Device and method for underwater positioning of defects in ultra-long water transfer tunnels

Abstract

The present invention discloses an underwater positioning device and method for defects in an ultra-long water transfer tunnel, the method comprising obtaining multi-source sensor data of an underwater vehicle, performing time synchronization and preprocessing on the data; establishing an acoustic environment characteristic model, identifying strong reflective identification information, and updating environmental parameters; calculating the dynamic weights of each data source, performing data consistency checks, and performing multi-source data fusion; performing trajectory optimization on the fused positioning result, performing identification point constraint processing, and performing backtracking correction on the historical trajectory; and performing multiple verifications and uncertainty assessments on the positioning result. The present invention effectively solves the multipath effect, thermocline influence, inertial navigation cumulative error, and image quality problems in complex underwater environments through multi-source data fusion and multi-level optimization strategies, and realizes high-precision positioning in ultra-long water transfer tunnels.

Description

Technical Field

[0001] The invention belongs to the field of underwater positioning, and in particular to an underwater positioning device and method for defects in an ultra-long water conveyance tunnel. Background Art

[0002] The ultra-long water transfer tunnel in the hydropower project is a key infrastructure connecting the water source and the generator set. Its safe operation is directly related to the power generation efficiency and engineering safety of the power station. This type of tunnel generally has the characteristics of "high head, long distance, and multiple bends", and is always full of water. Its structural integrity detection is of great significance to ensure the safe operation of the hydropower station.

[0003] At present, underwater robot technology has made significant progress in the field of tunnel inspection. The mainstream technical solutions include a combination of macroscopic surveys based on multi-beam sonar and local precision surveys based on high-definition cameras. In terms of positioning technology, the industry mainly uses inertial navigation systems (INS), acoustic positioning systems (LBL / USBL) and optical positioning technologies. Among them, the inertial navigation system provides continuous position solution through accelerometers and gyroscopes, the acoustic positioning system uses the time difference of sound wave propagation for three-dimensional positioning, and the optical positioning relies on image feature recognition to determine the relative position. These technologies have been well applied in conventional underwater environments.

[0004] However, the special environment of ultra-long water tunnels puts higher demands on underwater positioning systems. There are the following technical difficulties: First, in an environment with a significant thermocline effect, the sound wave propagation path bends, resulting in reduced acoustic positioning accuracy; second, multiple reflections from the tunnel wall cause serious multipath effects on acoustic signals, increasing the difficulty of signal processing; third, during the long-distance movement of underwater robots, the zero bias error of the inertial navigation system accumulates exponentially and is significantly affected by temperature changes; fourth, under turbid water conditions, optical identification is seriously disturbed, affecting image positioning accuracy; fifth, multi-source sensor data has problems such as inconsistent sampling frequency and uncertain delay, which affects the data fusion effect; finally, in the post-processing stage, the trajectory optimization algorithm is difficult to effectively process long-time series data, resulting in a significant decrease in positioning accuracy over time. These technical problems seriously restrict the actual application effect of underwater positioning systems in ultra-long water tunnels. Summary of the invention

[0005] The purpose of the invention is to provide an underwater positioning device and method for defects in an ultra-long water transfer tunnel to solve the above-mentioned problems existing in the prior art.

[0006] The technical solution is an underwater positioning method for defects in an ultra-long water transfer tunnel, comprising the following steps:

[0007] S1. Acquire multi-source sensor data of an underwater vehicle, perform time synchronization processing on the multi-source sensor data, and generate a synchronized data stream; perform signal preprocessing on the synchronized data stream, and generate a preprocessed data stream; wherein the multi-source sensor data includes inertial navigation data, acoustic positioning data, environmental sensor data, and strong reflective identification image data;

[0008] S2. Based on the preprocessed data stream, build an acoustic environment feature model, identify strong reflective sign information, update environmental parameters, and generate environmental model parameters and sign point location data;

