained, an inclination angle residual error regulation and control mapping relation is constructed, multi-dimensional partition adjustment and optimization are carried out on the inclination angle residual error regulation and control mapping relation, a residual error calibration function set is constructed, and output data of the high-precision inclination angle measurement sensor are calibrated based on the residual error calibration function set. The calibration precision of the inclination angle measurement sensor is improved from multiple dimensions, the influence of static errors and dynamic errors is effectively eliminated, the use reliability and applicability of the sensor are enhanced, and the problems that in a high-precision scene, error modeling is not accurate enough and the anti-interference capability is insufficient in the prior art are solved.

Description

Technical Field

[0001] This application relates to the technical field of inclination measurement sensors, and in particular to a calibration method and system for high-precision inclination measurement sensors. Background Art

[0002] As a sensor widely used in the fields of industrial automation, building monitoring, aerospace, intelligent devices, etc., the inclination measurement sensor can perceive and measure the inclination change of equipment or objects in space in real time. However, due to the influence of factors such as sensor manufacturing deviation, environmental interference (including temperature, vibration, etc.), and multi-axis error coupling during the inclination measurement process, the calibration of the inclination measurement sensor has become a technical difficulty.

[0003] In related technical means, the calibration of the inclination measurement sensor usually involves collecting the inclination data output by the sensor during testing on a standard platform, then performing static fitting on the collected data, comparing it with the standard reference value, extracting the static error characteristics, and using the static error correction formula to batch-correct multiple groups of test data. By rotating the posture of the sensor and collecting its outputs in various directions in a dynamic environment, the dynamic error characteristics, including non-linear drift, time delay, etc., are calculated to achieve relatively accurate inclination measurement.

[0004] For the above technical solutions, although the inclination measurement sensor can be effectively calibrated under low-precision requirements by collecting the data output by the standard platform and performing static and dynamic error fitting and correction, in high-precision scenarios, the existing technical methods have the following problems: on the one hand, the multi-dimensional error coupling and dynamic characteristic changes of the inclination measurement sensor are difficult to be fully described by simple linear fitting or a limited discrete mapping table, resulting in inaccurate error modeling; on the other hand, in terms of multi-axis coupling error processing and drift correction under complex dynamic conditions, the existing methods have insufficient robustness (anti-interference ability), resulting in the inability to correct multi-dimensional errors collaboratively and affecting the high-precision measurement results. Summary of the Invention

[0005] In order to improve the problems of inaccurate error modeling and insufficient anti-interference ability existing in the prior art in high-precision scenarios, this application provides a calibration method and system for high-precision inclination measurement sensors.

[0006] The present invention provides a calibration method for a high-precision inclination measurement sensor, including: obtaining the original output data of the high-precision inclination measurement sensor at multiple known attitude angles, generating an initial error map according to the original output data, and extracting an error distribution sequence based on the initial error map; performing piecewise fitting and trend normalization on the error distribution sequence to obtain a direction weight table, constructing a multi-dimensional attitude mapping table based on the direction weight table, performing error center-of-gravity vector analysis on the multi-dimensional attitude mapping table to generate an error mapping map; suppressing boundary deviations of the error mapping map according to preset constraint rules to generate an error suppression map, performing multi-axis matching and error energy difference compensation on the error suppression map to obtain a reconstructed reference set, and using the reconstructed reference set to perform back-projection correction on the error suppression map to obtain a corrected map; constructing an inclination residual regulation mapping relationship through the corrected map, performing multi-dimensional partition optimization on the inclination residual regulation mapping relationship to construct a residual calibration function set, and calibrating the output data of the high-precision inclination measurement sensor based on the residual calibration function set.

[0007] As a preferred solution, the steps of obtaining the original output data of the high-precision inclination measurement sensor at multiple known attitude angles, generating an initial error map according to the original output data, and extracting an error distribution sequence based on the initial error map include: using a preset multi-channel data acquisition system to collect the original output data of the high-precision inclination measurement sensor at multiple known attitude angles, performing attitude calibration and synchronous sampling on the original output data to obtain an attitude label set and a corresponding inclination response sequence; performing statistical debiasing and amplitude normalization on the inclination response sequence to obtain a data equilibrium set and an initial difference set, constructing an error field mapping array using the attitude label set and the initial difference set, and extracting an error perturbation intensity spectrum and a direction drift trajectory sequence according to the error field mapping array; constructing an initial error map using the error perturbation intensity spectrum and the data equilibrium set, and combining the initial error map with the direction drift trajectory sequence to extract an error distribution sequence.

[0008] As a preferred solution, the steps of segmenting and fitting the error distribution sequence and normalizing the trend to obtain a direction weight table, constructing a multi-dimensional pose mapping table based on the direction weight table, and performing error centroid vector analysis on the multi-dimensional pose mapping table to generate an error mapping atlas include: dividing the interval of the error distribution sequence and performing least-squares polynomial fitting to obtain a set of fitting functions, performing local normalization and trend balancing processing on the error distribution sequence based on the set of fitting functions to obtain a set of normalized residuals; using sliding window difference analysis and mutation detection algorithms to extract the residual fluctuation peak and valley points of the set of normalized residuals, constructing a set of feature points according to the residual fluctuation peak and valley points, constructing a set of drift correction factors and a direction weight table according to the set of feature points, and constructing a multi-dimensional pose mapping table using the direction weight table; using a preset threshold change rate rule to compare the residual change amplitude between adjacent feature points in the set of feature points to construct a section change sequence; using the section change sequence to perform clustering reconstruction on the multi-dimensional pose mapping table, extracting the error centroid vector sets of each pose block, and combining the error centroid vector sets with the set of drift correction factors to generate an error mapping atlas.

[0009] As a preferred solution, the steps of using the section change sequence to perform clustering reconstruction on the multi-dimensional pose mapping table, extracting the error centroid vector sets of each pose block, and combining the error centroid vector sets with the set of drift correction factors to generate an error mapping atlas include: using the section change sequence to cluster the error data in the multi-dimensional pose mapping table to obtain a plurality of pose blocks; calculating the error centroid vectors according to the error data in each pose block to obtain a set of error centroid vectors, combining the set of error centroid vectors with the set of drift correction factors to generate a number of error mapping data, and summarizing all the error mapping data to obtain an error mapping atlas.

[0010] As a preferred solution, the step of suppressing boundary deviation of the error mapping atlas according to a preset constraint rule to generate an error suppression atlas, performing multi-axis matching and error energy difference compensation on the error suppression atlas to obtain a reconstructed reference set, and using the reconstructed reference set to perform back-projection correction on the error suppression atlas to obtain a corrected atlas includes: suppressing boundary deviation of the error mapping atlas according to a preset constraint rule, extracting the residual distribution in the edge transition region, and constructing a boundary optimization set according to the residual distribution in the edge transition region; performing error density fitting and principal component analysis in the local angular domain on the boundary optimization set to generate an angular domain weighting matrix; generating an error suppression atlas by using the boundary optimization set, extracting a stable region based on the error suppression atlas, constructing a neutral error region according to the stable region, and extracting an error feature vector from the neutral error region; aligning and weighting the error feature vector with the angular domain weighting matrix to construct a reconstructed reference set, and performing regional back-projection correction on the error suppression atlas by using the reconstructed reference set to generate a corrected atlas.

[0011] As a preferred solution, the step of aligning and weighting the error feature vector with the angular domain weighting matrix to construct a reconstructed reference set, and performing regional back-projection correction on the error suppression atlas by using the reconstructed reference set to generate a corrected atlas includes: performing multi-axis rotation matching on the error data of the error feature vector at different pose angles with each pose in the angular domain weighting matrix to obtain an alignment result; performing weighting processing on the errors in different axes of the alignment result to obtain a weighted error result, and applying the weighted least squares method to calculate the error energy difference of each pose in different directions in the weighted error result to construct a reconstructed reference set; performing back-projection on the error correction factors in the reconstructed reference set and the regional errors in the error suppression atlas to generate a corrected atlas.

