Engineering measurement error dynamic correction method based on multi-sensor data fusion

Abstract

The invention discloses an engineering measurement error dynamic correction method based on multi-sensor data fusion, and relates to the technical field of engineering measurement. Three error sources of sensor physical characteristics, environmental interference and manual operation are defined as estimable state variables, and differentiation processing is performed according to different characteristics of errors: a dynamic model is established to embed the sensor and environmental errors into a state space, so that real-time compensation based on the model is realized; according to the scheme, a gross error recognition mechanism based on data-driven statistical characteristics is established, real-time diagnosis and processing are carried out on personal errors, joint optimal estimation is carried out on unified state vectors by adopting a recursive estimation algorithm, and high-precision corrected physical quantities and estimated values of all error states are synchronously output. The measurement precision and reliability are obviously improved; by means of robust processing of gross errors, stability of the system in a non-ideal environment and online self-diagnosis and health monitoring of the measurement system are enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of engineering surveying, and in particular to an engineering surveying error dynamic correction method based on multi-sensor data fusion. BACKGROUND

[0002] Engineering surveying is the basis of modern engineering construction, safety monitoring and quality control, and its precision and reliability are directly related to the success or failure of the project. With the development of modern engineering structures towards large-scale and complex, and the increasing demand for measurement accuracy and automation, traditional single-sensor measurement methods have been difficult to meet the needs.

[0003] Currently, in the field of engineering surveying, especially in dynamic and continuous measurement scenarios (such as bridge health monitoring, dam deformation monitoring, construction machinery precise positioning, etc.), the following technical difficulties exist: 1. Complex and interwoven error sources, limitations of traditional correction methods Engineering surveying errors mainly come from three aspects: sensor itself, measurement environment and human operation.

[0004] Sensor error: including sensor zero deviation, scale factor error, temperature drift and time drift, etc. inherent system error. Currently, laboratory calibration method is usually used for compensation, but the calibration parameters are obtained in a specific and stable environment. When the sensor is put into the actual engineering site, its performance will drift due to the influence of complex working conditions and time-varying environment, resulting in the invalidation of pre-calibration parameters and the problem of "measurement error".

[0005] Environmental error: fluctuations in temperature, humidity, air pressure and other environmental factors will significantly affect the measurement results. For example, the light wave ranging of total station is affected by atmospheric refraction, and GNSS positioning is affected by ionosphere and troposphere delay. Existing technologies mostly use empirical models for compensation, but these models are usually static or semi-static, and cannot respond to dynamic changes in the environment in real time and accurately, resulting in significant increase in measurement error when there is a large diurnal temperature difference or sudden weather changes.

[0006] Human error: in the measurement process, due to operation errors, target misillumination or transient strong interference (such as vibration), gross errors will be generated. This kind of error has burstiness and discreteness, and traditional data processing methods are difficult to identify and eliminate in real time, and a gross error can contaminate the entire data sequence, even leading to false alarms of the monitoring system.

[0007] 2. Multi-sensor application is simple, and deep data fusion is not achieved In order to improve reliability, multi-sensor (such as GNSS and IMU, total station and tilt sensor) combination measurement is adopted in engineering. However, many current applications still remain at the level of simple data complementation or weighted average, failing to fully exploit the advantages of multi-sensor data fusion. This simple combination cannot fundamentally separate various error sources. For example, it cannot distinguish whether the change in the observation value is due to the real displacement of the measured object, the drift of the sensor, or environmental interference. When an abnormal data is generated by a sensor due to environmental or human factors, the simple data fusion method does not have the "immunity" ability, and the abnormal data will directly pollute the fusion result, reducing the robustness of the system.

[0008] 3. The modified model is static and cannot adapt to dynamic changing scenarios Most of the existing error compensation models are static or offline models, which cannot adaptively adjust according to the changes of sensor characteristics and environmental conditions during measurement. This leads to the gradual degradation of the accuracy of the measurement system over time due to sensor aging and environmental changes, and the system cannot achieve continuous "dynamic" high-precision measurement.