[0009] S3. Based on the preprocessed data stream, environmental model parameters and identification point location data, the dynamic weight of each data source is calculated, and data consistency check is performed to obtain consistency score data; based on the consistency score data, multi-source data fusion operation is performed to generate fusion positioning results;

[0010] S4. Based on the fusion positioning result, trajectory optimization calculation is performed to obtain the initial optimized trajectory; based on the initial optimized trajectory and the location data of the marker points, the marker point constraint processing is performed to generate the constraint optimization result; based on the constraint optimization result, the historical trajectory is backtracked and corrected to generate the optimized trajectory data;

[0011] S5. Based on the optimized trajectory data and the fused positioning results, multiple verification operations are performed to conduct uncertainty assessment and generate the final positioning results.

[0012] The underwater defect positioning device for ultra-long water transfer tunnels includes:

[0013] at least one processor; and,

[0014] a memory communicatively connected to at least one of the processors; wherein,

[0015] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the underwater positioning method for defects in an ultra-long water transfer tunnel.

[0016] Beneficial effects: The present invention overcomes the inherent limitations of a single sensor through a multi-sensor collaborative working mechanism; adapts to complex underwater environments through dynamic acoustic models and identification; achieves the optimal combination of various types of data through adaptive weight distribution; ensures the accuracy of the positioning trajectory through multiple constraint processing; and ensures the reliability of the positioning result through uncertainty assessment. It not only achieves sub-meter positioning accuracy in ultra-long water transfer tunnels, but also has strong environmental adaptability and stability, providing reliable technical support for tunnel defect detection and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Flow chart of the method of the present invention.

[0018] Figure 2 This is a flow chart of step S1 of the present invention.

[0019] Figure 3 This is a flow chart of step S2 of the present invention.

[0020] Figure 4 This is a flow chart of step S3 of the present invention.

[0021] Figure 5 This is a flow chart of step S4 of the present invention.

[0022] Figure 6 This is a flow chart of step S5 of the present invention. DETAILED DESCRIPTION

[0023] like Figure 1 As shown, the present application proposes a method for underwater positioning of defects in an ultra-long water transfer tunnel, comprising the following steps:

[0024] S1. Acquire multi-source sensor data of an underwater vehicle, perform time synchronization processing on the multi-source sensor data, and generate a synchronized data stream; perform signal preprocessing on the synchronized data stream, and generate a preprocessed data stream; wherein the multi-source sensor data includes inertial navigation data, acoustic positioning data, environmental sensor data, and strong reflective identification image data;

[0025] S2. Based on the preprocessed data stream, build an acoustic environment feature model, identify strong reflective sign information, update environmental parameters, and generate environmental model parameters and sign point location data;

[0026] S3. Based on the preprocessed data stream, environmental model parameters and identification point location data, the dynamic weight of each data source is calculated, and data consistency check is performed to obtain consistency score data; based on the consistency score data, multi-source data fusion operation is performed to generate fusion positioning results;

[0027] S4. Based on the fusion positioning result, trajectory optimization calculation is performed to obtain the initial optimized trajectory; based on the initial optimized trajectory and the location data of the marker points, the marker point constraint processing is performed to generate the constraint optimization result; based on the constraint optimization result, the historical trajectory is backtracked and corrected to generate the optimized trajectory data;

[0028] S5. Based on the optimized trajectory data and the fused positioning results, multiple verification operations are performed to conduct uncertainty assessment and generate the final positioning results.

[0029] like Figure 2 As shown, according to one aspect of the present application, step S1 is further:

[0030] S11, using a sensor collector to respectively acquire inertial navigation data, acoustic positioning data, environmental sensor data and strong reflective sign image data of the underwater submersible, and combine them to form multi-source sensor data; wherein the inertial navigation data includes acceleration data and angular velocity data, the acoustic positioning data includes ultra-short baseline positioning data and long baseline positioning data, the environmental sensor data includes water temperature data, pressure data and sound speed data, and the strong reflective sign image data includes image data of the sign plate;

[0031] S12, based on multi-source sensor data, perform time alignment operations on data with different sampling frequencies, build a unified time base, and generate a synchronous data stream;

[0032] S13. Perform outlier detection and data noise reduction processing on the synchronous data stream to generate a preprocessed data stream.