[0012] As a preferred solution, the steps of constructing an inclination residual regulation mapping relationship through the correction map, performing multi-dimensional partition optimization on the inclination residual regulation mapping relationship to construct a set of residual calibration functions, and calibrating the output data of the high-precision inclination measurement sensor based on the set of residual calibration functions include: using the correction map to construct an initial inclination residual matrix and a position corresponding index table, and extracting multiple groups of attitude error difference sequences and angle delay factors based on the position corresponding index table; performing gradient clustering and direction fitting on the multiple groups of attitude error difference sequences to obtain an inclination residual regulation mapping relationship and a direction feedback vector group; using the angle delay factor and the direction feedback vector group to perform multi-dimensional partition optimization on the inclination residual regulation mapping relationship to obtain multiple groups of residual optimization coefficients; generating a set of residual calibration functions according to the multiple groups of residual optimization coefficients and the initial inclination residual matrix, and calibrating the output data of the high-precision inclination measurement sensor by using the set of residual calibration functions.

[0013] The present application also provides a calibration system for a high-precision inclination measurement sensor, including: an acquisition module, configured to acquire the original output data of the high-precision inclination measurement sensor at multiple known attitude angles, generate an initial error map according to the original output data, and extract an error distribution sequence based on the initial error map; an analysis module, configured to perform piecewise fitting and trend normalization on the error distribution sequence to obtain a direction weight table, construct a multi-dimensional attitude mapping table based on the direction weight table, perform error centroid vector analysis on the multi-dimensional attitude mapping table, and generate an error mapping map; a correction module, configured to suppress boundary deviation of the error mapping map according to a preset constraint rule to generate an error suppression map, perform multi-axis matching and error energy difference compensation on the error suppression map to obtain a reconstruction reference set, and perform back-projection correction on the error suppression map by using the reconstruction reference set to obtain a correction map; a calibration module, configured to construct an inclination residual regulation mapping relationship through the correction map, perform multi-dimensional partition optimization on the inclination residual regulation mapping relationship to construct a set of residual calibration functions, and calibrate the output data of the high-precision inclination measurement sensor based on the set of residual calibration functions.

[0014] Compared with the prior art, the present application has the following beneficial effects: accurate calibration and strong anti-interference ability. By obtaining the original output data of a high-precision inclination measurement sensor at multiple known attitude angles and generating an initial error map, segment fitting and trend normalization are performed on the error distribution sequence to construct a direction weight table and a multi-dimensional attitude mapping table, and then the error distribution is globally described by generating an error mapping map; the error mapping map is processed by using boundary deviation suppression and multi-axis matching means to generate an error suppression map, and then the error suppression map is corrected by back-projection through reconstructing a reference set to generate a correction map for calibration. Finally, an inclination residual control mapping relationship is constructed through the correction map, multi-dimensional partition optimization is performed, and a set of residual calibration functions is constructed to calibrate the sensor output data, so as to comprehensively improve the calibration accuracy of the inclination measurement sensor from multiple dimensions, effectively eliminate the influence of static errors and dynamic errors, enhance the reliability and applicability of sensor use, meet the measurement requirements of high-precision scenarios, and improve. However, in high-precision scenarios, the prior art has problems of inaccurate error modeling and insufficient anti-interference ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0016] The structures, proportions, sizes, etc. shown in the drawings of this specification are only used to cooperate with the content disclosed in the specification for those who are familiar with this technology to understand and read, and are not used to limit the limiting conditions under which the present invention can be implemented. Therefore, they do not have a substantial technical meaning. Any modification of the structure, change of the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope that can be covered by the technical content disclosed in the present invention.

[0017] Figure 1 is a schematic flowchart of a calibration method for a high-precision inclination measurement sensor provided by an embodiment of the present invention; Figure 2 is a schematic block diagram of the structure of a calibration system for a high-precision inclination measurement sensor provided by an embodiment of the present invention.

[0018] DESCRIPTION OF THE REFERENCE NUMERALS 10. Calibration system for high-precision inclination measurement sensor; 11. Acquisition module; 12. Analysis module; 13. Correction module; 14. Calibration module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] The flowcharts shown in the accompanying drawings are only illustrative examples, and do not necessarily include all the contents and operations / steps, nor do they necessarily need to be executed in the described order. For example, some operations / steps can also be decomposed, combined or partially merged, so the actual execution order may be changed according to the actual situation.

[0021] It should also be understood that the terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. As used in the specification of this application and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0022] It should be further understood that the term "and / or" used in the specification of this application and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0023] The technical solutions of the present invention will be further described below with reference to the accompanying drawings and through specific embodiments.

[0024] Embodiment 1: As Figure 1 shown, this application provides a calibration method for a high-precision inclination measurement sensor, including steps S100 to S400.

[0025] Step S100: Obtain the original output data of the high-precision inclination measurement sensor at multiple known attitude angles, generate an initial error map based on the original output data, and extract an error distribution sequence based on the initial error map.

[0026] In this step, a high-precision inclination measurement sensor is installed on a controllable standard attitude test platform. By placing the sensor in states of multiple preset attitude angles (such as 0°, 15°, 30°, etc.), the original output data of the inclination measurement sensor is collected in real time by a multi-channel data acquisition system, including the inclination response values and the corresponding sensor status data at each attitude angle. Specifically, the collected original output data is synchronously calibrated to ensure the consistency of time and attitude data, and the deviation of the original output data is calculated in combination with the standard reference value; based on this deviation information, an initial error map is generated to characterize the static error distribution characteristics of the sensor, and at the same time, in combination with the drift trend in data acquisition, the map is supplemented relying on the direction drift trajectory sequence to finally extract the error distribution sequence.

[0027] For example, the sensor is placed in known attitudes with a fixed inclination angle of 0° and a dynamic inclination change range of 15° to 30° through a multi-channel data acquisition system, and the original output data of the inclination measurement sensor is collected. During the collection process, a synchronous calibration algorithm (such as the average alignment algorithm) is executed in combination with the standard reference value to calculate the deviation between the original output data and the actual reference value. The original deviation data of the above static calibration can generate an initial error map, and at the same time, the trend analysis algorithm is used to detect the time drift of each inclination response value, and the error distribution sequence is obtained by correcting the deviation curve, which is finally used for the fitting process in the subsequent steps.

[0028] Step S200: Perform piecewise fitting and trend normalization on the error distribution sequence to obtain a direction weight table, construct a multi-dimensional attitude mapping table based on the direction weight table, and perform error centroid vector analysis on the multi-dimensional attitude mapping table to generate an error mapping map.

[0029] In this step, piecewise fitting processing is performed on the error distribution sequence, and the least squares method is used to calculate the error fitting function group for each segment to obtain the mathematical expression of the local trend, so as to achieve accurate fitting of the error distribution; specifically, the fitting result undergoes a trend normalization operation to eliminate the overall deviation in the error distribution and generate a direction weight table to characterize the error characteristics in different directions during the inclination change process; subsequently, the weight sequences in multiple directions are integrated and reconstructed by constructing a multi-dimensional attitude mapping table to form a global attitude data model. At the same time, the error data in the multi-dimensional attitude mapping table is classified and aggregated, and the error centroid vector in each attitude region is calculated. Finally, an error mapping map is generated based on these vectors to describe the error characteristics at each attitude.

[0030] For example, the error distribution sequence extracted in step S100 is segmented according to the inclination change range (such as 0° to 15°, 15° to 30°), and the least squares method is used to fit the error change trend of each segment to obtain the error fitting function group, and these functions can be in the form of cubic polynomials (such as It is represented by). Next, in each segmentation result, the mean normalization algorithm is used to eliminate the overall trend deviation to generate a direction weight table. The multi-dimensional attitude mapping table is constructed using the direction weight table. Through a preset region division algorithm (such as the K-means clustering algorithm), the error centroid vector within each attitude region is calculated, and these vectors are used to generate the final error mapping atlas, thereby accurately describing the inclination error distribution law of the sensor.