[0009] Therefore, there is an urgent need in the art for an advanced measurement method that can uniformly handle sensor, environmental and human errors, and can achieve dynamic, online and adaptive correction, to overcome the inherent defects of the prior art and meet the stringent requirements of modern precision engineering measurement for high precision, high reliability and high robustness. Therefore, an engineering measurement error dynamic correction method based on multi-sensor data fusion is proposed. SUMMARY

[0010] The main purpose of the present application is to provide an engineering measurement error dynamic correction method based on multi-sensor data fusion, which uniformly defines the three major error sources of sensor physical characteristics, environmental interference and human operation as estimable state variables, and then implements differentiated processing according to the different characteristics of the errors: by establishing a dynamic model to embed the sensor and environmental errors into the state space, real-time compensation based on the model is achieved; at the same time, a gross error identification mechanism based on data-driven statistical characteristics is established to diagnose and immunize human errors in real time. Finally, a recursive estimation algorithm is used to jointly optimize the estimation of the unified state vector, and the estimated values of the high-precision corrected physical quantities and each error state are output simultaneously, which can effectively solve the problems in the background art.

[0011] To achieve the above purpose, the technical scheme adopted by the present application is, The engineering measurement error dynamic correction method based on multi-sensor data fusion comprises the following steps: S1, define the measurement errors caused by sensor physical characteristics, environmental disturbances and human operation factors in engineering measurement as state variables that can be jointly estimated, and construct a unified state vector containing a physical state vector, a sensor error state vector and an environmental error state vector; S2, obtain observation data from heterogeneous sensors, establish a dynamic evolution model for describing sensor physical characteristics and environmental disturbances, and embed the dynamic evolution model into state equations and observation equations of the unified state vector; S3, establish a gross error identification mechanism for describing human operation factors based on data-driven statistical characteristics; S4, based on the unified state vector constructed in step S1 and the state equations and observation equations in step S2, use a recursive state estimation algorithm to perform real-time joint optimal estimation on the unified state vector; S5, in the recursive state estimation process, based on the established gross error identification mechanism, perform real-time diagnosis and Robust processing on the observation data; S6, extract the physical state vector from the unified state vector obtained from each step of real-time estimation as the final measurement result after dynamic error correction, and synchronously output the estimated values of the sensor error state vector and the environmental error state vector.

[0012] Further, the unified state vector is represented as: = ; Wherein, is a unified state vector; is a discrete time index, representing time t; is a physical state vector, representing the true physical quantity of the measured object; is a sensor error state vector, composed of to-be-estimated parameters of an error model describing sensor physical characteristics; is an environmental error state vector, composed of to-be-estimated parameters or states of an error model describing environmental disturbances.

[0013] Further, the sensor error state vector is constructed based on a first error model for describing the measurement error of the true physical quantity of the measured object caused by sensor physical characteristics, and is specifically represented as: = + + + ; Wherein, is the measurement error of the true physical quantity of the measured object caused by sensor physical characteristics; This refers to the measurement error of the true physical quantity of the measured object caused by the zero-position deviation of the sensor and the scale. = , Zero-position deviation factor As a scale factor, for The true value of a moment; This refers to the measurement error of the true physical quantity of the measured object caused by sensor drift. = , The measurement error of the true physical quantity of the measured object caused by linear drift over time. This refers to the measurement error of the true physical quantity of the measured object caused by the cumulative drift due to workload. The measurement error of the true physical quantity of the measured object is caused by the hysteresis and nonlinearity of the sensor. = , It is a nonlinear function. for The derivative; The noise is Gaussian white noise; correspondingly, the sensor error state vector is defined as... = .

[0014] Furthermore, the environmental error state vector is constructed based on a second error model, which describes the relationship between environmental error and environmental parameters. The second error model is in the form of a linear regression model or a nonlinear function, and the environmental parameters include at least temperature, air pressure, humidity, and vibration.