[0033] This embodiment achieves high-precision positioning data acquisition in complex underwater environments through the collaborative collection and processing mechanism of multi-source sensor data. Specifically, inertial navigation data provides continuous position and attitude references, acoustic positioning data overcomes cumulative errors, environmental sensor data compensates the acoustic propagation model in real time, and strong reflective identification image data provides reliable absolute position calibration points. Performing time alignment operations on data with different sampling frequencies and building a unified time base solves the problem of time inconsistency of multi-source data and ensures the synchronization of data under the same time base; time synchronization processing is crucial for subsequent data fusion and analysis because it ensures the corresponding relationship between various types of data at the same time point and improves the accuracy and consistency of data processing; outlier detection of synchronized data streams can identify and remove outliers in the data to prevent them from misleading subsequent analysis; data denoising processing improves the quality and reliability of data by filtering out noise in the data; the pre-processed data stream is cleaner and more accurate, providing a reliable foundation for subsequent model building and data fusion. This embodiment realizes efficient processing and fusion of multi-source sensor data of underwater submersibles, provides a solid data foundation for the precise positioning and trajectory optimization of underwater submersibles, and improves the overall performance and reliability of the system.

[0034] According to one aspect of the present application, step S11 is further:

[0035] S111, collecting inertial navigation data from an inertial navigation sensor, compensating and correcting the inertial navigation data through a dynamic calibration coefficient to generate a compensation signal; performing zero bias estimation and temperature drift correction on the compensation signal to generate a correction signal; based on the correction signal, calculating static and dynamic response characteristics, judging data validity, and generating effective inertial navigation data;

[0036] S112, receiving acoustic positioning data from the acoustic positioning system, calculating the signal arrival time difference and the sound wave propagation path, and generating path data; performing multipath component decomposition on the path data, identifying the direct wave and the reflected wave components, and generating decomposition data; constructing an acoustic wave propagation probability model based on the decomposition data, screening the optimal propagation path, and generating an optimized path; performing signal strength calibration and phase compensation based on the optimized path, and generating effective acoustic data;

[0037] S113, collecting environmental sensor data from environmental sensors, constructing an environmental parameter spatial distribution model, and generating distribution data; performing gradient calculation and outlier identification on the distribution data to generate verification data; calculating parameter change trends based on the verification data, judging data rationality, and generating valid environmental data;

[0038] S114, acquiring image data of a strong reflective mark from an image acquisition system, performing adaptive histogram equalization processing, and generating an enhanced image; performing edge strength calculation and contrast evaluation on the enhanced image, and generating feature data; calculating an image clarity index and an illumination uniformity index based on the feature data, and generating evaluation data; constructing an image quality scoring model based on the evaluation data, screening valid images, and generating valid image data;

[0039] S115. Based on the effective inertial navigation data, effective acoustic data, effective environmental data and effective image data, a time-space correspondence matrix is ​​constructed to generate corresponding data; data integrity verification and time consistency check are performed on the corresponding data to generate verification data; based on the verification data, a multi-source data association index is constructed to generate multi-source sensor data.

[0040] In one embodiment of the present application, the process of performing zero bias estimation and temperature drift correction is specifically as follows: Inertial Measurement Unit (IMU) compensation model Y(t) = X(t) + B(t) + K(T) · ΔT(t) + ε(t); wherein X(t) is the original inertial navigation signal, including acceleration a(t) and angular velocity ω(t); B(t) is the zero bias term, B(t) = B0 + ∫β(τ)dτ, B0 is the initial zero bias, β(τ) is the zero bias drift rate, d is the integral sign, τ is the integral variable, usually representing the time variable; K(T) is the temperature correlation coefficient matrix, ΔT(t) is the temperature change; ε(t) is the random error term, which obeys the Gaussian distribution N(0, σ 2 ); t is the current time; T is the temperature; Y(t) is the compensated signal.