[0031] Step S300: Perform boundary deviation suppression on the error mapping atlas according to the preset constraint rules to generate an error suppression atlas. Perform multi-axis matching and error energy difference compensation on the error suppression atlas to obtain a reconstruction reference set. Use the reconstruction reference set to perform back-projection correction on the error suppression atlas to obtain a corrected atlas.

[0032] In this step, perform boundary deviation suppression processing on the error mapping atlas. According to the preset constraint rules, eliminate the abnormal data in the error atlas and extract the residual distribution in the edge transition region to generate a boundary optimization set. Specifically, perform optimization analysis on the boundary optimization set, reduce the boundary residual value through a local fitting algorithm, and then construct an angular domain weighting matrix to describe the influence of the inclination change on the error distribution. Subsequently, generate an error suppression atlas based on the angular domain weighting matrix, construct a neutral error region by extracting the stable region in the suppression atlas, and further extract the error feature vectors from the neutral error region. Use the error feature vectors to perform multi-axis matching with the angular domain weighting matrix to balance the dynamic coupling characteristics of the error distribution, and combine the error energy difference compensation algorithm to calculate the dynamic error correction factor, and finally construct a reconstruction reference set. Apply the reconstruction reference set to the error suppression atlas and optimize the atlas through back-projection correction to generate a new corrected atlas, comprehensively improving the calibration accuracy.

[0033] For example, extract the data in the edge transition region of the error mapping atlas, use the cubic spline fitting algorithm to eliminate the abnormal error points, and construct an angular domain weighting matrix. Adjust the sensitive weights in each direction according to the inclination change characteristics in the weighting matrix. After generating the error suppression atlas, extract the stable region through a region division algorithm (such as the maximum stable region detection method) to construct a neutral error region. Use the direction-by-point matching algorithm for multi-axis dynamic matching of the data in the neutral error region, and at the same time calculate the error energy difference compensation factor through the weighted least squares method, and finally construct a reconstruction reference set. Based on the reconstruction reference set, use the back-projection correction algorithm to optimize the error distribution and generate the final corrected atlas.

[0034] Step S400: Construct an inclination residual regulation mapping relationship through the corrected atlas, perform multi-dimensional partition optimization on the inclination residual regulation mapping relationship to construct a residual calibration function set, and calibrate the output data of the high-precision inclination measurement sensor based on the residual calibration function set.

[0035] In this step, a correction map is used to construct the inclination residual regulation mapping relationship. Key data points in the correction map are extracted, the variation law of the residuals is analyzed by a gradient clustering algorithm, and an initial inclination residual matrix is established to represent the correction residuals corresponding to each attitude angle. Specifically, the residual matrix is combined with the position corresponding index table to extract the error difference sequence and the corresponding angle delay factor. Subsequently, a direction fitting process is performed on the difference sequence to form the inclination residual regulation mapping relationship, and the mapping relationship is optimized in multiple dimensions in combination with the angle delay factor. Finally, the residual optimization coefficient is extracted to construct a set of residual calibration functions. Based on the set of residual calibration functions, the output data of the sensor is calibrated to achieve the final optimization of the sensor data.

[0036] For example, key error points are extracted from the correction map, the variation trend of the residuals is aggregated into different gradient intervals by the K-means clustering algorithm, and the inclination residual regulation mapping relationship is established using the cubic polynomial fitting method. In combination with the angle delay factor, the boundary data of the inclination residual regulation mapping relationship is corrected by a multi-dimensional partition optimization algorithm, and finally, based on the corrected mapping relationship, the residual optimization coefficient is calculated to construct a set of residual calibration functions for calibrating the sensor output data, thereby completing the calibration process.

[0037] In this embodiment, the original output data of a high-precision inclination measurement sensor at multiple known attitude angles is collected, and an initial error map is generated according to the original output data. An error distribution sequence is extracted based on the initial error map. Then, the error distribution sequence is subjected to piecewise fitting and trend normalization to obtain a direction weight table, and a multi-dimensional attitude mapping table is constructed based on the direction weight table. Error centroid vector analysis is performed on the multi-dimensional attitude mapping table to generate an error mapping map. Then, boundary deviation suppression is performed on the error mapping map according to preset constraint rules to generate an error suppression map, and multi-axis matching and error energy difference compensation are performed on the error suppression map to obtain a reconstruction reference set. The error suppression map is corrected by back projection using the reconstruction reference set to obtain a correction map. Finally, an inclination residual regulation mapping relationship is constructed through the correction map, and the inclination residual regulation mapping relationship is optimized in multiple dimensions to construct a set of residual calibration functions. Based on the set of residual calibration functions, the output data of the high-precision inclination measurement sensor is calibrated.

[0038] Through a step-by-step calibration strategy and a gradual optimization method for the entire process, significant effects are demonstrated in the calibration of static errors and dynamic errors. By generating an initial error map and extracting the error distribution sequence, the error characteristics of the original output data can be accurately analyzed at multiple known attitude angles; by using piecewise fitting and trend normalization to obtain a direction weight table and constructing a multi-dimensional attitude mapping table, the correlation between multi-dimensional inclination data can be fully captured; based on the boundary deviation suppression of the error mapping map and the process of multi-axis matching and error energy difference compensation, the dynamic coupling errors in complex environments are further eliminated; by reconstructing the reference set for back-projection correction and generating a correction map, the overall accuracy of inclination measurement is effectively improved; finally, an inclination residual regulation mapping relationship is generated and a residual calibration function set is applied to calibrate the output data of the high-precision inclination measurement sensor, which can significantly reduce the measurement error, enhance the system robustness and applicability, so as to meet the usage requirements of high-precision scenarios. However, in high-precision scenarios, the existing technologies have problems of inaccurate error modeling and insufficient anti-interference ability.

[0039] Embodiment 2: In step S100, a preset multi-channel data acquisition system is used to collect the original output data of the high-precision inclination measurement sensor at multiple known attitude angles, and the attitude calibration and synchronous sampling are performed on the original output data to obtain an attitude label set and the corresponding inclination response sequence.

[0040] By using the multi-channel data acquisition system to sample the high-precision inclination measurement sensor on a controllable standard platform, the system includes multiple channels for recording the output data of the inclination measurement sensor at multiple attitude angles and the external experimental environment data (such as temperature, vibration force, etc.). Specifically, by placing the sensor at positions of at least three or more known attitude angles (for example, the earth horizontal 0°, standard inclination 15°, 30°, 45°), collecting its output inclination response values, and associating them with the standard reference attitude angle values in real time, the data labels of each corresponding attitude are obtained; during the sampling process, the output data is calibrated through a synchronous sampling algorithm, which ensures that the inclination response values at the time points are consistent and synchronous with the attitude label set, so as to obtain the attitude label set and the corresponding inclination response sequence.

[0041] For example, a multi-channel data acquisition system can record the multi-axis response outputs of sensors in real time (such as X-axis inclination values, Y-axis inclination values, etc.) and specific tags (such as 0°, 15°, 30°). During specific sampling, the attitude calibration system dynamically compares the standard reference inclination values with the sensor output values and adjusts the time series of the sensor inclination response values using a synchronous alignment algorithm. For example, in actual tests, the response values output by the sensor are a set of data in units of angles. For example, the inclination response sequence at 0° is x1, x2, x3,... and the corresponding response sequence at 15° is y1, y2, y3,... This set of response values is precisely matched with the time points and inclination tags, thereby constructing a complete set of attitude tags and the corresponding inclination response sequences.