[0015] Furthermore, when the environmental parameters are linearly correlated with the environmental error, the second error model is specifically expressed as follows: = + + + + ; in, This is a correction factor; This is the temperature sensitivity coefficient; Real-time temperature; For reference temperature; This is the barometric pressure sensitivity coefficient; Real-time air pressure; Reference air pressure; Humidity sensitivity coefficient; Real-time humidity; For reference humidity; This is the vibration sensitivity coefficient; a real-time vibration parameter; a reference vibration parameter.

[0016] Further, when the environmental parameter is nonlinearly related to the environmental error, the second error model is specifically represented as: = + ; wherein, is a nonlinear function; is a constant term.

[0017] Further, the observation equation is specifically represented as: = + + + ; wherein, is a sensor observation vector; is an observation matrix of a physical state vector; is an observation matrix of a sensor error state vector; is an observation matrix of an environmental error state vector; is observation noise.

[0018] Further, the identification mechanism based on the data-driven statistical characteristics is constructed based on a third error model, specifically: in the updating step of the recursive state estimation algorithm, the observation innovation and its theoretical standard deviation are calculated; if the absolute value of the normalized observation innovation = is greater than a set threshold value, it is determined that there is a human gross error; wherein the third error model is used to describe the statistical abnormal characteristics of the gross error caused by human operation in the observation data sequence and the impact mode on the data fusion process, and the third error model is specifically represented as: = ; wherein, is the size of the gross error occurring at time ; is a Dirac function, indicating that the error only occurs at a specific time; is the number of gross errors occurring within the observation period.

[0019] Further, the recursive state estimation algorithm is a Kalman filter, an extended Kalman filter or an unscented Kalman filter. ​The specific way of the robust processing is: when it is determined that a certain observation value has gross error, it is removed from the current state update, or its weight is reduced by increasing the variance value in the observation noise covariance matrix corresponding to it.

[0020] A computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method for dynamic correction of engineering measurement errors based on multi-sensor data fusion.

[0021] The present application has the following beneficial effects, Compared with the prior art, the present application can dynamically track and compensate the time-varying parts of the sensor errors and environmental errors (such as sensor temperature drift and deformation caused by changes in environmental temperature) by real-time joint estimation of the state vectors of the sensor errors and environmental errors, instead of using fixed calibration values, thereby overcoming the problem of failure of traditional static calibration in time-varying environments, and continuously outputting high-precision measurement results in complex working conditions.

[0022] Compared with the prior art, the present application can not only output the corrected physical quantities, but also output the estimated values of the state vectors of the sensor errors and environmental errors simultaneously, so as to monitor the health indicators of the sensors such as zero offset and scale factor, and the intensity of environmental interference in real time, thereby realizing predictive maintenance and credibility evaluation of measurement results.

[0023] Compared with the prior art, the present application adopts a data-driven gross error identification and robust processing mechanism, so that the system can diagnose and suppress abnormal data caused by operation errors or transient disturbances in real time, and ensure that the system output is still stable and reliable even in non-ideal measurement environments, and will not be completely disabled or divergent due to individual gross errors.