[0041] Acoustic multipath decomposition method: Acoustic signal model S(t) = Σ[Ai·D(t-τi)·exp(-μi·Li)] + N(t); where D(t) is the direct sound wave signal; Ai is the amplitude attenuation coefficient of the i-th path; τi is the time delay of the i-th path; μi is the attenuation coefficient of the i-th path; Li is the propagation distance of the i-th path; N(t) is the background noise; t is the time; i is the path number; S(t) is the received superposition signal.

[0042] This embodiment achieves the acquisition of high-quality raw data in an underwater environment through the collaborative working mechanism of a multi-sensor acquisition system. In the inertial navigation data acquisition, dynamic calibration coefficients are used for compensation correction, combined with zero bias estimation and temperature drift correction, which effectively solves the problem of cumulative errors of the inertial navigation system when working for a long time. In the processing of acoustic positioning data, by analyzing the signal arrival time difference and the sound wave propagation path, combined with multipath component decomposition technology, the direct wave and reflected wave components can be accurately identified, and the optimal path can be screened through the sound wave propagation probability model. This embodiment improves the positioning accuracy in the complex environment of the tunnel. For environmental sensor data, by establishing an environmental parameter spatial distribution model, combined with gradient calculation and outlier recognition, the impact of underwater environmental changes on positioning can be monitored and compensated in real time. In the acquisition of strong reflective sign images, adaptive histogram equalization and quality scoring models are used to obtain clear sign images even under turbid water conditions. This embodiment provides a high-quality data foundation for subsequent data fusion and positioning calculations.

[0043] In one embodiment of the present application, step S111 also includes: based on the effective inertial navigation data, extracting the zero-position signal of the acceleration and angular velocity to generate zero-position data; performing wavelet decomposition on the zero-position data, identifying the drift trend, and generating drift data; establishing a random walk model based on the drift data to generate error model data; calculating the compensation coefficient according to the error model data, correcting the original signal, and generating zero-bias correction data.

[0044] According to one aspect of the present application, step S112 is further:

[0045] S112a, receiving acoustic positioning data from an acoustic positioning system, calculating the signal arrival time difference and the sound wave propagation path, and generating path data; based on the path data, calculating the geometric propagation characteristics of the underwater acoustic channel, and generating geometric feature data; performing horizontal layered modeling on the geometric feature data, identifying the sound wave reflection surface, and generating reflection surface data; calculating the propagation delay of the direct wave, the surface reflection wave, and the bottom reflection wave based on the reflection surface data, and generating delay data; performing multipath effect compensation calculation according to the delay data, and generating compensation path data;

[0046] S112b. Based on the compensation path data, the amplitude attenuation coefficient of each path signal is calculated to generate attenuation coefficient data; the attenuation coefficient data is normalized to generate normalized data; a multipath signal superposition model is established based on the normalized data to generate superposition data; a multipath suppression operation is performed according to the superposition data to generate suppressed acoustic data.

[0047] According to one aspect of the present application, step S13 is further:

[0048] S131, read the synchronous data stream, construct the data feature space mapping matrix, and generate the feature matrix; perform singular value decomposition operation on the feature matrix, extract the main feature components, and generate decomposition data; establish an anomaly detection threshold model based on the decomposition data and generate threshold parameters.

[0049] S132, reading the synchronous data stream and threshold parameters, calculating the data deviation index, and generating deviation data; performing spatial correlation analysis on the deviation data, identifying local abnormal patterns, and generating abnormal patterns; performing adaptive abnormal replacement operations based on the abnormal patterns, and generating corrected data.