[0042] Perform statistical debiasing and amplitude normalization on the inclination response sequences to obtain a data equilibrium set and an initial difference set. Use the set of attitude tags and the initial difference set to construct an error field mapping array, and extract the error perturbation intensity spectrum and the direction drift trajectory sequence according to the error field mapping array.

[0043] By processing the inclination response sequences with statistical debiasing algorithms and amplitude normalization algorithms, the inclination responses of each known attitude are calibrated into a value form within a unified range, so that the multi-dimensional inclination responses are all consistent, thereby eliminating the systematic errors caused by the influence of the external experimental environment. Specifically, the statistical debiasing algorithm adjusts the average value offset of the inclination response sequence to zero to eliminate the constant bias in the static inclination output. For example, the mean offset of the inclination response sequence can be calculated. , and correct all inclination outputs to ; At the same time, through the amplitude normalization algorithm, the value ranges of each inclination response sequence are scaled to a unified standard range [0, 1] for subsequent calculations. Combining the above processing, a data equilibrium set is generated. At the same time, by analyzing the differences between the data equilibrium set and the set of attitude tags, the initial difference characteristics are extracted and an error field mapping array is constructed. This array is used to represent the error variation law within each attitude angle range. Finally, perform a frequency domain analysis algorithm through the error field mapping array to extract the error perturbation intensity spectrum; extract the drift curves in each attitude direction through a trajectory modeling algorithm to generate a direction drift trajectory sequence.

[0044] For example, for the inclination response sequence (where is the inclination value at a specified time point), use the statistical debiasing method to calculate the mean , and adjust each inclination value to the debiased value . At the same time, perform amplitude normalization processing on the debiased inclination values. Through the formula: Standardize the input values so that all inclination values fall within the range [0, 1]. Finally, obtain a data equilibrium set. After that, the data balance set and the pose label set are subjected to differential matching, and data processing is carried out relying on the error field mapping array. For example, the dynamic error characteristics can adopt the Fourier frequency domain analysis method to extract the error perturbation intensity spectrum, and the direction drift sequence can be recorded by straight line fitting trajectory modeling.

[0045] An initial error map is constructed by using the error perturbation intensity spectrum and the data balance set, and the initial error map is combined with the direction drift trajectory sequence to extract the error distribution sequence.

[0046] By taking the error perturbation intensity spectrum and the data balance set as the static error and the dynamic drift source respectively, the error characteristics are classified to construct the initial error map; specifically, the multi-dimensional partition analysis is carried out on the error perturbation intensity spectrum by using the static error distribution extraction algorithm, and the error amplitudes at different frequencies are represented by a curve calibrator, so as to generate the static map module; at the same time, the dynamic pose modeling is carried out through the direction drift trajectory sequence, the drift data is matched with the tilt pose label, and the dynamic map drift characteristic module is constructed. Finally, relying on the static map module and the dynamic map drift module, a complete initial error map is generated; based on the initial error map combined with the dynamic and static error characteristics, the map information is refined, and the error distribution sequence is determined through trend calculation.

[0047] For example, for the error perturbation intensity spectrum (representing the intensity change data of different frequency domains), through partition analysis, a cubic polynomial fitting is adopted to generate the form of the static map , and the drift characteristic data is calculated through the dynamic modeling of the direction drift trajectory sequence. For example, the direction change curve can be expressed as , where is the amplitude factor, is the phase offset. Through the combination of static and dynamic errors, an initial error map is finally generated, and the trend analysis of the map information is carried out, and the extracted error characteristics are generated into an error distribution sequence.

[0048] In step S200, the error distribution sequence is subjected to interval division and least square polynomial fitting to obtain a set of fitting functions, and based on the set of fitting functions, local normalization and trend balance processing are carried out on the error distribution sequence to obtain a group of normalized residuals.

[0049] By carrying out interval division on the error distribution sequence, the error sequence is segmented into multiple intervals by using the interval division algorithm, usually according to the error amplitude or the distribution uniformity of data points. The error data in each interval will be respectively subjected to least square polynomial fitting. Specifically, a quadratic or cubic polynomial (for example to fit the error change trend within the interval. In this way, the local trend of the error distribution can be accurately captured, and a corresponding set of fitting functions can be generated for each interval. After fitting, the fitting results of each interval are locally normalized using a normalization algorithm to eliminate the differences in absolute error values between different intervals and perform trend balancing processing to ensure that the error sequences of all intervals are processed under the same dimension. Finally, a set of normalized residuals is obtained as the basis for subsequent processing.

[0050] For example, for the error distribution sequence where is the error at each sampling point, first, the error sequence is divided into intervals. For example, it is divided into three intervals: Interval 1, Interval 2, and Interval 3. For the data within each interval, the least squares fitting formula is used for polynomial fitting to obtain a set of fitting functions , which can accurately describe the error distribution of each interval. After fitting, the error is normalized using the normalization formula to obtain a set of normalized residuals .

[0051] The residual fluctuation peak and valley points of the set of normalized residuals are extracted using a sliding window difference analysis and mutation detection algorithm. A set of feature points is constructed based on the residual fluctuation peak and valley points. A set of drift correction factors and a direction weight table are constructed based on the set of feature points. A multi-dimensional attitude mapping table is constructed using the direction weight table.

[0052] Through the sliding window difference analysis algorithm, the set of normalized residuals is divided into multiple windows of a fixed length. The difference and fluctuation trend of the residual values are calculated within each window. Specifically, the sliding window algorithm calculates the error difference for each window in turn and records the fluctuation information within each window. When the difference exceeds a set threshold, it is marked as a mutation point, that is, a residual fluctuation peak and valley point. Next, a clustering analysis is performed on all marked fluctuation peak and valley points using a mutation detection algorithm to extract a set of feature points, which represent important change intervals in the error distribution. Based on the set of feature points, drift correction factors are calculated, and a direction weight table is generated. The direction weight table is used to measure the sensitivity of errors in different directions. Finally, in combination with the direction weight table, a multi-dimensional attitude mapping table is constructed to associate the error characteristics of each attitude with their corresponding weight values to form a high-dimensional data model.

[0053] For example, considering the set of normalized residuals , , …, , a difference analysis is performed using a sliding window size of 5. For the error values within the window, each difference is calculated, and when the difference exceeds a set threshold (for example When it is > 0.1), it is marked as a mutation point. For example, assume that at a certain time point , the error difference is greater than the threshold of 0.1, then is regarded as a peak or valley value and is one of the feature points. Through cluster analysis, find the most important set of feature points , and calculate the drift correction factor set according to these feature points, which is used to describe the drift behavior at each attitude angle. Subsequently, by constructing a direction weight table and combining it with the set of feature points, a multi-dimensional attitude mapping table is generated, associating the error characteristics of each attitude with its weight, thereby obtaining the overall error model.

[0054] Use the preset threshold change rate rule to compare the residual change amplitude between adjacent feature points in the set of feature points, and extract the section index with significant residual change to construct a section change sequence.

[0055] By applying the threshold change rate rule to perform step-by-step analysis on the set of feature points, compare the residual change amplitude between adjacent feature points. Specifically, set a predefined change rate threshold (for example, the change rate threshold is 20%), and calculate the residual change amplitude between adjacent feature points. When the change amplitude between adjacent feature points exceeds the threshold, it is marked as a section with significant change, and the index of this section is extracted to form a section change sequence. The section change sequence provides key information for subsequent error analysis and reconstruction processes.

[0056] For example, for the set of feature points , set the change rate threshold to 20%. Calculate the residual change amplitude between adjacent feature points. For example, for adjacent feature points and , calculate their residual change amplitude: If > 0.2, then it is considered that and have a significant change, mark this section as a changed section and record its index. For example, if the section change occurs between , then the section change sequence is 1, 2, 3.

[0057] Use the section change sequence to perform cluster reconstruction on the multi-dimensional attitude mapping table, extract the set of error center of gravity vectors for each attitude block, and combine the set of error center of gravity vectors with the drift correction factor set to generate an error mapping atlas.