[0024] Compared with the prior art, the present application adopts a recursive state estimation algorithm, which provides the optimal estimation under given all available information in a statistical sense. Through deep fusion of heterogeneous sensor data, the accuracy and reliability of the system output are better than the best capability of any single sensor working independently. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 The figure is a flowchart of the method for dynamic correction of engineering measurement errors based on multi-sensor data fusion of the present application. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0027] Example 1: Referring to Figure 1 The flowchart of the application based on the multi-sensor data fusion engineering measurement error dynamic correction method is shown, and the specific implementation process of the correction method is as follows: Step 1, error state and system initialization 1.1) Analyze the engineering measurement scene, and determine that the main error sources are sensor physical characteristics (such as zero offset, scale factor), environmental interference (such as temperature, air pressure), and human operation factors (such as gross error); 1.2) Define the physical state vector (such as position, velocity), the sensor error state vector (such as zero offset factor , scale factor ), and the environmental error state vector (such as temperature-induced error), wherein the sensor error state vector is constructed based on a first error model, and the first error model is used to describe the measurement error of the real physical quantity of the measured object caused by the sensor physical characteristics, and is specifically represented as: = + + + ; Wherein, is the measurement error of the real physical quantity of the measured object caused by the sensor physical characteristics; is the measurement error of the real physical quantity of the measured object caused by the zero offset and scale of the sensor, , is the zero offset factor, is the scale factor, is the real value at time t; is the measurement error of the real physical quantity of the measured object caused by the drift of the sensor, , is the measurement error of the real physical quantity of the measured object caused by the linear drift over time, is the measurement error of the real physical quantity of the measured object caused by the cumulative drift over work, is the measurement error of the real physical quantity of the measured object caused by the hysteresis and nonlinearity of the sensor, , is a nonlinear function, is the derivative of ; is a Gaussian white noise; correspondingly, the sensor error state vector is defined as =​​​ .

[0028] The environmental error state vector is constructed based on a second error model, which describes the relationship between the environmental error and environmental parameters, in the form of a linear regression model or a nonlinear function, and the environmental parameters at least include temperature, air pressure, humidity, and vibration.

[0029] When the environmental parameters are linearly related to the environmental error, the second error model is specifically expressed as: = + + + + ; wherein, is a correction coefficient; is a temperature sensitivity coefficient; is a real-time temperature; is a reference temperature; is an air pressure sensitivity coefficient; is a real-time air pressure; is a reference air pressure; is a humidity sensitivity coefficient; is a real-time humidity; is a reference humidity; is a vibration sensitivity coefficient; is a real-time vibration parameter; is a reference vibration parameter.

[0030] When the environmental parameters are nonlinearly related to the environmental error, the second error model is specifically expressed as: = + ; wherein, is a nonlinear function; is a constant term; 1.3) Construct a unified state vector, expressed as: = ; wherein, is a unified state vector; is a discrete time index, representing the time; is a physical state vector, representing the real physical quantity of the measured object; is a sensor error state vector, composed of the to-be-estimated parameters of the error model describing the physical characteristics of the sensor; is an environmental error state vector, composed of the to-be-estimated parameters or states of the error model describing the environmental interference; 1.4) Initialize the state estimation value and its error covariance matrix .

[0031] Step 2, dynamic model construction and multi-source data synchronization 2.1) Configure and synchronize heterogeneous sensors (such as GNSS receivers, IMU) and environmental sensors (such as temperature, barometer).

[0032] 2.2) Establish the state equation, expressed as: = + + ; where, is the state transition matrix; is the environmental parameter input (such as temperature value); is the process noise; 2.3) Establish the observation equation, specifically expressed as: = + + + ; where, is the sensor observation vector; is the observation matrix of the physical state vector; is the observation matrix of the sensor error state vector; is the observation matrix of the environmental error state vector; is the observation noise.

[0033] Step 3, construction of artificial gross error identification mechanism 3.1) Based on the third error model, construct an identification mechanism based on data-driven statistical characteristics, the third error model is used to describe the statistical abnormal characteristics of gross errors caused by human operation in the observation data sequence and its impact mode on the data fusion process, the third error model is specifically expressed as: = ; where, is the size of the gross error occurring at time ; is the Dirac function, indicating that the error only occurs at a specific time; is the number of gross errors occurring during the observation period; 3.2) Based on the data-driven idea, design a gross error detection statistic with observation innovation as the core; 3.3) In the update step of the recursive state estimation algorithm, calculate the observation innovation and its theoretical standard deviation ; 3.4)If the absolute value of the normalized observation innovation = is greater than a set threshold, then it is determined that there is a gross error.