[0058] By combining the section change sequence with the multi-dimensional attitude mapping table, the K-means clustering algorithm is used to cluster and reconstruct the error data, ensuring that attitude data with similar error patterns are grouped into the same category. Specifically, based on the significant change sections in the section change sequence, the K-means clustering algorithm groups its error data and calculates the error centroid vector within each group. The error centroid vector is the weighted average of all attitude error values within this block, and the weights are dynamically adjusted according to the magnitude or stability of the errors. Finally, these error centroid vectors are combined with the set of drift correction factors, and the error mapping data is generated through a weighted fusion algorithm. All the error mapping data is summarized to obtain the complete error mapping atlas.

[0059] For example, for a known attitude data set , first, clustering is performed according to the section change sequence. For example, the data is divided into multiple categories through K-means clustering, and each category represents an attitude block with similar error characteristics. For each attitude block, its error centroid vector (where is the error value of each attitude within this block, is the number of samples within the block) is calculated to obtain the set of error centroid vectors for this block. Then, this set of error centroid vectors is combined with the set of drift correction factors , and the weighted fusion algorithm (where is the weight coefficient, is the error centroid vector) is used to generate the error mapping data. Finally, all the error mapping data is summarized to obtain the error mapping atlas.

[0060] Among them, the steps of using the section change sequence to cluster and reconstruct the multi-dimensional attitude mapping table, extracting the set of error centroid vectors for each attitude block, and combining the set of error centroid vectors with the set of drift correction factors to generate the error mapping atlas specifically include: Using the section change sequence to cluster the error data in the multi-dimensional attitude mapping table, and using the K-means clustering algorithm to group attitude data with similar error patterns into the same category to obtain multiple attitude blocks.

[0061] Classify and reconstruct the error data in the multi-dimensional attitude mapping table through the section change sequence. By analyzing the error change patterns of the significant sections in the section change sequence, after clarifying the distribution characteristics of the data, divide the error data in the mapping table. Specifically, select the K-means clustering algorithm to perform clustering analysis on the error data of each attitude point, set the number of clustering centers according to the input section change sequence, and divide the attitude points into multiple categories according to the similarity of the error patterns. Each category represents an attitude block, which can initially describe the attitude point area with similar error characteristics. The clustering process classifies according to the distance between points (such as Euclidean distance), minimizing the distance difference of data points within each category, thereby ensuring the error consistency within the block.

[0062] For example, assume that the section change sequence is composed of the indices of significant change sections, such as the sequence 1, 2, 3, 7, 8, 9... For the error data in the multi-dimensional attitude mapping table (for example, the error value of each attitude point is a multi-dimensional vector ), perform clustering analysis using the K-means clustering algorithm. First, select the number of significant sections in the section change sequence as the initial number of clustering centers , and the error data of each attitude point is regarded as a point. Through the Euclidean distance formula: where, represents the error value of the attitude point, is the average error of the clustering center. The algorithm iteratively optimizes the clustering center and finally divides into multiple attitude blocks. For example, block 1 includes points 1, 2, 3, and block 2 includes points 7, 8, 9, obtaining the attitude block data after division.

[0063] Calculate the error center of gravity vector according to the error data within each attitude block, that is, calculate the weighted average of each attitude error within each block, where the weight can be dynamically adjusted according to the size or stability of the error, obtaining a set of error center of gravity vectors. Combine the set of error center of gravity vectors with the set of drift correction factors, and use the weighted fusion algorithm to fuse the error center of gravity of each attitude block with the corresponding drift correction factor to generate several error mapping data, and summarize all the error mapping data to obtain the error mapping atlas.

[0064] Perform weighted average processing on the data of each attitude block through the calculation of the error center of gravity vector to accurately express the error characteristics within the block. Specifically, for all attitude error values within the block, assign a dynamic weight to each data point, and the weight can be adjusted based on the size of the error (for example, the larger the error amplitude, the smaller the weight) or the stability of the error performance (for example, the higher the stability, the larger the weight). Through the formula: where, Represents the error centroid vector of the block, is the weight of the data point, is the error vector of the data point within the block. After obtaining the set of error centroid vectors, it is combined with the set of drift correction factors, and data integration is performed through a weighted fusion algorithm. The formula for weighted fusion is: where, is the fused error mapping data, is the error centroid vector of the block, represents the drift correction factor. Finally, the error mapping data of all blocks are summarized to obtain a complete error mapping atlas, which can comprehensively express the calibration rules of different blocks in the multi-dimensional error space.

[0065] For example, assume that block 1 includes the set of attitude point errors , and block 2 includes . For block 1, the error centroid is calculated by weighted average: where is the weight value, is the maximum error value within the block. For block 2, the calculation result is: After that, the error centroid vector is combined with the set of drift correction factors , and the weighted fusion formula is used: All the error mapping data are summarized to obtain an error mapping atlas for subsequent calibration planning.

[0066] In step S300, according to the preset constraint rules, boundary deviation suppression is performed on the error mapping atlas, the residual distribution in the edge transition region is extracted, and a boundary optimization set is constructed based on the residual distribution in the edge transition region.

[0067] By performing boundary deviation analysis on the error mapping atlas, according to the preset constraint rules (such as the upper and lower limits of error amplitude or the boundary gradient change rule), the abnormal deviation values in the boundary region of the error mapping atlas are detected and suppressed. Such abnormal deviations usually appear in the edge region of the mapping atlas, so they are classified as the edge transition region. Specifically, by calculating the residual distribution characteristics in the boundary region, such as amplitude gradient, change rate, and local drift trend, the residual data points in the edge transition region are extracted; then, the residual data points are normalized and cluster analyzed to form a boundary optimization set, providing stable residual basic data for subsequent boundary optimization fitting and redistribution processing.

[0068] For example, for the error mapping atlas, consider the data distribution in its boundary region as , and these data have significant deviation values. According to the preset amplitude limit rule (e.g., the absolute value of the error ) and the gradient change rule (e.g., the gradient change rate in the boundary region ), the data points that meet the rules are extracted, and their corresponding residual values are recorded. For example, for a set of data 10, 12, 25, 15, 10, 8, the point at the data "25" is regarded as an outlier; the remaining data meet the rules by calculating the gradient change, and these qualified data points are added to the boundary optimization set for subsequent processing.

[0069] Perform error density fitting and principal component analysis in the local angular domain on the boundary optimization set to generate an angular domain weighting matrix.

[0070] By performing error density fitting on the residual data in the boundary optimization set, a fine model of the error distribution within the boundary region is built; specifically, solve the density function of the error between adjacent points, and represent the local discrete residual data as a set of continuous density curves (such as Gaussian models) through the least squares fitting method. In addition, the principal component analysis (PCA) algorithm is also used. By defining the principal direction vector, the errors in multiple directions are redistributed and mapped into a unified angular domain to more accurately describe the error characteristics of the boundary region. Finally, combining the overall density fitting and directional redistribution, an angular domain weighting matrix is generated, and this matrix is used to assign weights to the errors in different directions, reflecting the influence degree of the errors in each direction on the calibration.

[0071] For example, assume that the residual values in the boundary optimization set are , and continuous density curves are generated in the form of least squares fitting. For example, the density function is: where is the peak constant, is the mean of the residuals, is the standard deviation of the residuals. Then, perform principal component analysis on the density function, and normalize and map the residuals through the principal direction vector , for example: where is the angle between the deviation vector and the principal direction. Combine the above directionally corrected values to generate an angular domain weighting matrix for subsequent calibration calculations.

[0072] Generate an error suppression atlas using the boundary optimization set, extract the stable region based on the error suppression atlas, construct a neutral error region according to the stable region, and extract the error feature vector from the neutral error region.

[0073] An error suppression map is generated by processing the entire error mapping map using the generated boundary optimization set. The error suppression map smooths the boundary region using an interpolation method and feeds back the residual optimization result into the error map to suppress the influence of abnormal boundary values. Specifically, stable regions in the map (i.e., regions where the error amplitude changes little and has high consistency) are identified, and these stable regions are defined as neutral error regions. Further, error feature vectors are extracted from the neutral error regions as calibration important data inputs. The error feature vectors are data groups containing key error information, which are used to represent the error distribution in a multi-dimensional space and assist in constructing a calibration model.

[0074] For example, in the error suppression map, the stable region can be determined by calculating the change rate of the error, for example: where is the maximum allowable change rate threshold (e.g., 0.1). Assuming the position of the stable region in the map is the interval , then the neutral error region can be calibrated as this part of the data in the map . Feature vectors such as standard deviation, mean, or key vectors after principal component analysis are extracted from the neutral error region for subsequent processing.

[0075] The error feature vectors are aligned and weighted with the angular domain weighting matrix to construct a reconstruction reference set, and the error suppression map is corrected by region back-projection using the reconstruction reference set to generate a corrected map.

[0076] By aligning the error feature vectors extracted from the neutral error region with the corresponding angular domain weighting matrix, the mapping relationships in each direction and scale are implemented. Specifically, a multi-axis rotation matching method is used to map the error feature vectors onto the angular domain weighting matrix, and after ensuring data dimension consistency, weighting processing is performed. The weighted result constitutes the reconstruction reference set, which contains key correction factors in the calibration process. After generating the reconstruction reference set, it is applied to the error suppression map, and the regional error distribution of the map is iteratively adjusted through the region back-projection correction algorithm to generate the finally corrected error map.

[0077] For example, assuming the extracted error feature vector is , and the corresponding angular domain weighting matrix is . Alignment is achieved through the rotation matrix , for example: After rotating the error feature vector to a unified angular domain, it is weighted with the feature matrix: Generate a reconstructed reference set for calibration. In the error suppression map, the reconstructed reference set is subjected to regional back-projection through interpolation and optimization algorithms, such as weight interpolation: where is the projection weight. Obtain the corrected error distribution and generate a complete corrected map.

[0078] Among them, the steps of aligning and weighting the error feature vector with the angular domain weighting matrix to construct a reconstructed reference set and using the reconstructed reference set to correct the regional back-projection of the error suppression map to generate a corrected map specifically include: Perform multi-axis rotation matching on the error data of the error feature vector at different pose angles with each pose in the angular domain weighting matrix to ensure the comparability and consistency of the error data under different angular domains, and obtain the alignment result.

[0079] By using the multi-axis rotation matching algorithm to map the error feature vector to a unified coordinate system, ensure that the error data under each pose angle is consistent and comparable with each pose in the angular domain weighting matrix. Specifically, for the error feature vectors at different pose angles, use the rotation matrix to adjust them to a unified reference pose coordinate system, thereby eliminating the directional deviation of the error feature vector caused by different pose angles and generating an alignment result.

[0080] The rotation matrix is defined as follows: where is the rotation matrix for rotating the error feature vector from one pose angle to the target coordinate system; is the pose angle specified through experiments or label data; and are the trigonometric function components of the angle, representing the rotation direction. The vector obtained by projecting the error feature vector into the unified coordinate system is represented as , generating an alignment result.

[0081] For example, assume that a certain sensor collects an error feature vector data set , at three different pose angles (e.g., ). For each vector data, adjust it to a unified reference coordinate system through the rotation matrix, such as: In this way, generate an aligned feature vector set , ensuring that the error feature data can be uniformly compared and analyzed.

[0082] Weight the errors in different axes of the alignment result to obtain a weighted error result, and apply the weighted least squares method to calculate the error energy difference of each pose in different directions in the weighted error result, so as to construct a reconstruction reference set.

[0083] By performing weighted processing on the multi-dimensional error data in the alignment result to balance the errors in each direction. During the weighted processing, the dynamic weights are assigned according to the magnitude or stability of the error data. High-error data is assigned low weights, and stable data (with a lower change rate) is assigned high weights, thereby reducing the impact of high-error points on the overall error representation. Subsequently, the weighted least squares method is used to fit the relationship between the pose angle and the error energy, calculate the error energy difference of each pose in different directions, and construct a reconstruction reference set.

[0084] The weighted processing formula is as follows: where, is the weight value of the th error data point; is the error value of the th data point; is the maximum absolute error value in the dataset.

[0085] After that, apply the weighted least squares method to fit the error energy difference data, and the formula is as follows: where, is the objective function of the weighted least squares; is the data point weight; , are the input value and output value of the data, representing the pose angle and the error feature respectively; is the fitting function used to characterize the relationship between the pose angle and the error.

[0086] For example, for the weighted error data set , first calculate the weight of each data point through the weight formula. For example, for a certain point , the maximum error in the dataset is , then the weight is: Subsequently, perform weighted least squares fitting on the weighted data points to generate error energy difference data and construct a reconstruction reference set for subsequent calibration work.

[0087] Back-project the error correction factors in the reconstruction reference set and the regional errors in the error suppression map to correct the error distribution in the map to improve the overall accuracy. After back-projection correction, a corrected map is generated.

[0088] By mapping the error correction factors in the reconstruction reference set to the regional error distribution of the error suppression map, the comprehensive correction of regional errors is achieved. This back-projection correction uses an iterative interpolation algorithm to adjust the data in the map, takes the calibration data in the reconstruction reference set as the optimization target, aligns the values of each error region with the target reference data, so as to achieve the balanced distribution of errors. Finally, a corrected error map is generated, which overall improves the calibration accuracy.

[0089] The back-projection formula is as follows: Where, is the corrected error value; is the original error value; is the mapping weight, which is related to the sensitivity of regional errors; is the error correction factor in the reconstruction reference set for optimization.

[0090] For example, assume that the regional error value in the error suppression map is , and the error correction factor in the reconstruction reference set is , and the mapping weight . Through the back-projection formula for regional correction, for example: The corrected error region data , , are obtained. The data of all corrected regions are summarized to form a final corrected map for subsequent error calibration processing.

[0091] In step S400, an initial inclination residual matrix and a position correspondence index table are constructed using the corrected map, and multiple sets of attitude error difference sequences and angle delay factors are extracted based on the position correspondence index table.

[0092] By using the data in the corrected map, an initial inclination residual matrix is first constructed, which is used to record the change of residual error values corresponding to each attitude angle. Specifically, the key data points in the corrected map are structurally transformed, and the residual error values at different inclinations are divided into the rows and columns of the matrix to form a multi-dimensional initial residual matrix. The rows of the matrix represent the angles of different inclination changes, and the columns of the matrix represent the changes in time series data, forming a two-dimensional data table. In addition, based on the inclination change and experimental settings, a position correspondence index table is generated to indicate the specific position relationship of the data points in the inclination space. Subsequently, according to the initial inclination residual matrix and the index table, multiple attitude error difference sequences are extracted, and the calculation of the difference sequences follows the following formula: wherein represents the residual error difference between two inclination positions; , are the error values at adjacent positions in the matrix respectively. In addition, in combination with the position corresponding index table, the angular delay factor of any point is calculated, and the definition formula of the angular delay factor is as follows: wherein, is the angular delay factor; , are the inclinations at two adjacent positions; , are the times at two adjacent position points.

[0093] For example, assume that the data points recorded in the correction map are the inclination residual values at different times , and the corresponding positions are the inclination changes and the time series . By calculating the error difference between adjacent points: For example, for the first point: . Then, in combination with the time and angle data, the angular delay factor is calculated: The above difference sequence and the angular delay factor are recorded into the initial residual matrix and the position corresponding index table for the next step of analysis.

[0094] Gradient clustering and direction fitting are performed on multiple groups of attitude error difference sequences to obtain the inclination residual regulation mapping relationship and the direction feedback vector group.