[0034] Step 4, integrated recursive fusion estimation Using recursive state estimation algorithm (such as extended Kalman filter EKF), time update (prediction) is performed, specifically: = + + ; = + ; wherein, is the process noise covariance matrix.

[0035] Step 5, real-time diagnosis and robust update 5.1)Obtain the sensor observation value at the current time ; 5.2)Calculate the observation innovation : = - ; 5.3)Perform gross error diagnosis, based on the judgment principle of step 3.4, check each component of the observation innovation ; 5.4)Robust processing, specifically: If no gross error is found, use the standard observation noise covariance matrix .

[0036] If the ith observation is found to be a gross error, take the "weight reduction" strategy, i.e. multiply the diagonal elements in the matrix by a very large number (such as 10 6 ), or take the "removal" strategy, remove the observation value and the corresponding row in the observation matrix; 5.5)Perform measurement update (correction), specifically: = ; = + ; = .

[0037] Step 6, dynamic correction result output and system diagnosis 6.1) Output the final correction result: extract the physical state vector from the updated state vector as the high-precision final measurement result after dynamic error correction; 6.2) Synchronize the output error estimate: output the estimate of the sensor error state vector and the environmental error state vector ; 6.3) System diagnosis: based on the error estimate output in S6.2, monitor the sensor health status (e.g. whether the zero offset is too large) and the intensity of environmental interference in real time, and provide data support for system maintenance.

[0038] Example 2: Application scenario In large earthwork projects, it is necessary to accurately measure the real-time moving speed of transport vehicles in the construction site for optimization scheduling, calculation of cubic capacity and safety management. Various sensors are installed on the vehicles.

[0039] Specific implementation case: dynamic correction of vehicle moving speed Step 1, error state and system initialization 1.1) Error source analysis: Sensor physical characteristics: GNSS speed solution has noise and scale error; IMU (Inertial Measurement Unit) accelerometer has zero offset.

[0040] Environmental interference: GNSS signal tropospheric delay is related to temperature and air pressure.

[0041] Human operation factors: GNSS signal may produce large error (jump) due to temporary obstruction (such as passing through a bridge hole).

[0042] 1.2) Define state vector: Physical state vector : [v, a] T (speed, acceleration) Sensor error state vector : [b a , k gnss ] T (IMU accelerometer zero offset, GNSS speed scale factor) Environmental error state vector : [E tropo ] speed error caused by tropospheric delay) 1.3) Unified state vector: X = [v, a, b a , k gnss , E tropo ] T 1.4) Initialization: Assume the vehicle is stationary, initialize = [0, 0, 0, 0, 0] T , and give a large initial covariance .

[0043] Step 2, Dynamic model construction and multi-source data synchronization 2.1) Sensor configuration: Synchronize the collection of GNSS receiver speed observations z gnss , IMU acceleration observations z acc , and temperature / barometer data Temp / Press.

[0044] 2.2) Establish the state equation: = + × + ; Assume a first-order decay; is the environmental input; The physical state (speed, acceleration) uses a uniform motion model.

[0045] Sensor errors (zero bias, scale factor) are modeled as random walks.

[0046] Environmental errors (tropospheric delay) are modeled as a first-order process related to temperature.

[0047] 2.3) Establish the observation equation: = × + ; GNSS observations z gnss are affected by the true speed v, scale error k gnss , and environmental error E tropo .

[0048] IMU acceleration observations z acc are affected by the true acceleration a and accelerometer zero bias b a .

[0049] Step 3, Construction of artificial gross error identification mechanism Discrimination criteria: mainly for GNSS speed observations. Calculate its innovation v gnss (k) = z gnss (k) - .

[0050] If |v gnss (k) | > 5 x σ ν,gnss (e.g. velocity jump over 3 m / s), it is determined as a gross error.