[0095] By performing gradient clustering on the extracted attitude error difference sequences, the difference sequences with similar gradient change trends are grouped into the same category. Specifically, the K-means gradient clustering algorithm is selected, and the change rate of the error difference is used as the classification basis to divide the data points into multiple groups. Subsequently, direction fitting is performed on each group of data, and the polynomial fitting algorithm is used to mathematically model the error trend to generate the inclination residual regulation mapping relationship, which can describe the relationship between the inclination change and its error residual. On this basis, the direction feedback vector group is calculated according to the fitting result for further optimizing the inclination error data.

[0096] The gradient clustering formula is as follows: wherein, is the gradient change rate; , are two adjacent data points in the difference sequence respectively.

[0097] The direction fitting formula is as follows: Wherein, is the fitting function, representing the error trend of the difference sequence; is the inclination value; is the fitting parameter, solved by the least squares method.

[0098] For example, for the attitude error difference sequence , calculate the gradient change rate, for example: According to the gradient change rate, divide the data into two groups, for example, group 1: , group 2: . Perform direction fitting on the data of group 1, for example: ; The fitting result generates the inclination residual regulation mapping relationship, and further calculates the direction feedback vector of this group.

[0099] Use the angle delay factor and the direction feedback vector group to perform multi-dimensional partition optimization on the inclination residual regulation mapping relationship to obtain multiple groups of residual tuning coefficients.

[0100] By performing multi-dimensional partition processing on the inclination residual regulation mapping relationship, the error data is divided into each direction dimension for independent optimization. Specifically, based on the angle delay factor, weight adjustment is performed on the residual data to further refine the fitting result of the regulation mapping relationship. Combining the direction feedback vector group, apply the multi-dimensional partition optimization algorithm to optimize the inclination residual, and generate a set of residual tuning coefficients under each dimension.

[0101] The residual tuning formula is as follows: Wherein, is the residual tuning coefficient; is the residual weight; is the inclination residual regulation function; is the angle delay factor.

[0102] For example, for the direction feedback vector group and the angle delay factor data, assume that a certain inclination mapping relationship is: Combined with the weight value of the direction feedback vector group and the delay factor data Calculate the tuning coefficient: Record the tuning result as a set of residual tuning coefficients for further calibration.

[0103] Generate a set of residual calibration functions based on multiple groups of residual tuning coefficients and the initial tilt residual matrix, and use the set of residual calibration functions to calibrate the output data of the high-precision tilt measurement sensor.

[0104] By combining the residual calibration coefficients obtained after tuning with the initial tilt residual matrix, use interpolation and fitting algorithms to generate a set of residual calibration functions. The calibration functions can describe the error correction rules of the sensor in different tilt regions and are used to calibrate the output data of the sensor to high-precision values. Finally, apply the set of calibration functions to all the output data of the sensor for comprehensive error calibration.

[0105] The formula for the residual calibration function is as follows: Where: is the final calibration function; is the residual tuning coefficient; is the set of fitting functions.

[0106] For example, assume the residual tuning coefficient is , and the initial tilt residual matrix is . For a certain tilt value , the final calibration function is: Apply the calibration function to the output data of the sensor, and the calibrated data can meet the high-precision requirements.

[0107] In this embodiment, through a series of precise error analyses and optimization algorithms, the calibration of a high-precision inclination measurement sensor is successfully achieved. First, a multi-channel data acquisition system is used to collect the original output data of the high-precision inclination measurement sensor at multiple known attitude angles, and an accurate inclination response sequence is obtained through attitude calibration and synchronous sampling. Then, for the error distribution sequence, the least squares method is used to fit and balance the trend, generating a set of error center-of-gravity vectors. Combining with the drift correction factor, a multi-dimensional attitude mapping table is constructed. On this basis, using the gradient clustering and direction fitting methods, the inclination residual regulation mapping relationship is obtained, and multi-dimensional partition optimization is performed through the angle delay factor and the direction feedback vector group, and finally the residual optimization coefficient is generated. By combining the residual optimization coefficient with the initial inclination residual matrix, a set of residual calibration functions is constructed, and then the output data of the sensor is accurately calibrated. The whole process not only effectively eliminates the influence of various error sources, but also realizes high-precision calibration, improving the stability and accuracy of inclination measurement. Finally, through back-projection correction and weighted least squares method to optimize the calibration function, the high-precision performance of the sensor in a complex dynamic environment is ensured. In this embodiment, through fine error mapping, partition optimization and calibration function generation, the robustness and high precision of the calibration process are ensured.

[0108] Embodiment 3: This application also provides a calibration system 10 for a high-precision inclination measurement sensor, including an acquisition module 11, an analysis module 12, a correction module 13 and a calibration module 14.

[0109] The acquisition module 11 is mainly used to acquire the original output data of the high-precision inclination measurement sensor at multiple known attitude angles, generate an initial error map according to the original output data, and extract the error distribution sequence based on the initial error map.

[0110] The analysis module 12 is mainly used to perform piecewise fitting and trend normalization on the error distribution sequence to obtain a direction weight table, construct a multi-dimensional attitude mapping table based on the direction weight table, perform error center-of-gravity vector analysis on the multi-dimensional attitude mapping table, and generate an error mapping map.

[0111] The correction module 13 is mainly used to suppress the boundary deviation of the error mapping map according to the preset constraint rules to generate an error suppression map, perform multi-axis matching and error energy difference compensation on the error suppression map to obtain a reconstructed reference set, and use the reconstructed reference set to perform back-projection correction on the error suppression map to obtain a corrected map.

[0112] The calibration module 14 is mainly used to construct an inclination residual regulation mapping relationship through the corrected map, perform multi-dimensional partition optimization on the inclination residual regulation mapping relationship to construct a set of residual calibration functions, and calibrate the output data of the high-precision inclination measurement sensor based on the set of residual calibration functions.

[0113] In this embodiment, through the collaborative work of the acquisition module 11, the analysis module 12, the correction module 13, and the calibration module 14, a calibration system 10 for a high-precision inclination measurement sensor is constructed, realizing the whole process from data acquisition to calibration. Specifically, the acquisition module 11 collects the original output data of the high-precision inclination measurement sensor at multiple known attitude angles and generates an initial error map and an error distribution sequence, providing a data basis for the subsequent calibration process; the analysis module 12 performs piecewise fitting and trend normalization on the error distribution sequence, accurately identifies the error characteristics at different attitudes by generating a direction weight table and a multi-dimensional attitude mapping table, and generates an error mapping map, providing a global data description for the error correction operation; the correction module 13 further optimizes the error mapping map through boundary deviation suppression, multi-axis matching, and error energy difference compensation, generates an error suppression map, and generates a correction map through back-projection correction, comprehensively reducing the coupling influence of multi-dimensional errors; the calibration module 14 constructs an inclination residual regulation mapping relationship based on the correction map and generates a set of residual calibration functions in combination with the multi-dimensional partition optimization technology, realizing the accurate calibration of the sensor output data. Through modular design, each module of the entire calibration system has clear responsibilities, can efficiently complete the full process operations of data acquisition, error analysis, correction, and calibration, and greatly improves the accuracy and robustness of the inclination measurement sensor, providing a reliable guarantee for high-precision inclination measurement in complex dynamic environments.

[0114] It should be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system and each module can refer to the corresponding processes in the foregoing Embodiment 1 and will not be elaborated herein.

[0115] The structures, proportions, sizes, etc. shown in the drawings of this specification are only used to cooperate with the content disclosed in the specification for those skilled in this technology to understand and read, and are not used to limit the limited conditions under which the present invention can be implemented. Therefore, they do not have technical essential significance. Any modification of the structure, change of the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope that can be covered by the technical content disclosed by the present invention.