[0051] Step 4, integrated recursive fusion estimation and robust update

[0052] 4.1) Prediction: using the optimal estimation of last time according to the state equation of S2.2 predict the current state .

[0053] 4.2) Collect observations: get the raw data of GNSS and IMU.

[0054] 4.3) Gross error diagnosis and processing: Case 1: vehicle is driving normally, GNSS innovation is in normal range. System uses standard update.

[0055] Case 2: vehicle suddenly enters under a bridge, GNSS signal is lost, velocity value changes dramatically, triggering gross error alarm. System immediately increases the variance R of GNSS observation noise gnss . During this period, the system mainly relies on the integration of IMU acceleration to calculate the velocity, and takes this opportunity to better estimate the IMU bias b a .

[0056] Case 3: vehicle drives out of the bridge, GNSS signal is restored, innovation is normal. System automatically restores normal use of GNSS data.

[0057] 4.4) Update: using the observation (possibly after robust processing) and Kalman gain, update the state vector to get the optimal estimation .

[0058] Step 5, output of dynamic correction results and system diagnosis 5.1) output the final correction result: the final output of high-precision velocity is . The velocity has been corrected as follows: 5.11) deducted the influence of GNSS scale error and environmental error .

[0059] 5.12) seamlessly connected by IMU when GNSS signal is poor, and corrected the IMU zero bias during the effective period of GNSS, suppressing the integral drift.

[0060] 5.2) synchronous output and diagnosis: output : If its value keeps increasing, the early warning IMU is likely to fail soon.

[0061] Output : Its trend of change reflects the intensity of the current atmospheric environment interference on GNSS measurement.

[0062] Through the above cases, it can be proved that the method proposed by the application can fuse multiple imperfect sensor data together, dynamically correct various errors, finally output a continuous and reliable, higher precision speed measurement value, and at the same time, real-time monitor the health status of the whole measurement system.

[0063] The above shows and describes the basic principles and main features of the application and the advantages of the application. Those skilled in the art should understand that the application is not limited to the above examples, and the above examples and descriptions in the specification are only to illustrate the principles of the application. Without departing from the spirit and scope of the application, various changes and improvements can be made to the application, and these changes and improvements all fall within the scope of the claimed application. The scope of protection of the application is defined by the appended claims and their equivalents.

Claims

1. A method for dynamic correction of engineering surveying errors based on multi-sensor data fusion, characterized in that, The method comprises the following steps: S1, defining the measurement errors caused by the physical characteristics of sensors, environmental interference and human operation factors in engineering measurement as state variables that can be jointly estimated, and constructing a unified state vector comprising a physical state vector, a sensor error state vector and an environmental error state vector; S2, obtaining observation data from heterogeneous sensors, establishing a dynamic evolution model for describing the physical characteristics of sensors and environmental interference, and embedding the dynamic evolution model into the state equation and the observation equation of the unified state vector; S3, establishing a gross error identification mechanism for describing human operation factors based on data-driven statistical characteristics; S4, based on the unified state vector constructed in step S1 and the state equation and the observation equation in step S2, using a recursive state estimation algorithm to perform real-time joint optimal estimation on the unified state vector; S5, in the recursive state estimation process, based on the established gross error identification mechanism, performing real-time diagnosis and Robust processing on the observation data; S6, extracting the physical state vector from the unified state vector obtained in each step of real-time estimation as the final measurement result after dynamic error correction, and synchronously outputting the estimated values of the sensor error state vector and the environmental error state vector.

2. The method for dynamic correction of engineering surveying errors based on multi-sensor data fusion according to claim 1, characterized in that, The unified state vector is expressed as: = ； wherein, is a unified state vector; is a discrete time index, representing time instant; is a physical state vector, characterizing real physical quantities of the measured object; is a sensor error state vector, consisting of estimated parameters of error models describing physical characteristics of the sensors; is an environmental error state vector, consisting of estimated parameters or states of error models describing environmental disturbances.