[0116] The above is the case. The above embodiments are only used to illustrate the technical solutions of the present invention, not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A calibration method for a high-precision inclination measurement sensor, characterized in that, Including: Obtain the original output data of a high-precision inclination measurement sensor at multiple known attitude angles, generate an initial error map based on the original output data, and extract an error distribution sequence based on the initial error map; Perform piecewise fitting and trend normalization on the error distribution sequence to obtain a direction weight table, construct a multi-dimensional attitude mapping table based on the direction weight table, and perform error centroid vector analysis on the multi-dimensional attitude mapping table to generate an error mapping map; Suppress boundary deviations of the error mapping map according to preset constraint rules to generate an error suppression map, perform multi-axis matching and error energy difference compensation on the error suppression map to obtain a reconstruction reference set, and use the reconstruction reference set to perform back-projection correction on the error suppression map to obtain a corrected map; Construct an inclination residual regulation mapping relationship through the corrected map, perform multi-dimensional partition optimization on the inclination residual regulation mapping relationship to construct a residual calibration function set, and calibrate the output data of the high-precision inclination measurement sensor based on the residual calibration function set.

2. The calibration method of the high-precision inclination measurement sensor according to claim 1, characterized in that The steps of obtaining the original output data of a high-precision inclination measurement sensor at multiple known attitude angles, generating an initial error map based on the original output data, and extracting an error distribution sequence based on the initial error map include: Use a preset multi-channel data acquisition system to collect the original output data of a high-precision inclination measurement sensor at multiple known attitude angles, perform attitude calibration and synchronous sampling on the original output data to obtain an attitude label set and a corresponding inclination response sequence; Perform statistical debiasing and amplitude normalization on the inclination response sequence to obtain a data equilibrium set and an initial difference set, construct an error field mapping array using the attitude label set and the initial difference set, and extract an error perturbation intensity spectrum and a direction drift trajectory sequence based on the error field mapping array; Construct an initial error map using the error perturbation intensity spectrum and the data equilibrium set, and combine the initial error map with the direction drift trajectory sequence to extract an error distribution sequence.

3. The calibration method of the high-precision inclination measurement sensor according to claim 1, wherein, The steps of performing piecewise fitting and trend normalization on the error distribution sequence to obtain a direction weight table, constructing a multi-dimensional attitude mapping table based on the direction weight table, and performing error centroid vector analysis on the multi-dimensional attitude mapping table to generate an error mapping map include: Perform interval division and least squares polynomial fitting on the error distribution sequence to obtain a set of fitting functions, perform local normalization and trend balance processing on the error distribution sequence based on the set of fitting functions to obtain a set of normalized residuals; Adopt a sliding window difference analysis and mutation detection algorithm to extract the peak and valley points of the residual fluctuations of the set of normalized residuals, construct a set of feature points according to the peak and valley points of the residual fluctuations, construct a set of drift correction factors and a direction weight table according to the set of feature points, and construct a multi-dimensional attitude mapping table using the direction weight table; Compare the residual change amplitudes between adjacent feature points in the set of feature points using a preset threshold change rate rule to construct a section change sequence; Cluster and reconstruct the multi-dimensional pose mapping table using the section change sequence, extract the error centroid vector sets of each pose block, and combine the error centroid vector sets with the drift correction factor set to generate an error mapping atlas.

4. The calibration method of the high-precision inclination measurement sensor according to claim 3, characterized in that The step of clustering and reconstructing the multi-dimensional pose mapping table using the section change sequence, extracting the error centroid vector sets of each pose block, and combining the error centroid vector sets with the drift correction factor set to generate an error mapping atlas includes: Cluster the error data in the multi-dimensional pose mapping table using the section change sequence to obtain multiple pose blocks; Calculate the error centroid vectors based on the error data within each pose block to obtain an error centroid vector set, combine the error centroid vector set with the drift correction factor set to generate a number of error mapping data, and summarize all the error mapping data to obtain an error mapping atlas.

5. The calibration method of the high-precision inclination measurement sensor according to claim 1, characterized in that The step of suppressing boundary deviation of the error mapping atlas according to a preset constraint rule to generate an error suppression atlas, performing multi-axis matching and error energy difference compensation on the error suppression atlas to obtain a reconstruction reference set, and using the reconstruction reference set to perform back-projection correction on the error suppression atlas to obtain a corrected atlas includes: Suppress the boundary deviation of the error mapping atlas according to a preset constraint rule, extract the residual distribution in the edge transition region, and construct a boundary optimization set based on the residual distribution in the edge transition region; Perform error density fitting and principal component analysis in the local angle domain on the boundary optimization set to generate an angle domain weighted matrix; Generate an error suppression atlas using the boundary optimization set, extract a stable region based on the error suppression atlas, construct a neutral error region according to the stable region, and extract error feature vectors from the neutral error region; Align and weight the error feature vectors with the angle domain weighted matrix to construct a reconstruction reference set, and use the reconstruction reference set to perform regional back-projection correction on the error suppression atlas to generate a corrected atlas.

6. The calibration method of the high-precision inclination measurement sensor according to claim 5, characterized in that, The step of aligning and weighting the error feature vectors with the angle domain weighted matrix to construct a reconstruction reference set, and using the reconstruction reference set to perform regional back-projection correction on the error suppression atlas to generate a corrected atlas includes: Perform multi-axis rotation matching on the error data of the error feature vectors at different pose angles with each pose in the angle domain weighted matrix to obtain an alignment result; Perform weighting processing on the errors in different axes of the alignment result to obtain a weighted error result, and apply the weighted least squares method to calculate the error energy difference of each pose in different directions in the weighted error result to construct a reconstruction reference set; Perform back-projection on the error correction factors in the reconstruction reference set and the regional errors in the error suppression atlas to generate a corrected atlas.

7. The calibration method of the high-precision inclination measurement sensor according to claim 1, characterized in that The step of constructing an inclination residual regulation mapping relationship through the correction map, performing multi-dimensional partition optimization on the inclination residual regulation mapping relationship to construct a set of residual calibration functions, and calibrating the output data of the high-precision inclination measurement sensor based on the set of residual calibration functions includes: Constructing an initial inclination residual matrix and a position correspondence index table by using the correction map, and extracting multiple groups of attitude error difference sequences and angle delay factors based on the position correspondence index table; Performing gradient clustering and direction fitting on the multiple groups of attitude error difference sequences to obtain an inclination residual regulation mapping relationship and a direction feedback vector group; Performing multi-dimensional partition optimization on the inclination residual regulation mapping relationship by using the angle delay factor and the direction feedback vector group to obtain multiple groups of residual optimization coefficients; Generating a set of residual calibration functions according to the multiple groups of residual optimization coefficients and the initial inclination residual matrix, and calibrating the output data of the high-precision inclination measurement sensor by using the set of residual calibration functions.

8. A calibration system for a high-precision inclination measurement sensor, characterized in that, Including: An acquisition module, configured to acquire the original output data of the high-precision inclination measurement sensor at multiple known attitude angles, generate an initial error map according to the original output data, and extract an error distribution sequence based on the initial error map; An analysis module, configured to perform piecewise fitting and trend normalization on the error distribution sequence to obtain a direction weight table, construct a multi-dimensional attitude mapping table based on the direction weight table, perform error center of gravity vector analysis on the multi-dimensional attitude mapping table, and generate an error mapping map; A correction module, configured to suppress boundary deviation of the error mapping map according to a preset constraint rule to generate an error suppression map, perform multi-axis matching and error energy difference compensation on the error suppression map to obtain a reconstruction reference set, and perform back-projection correction on the error suppression map by using the reconstruction reference set to obtain a correction map; A calibration module, configured to construct an inclination residual regulation mapping relationship through the correction map, perform multi-dimensional partition optimization on the inclination residual regulation mapping relationship to construct a set of residual calibration functions, and calibrate the output data of the high-precision inclination measurement sensor based on the set of residual calibration functions.

Images