3. The method for dynamic correction of engineering surveying errors based on multi-sensor data fusion according to claim 2, characterized in that, The sensor error state vector is constructed based on a first error model, and the first error model is used to describe the measurement error of the real physical quantity of the measured object caused by the physical characteristics of the sensor, and is specifically expressed as: = + + + ； wherein, is the measurement error of the real physical quantity of the measured object caused by the physical characteristics of the sensor; is the measurement error of the real physical quantity of the measured object caused by the zero offset and scale of the sensor, = 0, , is the zero offset factor, is the scale factor, is the real value at the time instant; is the measurement error of the real physical quantity of the measured object caused by the drift of the sensor, = 0, , is the measurement error of the real physical quantity of the measured object caused by the linear drift over time, is the measurement error of the real physical quantity of the measured object caused by the cumulative drift over work amount, is the measurement error of the real physical quantity of the measured object caused by the hysteresis and nonlinearity of the sensor, = 0, , is the non-linear function, is the derivative of ; is the Gaussian white noise; correspondingly, the sensor error state vector is defined as = 0, .​ 4. The method for dynamic correction of engineering surveying errors based on multi-sensor data fusion according to claim 2, characterized in that, The environmental error state vector is constructed based on a second error model, and the second error model describes the relationship between environmental error and environmental parameters, which is a linear regression model or a nonlinear function, and the environmental parameters at least include temperature, air pressure, humidity and vibration.

5. The method for dynamic correction of engineering surveying errors based on multi-sensor data fusion according to claim 4, characterized in that, When the environmental parameters and the environmental error are linearly related, the second error model is specifically expressed as: = + + + + ； wherein, is a correction factor; is a temperature sensitivity factor; is a real-time temperature; is a reference temperature; is a barometric pressure sensitivity factor; is a real-time barometric pressure; is a reference barometric pressure; is a humidity sensitivity factor; is a real-time humidity; is a reference humidity; is a vibration sensitivity factor; is a real-time vibration parameter; is a reference vibration parameter.

6. The method for dynamic correction of engineering surveying errors based on multi-sensor data fusion according to claim 4, characterized in that, When the environmental parameters and the environmental error are nonlinearly related, the second error model is specifically expressed as: = + ； wherein is a non-linear function; is a constant term.

7. The method for dynamic correction of engineering surveying errors based on multi-sensor data fusion according to claim 1, characterized in that, The observation equation is specifically expressed as: = + + + ； wherein, is a sensor observation vector; is an observation matrix for the physical state vector; is an observation matrix for the sensor error state vector; is an observation matrix for the environmental error state vector; is an observation noise.

8. The method for dynamic correction of engineering surveying errors based on multi-sensor data fusion according to claim 1, characterized in that, The identification mechanism based on the data-driven statistical characteristics is constructed based on a third error model, specifically, in the updating step of the recursive state estimation algorithm, the observation innovation is calculated and its theoretical standard deviation ; if the absolute value of the standardized observation innovation = is greater than a set threshold, it is determined that there is a human gross error; wherein the third error model is used to describe the statistical abnormal characteristics of the gross error caused by human operation in the observation data sequence and the impact mode on the data fusion process, and the third error model is specifically represented as: = ； where, is the size of the blunder occurring at time ; is the Dirac function, indicating that the error occurs only at a specific time; is the number of blunders occurring during the observation period.

9. The method for dynamic correction of engineering surveying errors based on multi-sensor data fusion according to claim 1, characterized in that, The recursive state estimation algorithm is a Kalman filter, an extended Kalman filter or an unscented Kalman filter; The specific way of the Robust processing is: when it is determined that there is a gross error in a certain observation value, it is excluded from the current state update, or its weight is reduced by increasing the variance value corresponding to it in the observation noise covariance matrix.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to realize the steps of the engineering measurement error dynamic correction method based on multi-sensor data fusion in any one of claims 1 to 9.

Images