[0050] S133, read the correction data, establish a multi-scale signal decomposition model, and generate decomposition coefficients; perform energy distribution analysis on the decomposition coefficients, identify noise characteristics, and generate noise characteristics; build an adaptive signal reconstruction model based on the noise characteristics and generate reconstructed data.

[0051] S134, reading the reconstructed data, calculating the signal-to-noise ratio evaluation index, and generating evaluation data; performing signal quality optimization processing on the evaluation data to generate optimized data; establishing a data credibility scoring model based on the optimized data to generate a preprocessed data stream.

[0052] In the data preprocessing stage, this embodiment effectively improves the quality of the original data through multi-level signal processing technology. First, by constructing a data feature space mapping matrix and performing singular value decomposition, the main characteristic components of the data can be extracted and an abnormal detection threshold model can be established, so that the mutation points and abnormal values ​​in various sensor data can be accurately identified and processed. In terms of noise processing, multi-scale signal decomposition and energy distribution analysis are used to accurately identify and filter out various noise components. Through the adaptive signal reconstruction model, not only the effective information of the original signal is retained, but also the signal-to-noise ratio is improved. This embodiment provides high-quality data input for subsequent positioning calculations and improves the overall performance of the system.

[0053] like Figure 3 As shown, according to one aspect of the present application, step S2 is further:

[0054] S21, constructing an acoustic propagation model based on the environmental sensor data in the preprocessed data stream, and generating acoustic propagation model parameters, i.e., environmental model parameters;

[0055] S22, based on the strong reflective mark image data in the preprocessed data stream, image feature extraction and mark position calculation are performed to generate mark point position data;

[0056] S23. Based on the acoustic propagation model parameters and the environmental sensor data in the preprocessed data stream, an environmental parameter state transfer relationship is constructed to generate an environmental state vector.

[0057] This embodiment establishes an accurate acoustic environment characteristic model and identification system in the special environment of an ultra-long water transfer tunnel. By constructing an acoustic propagation model, the propagation characteristics of acoustic signals under different environmental conditions can be accurately described, providing a reliable basis for subsequent data processing and analysis, and improving the accuracy of acoustic positioning and detection. Through image feature extraction, the position of highly reflective signs can be accurately located, providing important reference data for environmental modeling and positioning, and the positioning accuracy of the overall system can be improved by solving the position data of the marking points. By constructing the state transfer relationship of environmental parameters, changes in environmental conditions can be reflected in real time; the generation of environmental state vectors enables the system to dynamically adjust and optimize the acoustic propagation model, improving the adaptability and robustness of the system. This embodiment realizes efficient processing and modeling of environmental data, improves the overall performance and reliability of the system, and provides solid technical support for accurate positioning and detection in complex environments.

[0058] According to one aspect of the present application, step S21 is further:

[0059] S211, based on the environmental sensor data in the preprocessed data stream, the three-dimensional space is divided into uniform grid units, the center coordinates of each grid unit are calculated, and grid coordinate data are generated; based on the environmental sensor data of each grid unit, the corresponding sound speed value is calculated to generate sound speed distribution data; based on the grid coordinate data and the sound speed distribution data, a spatial interpolation operation is performed to generate node data; based on the node data, the sound speed change rate between adjacent grid points is calculated to generate gradient field data;

[0060] S212, based on the gradient field data, calculating the rate of change of the speed of sound in each direction, extracting the bending characteristics of the sound wave propagation path, and generating curvature data; based on the curvature data, calculating the energy attenuation coefficient of the sound wave during the propagation process, and generating attenuation data; based on the attenuation data, calculating the energy loss at different propagation distances, and generating energy distribution;

[0061] S213, based on the gradient field data and energy distribution, calculating the reflection angle of the sound wave at different medium interfaces to generate reflection data; based on the reflection data, analyzing the multipath propagation path, calculating the time delay difference of each path, and generating the propagation path; based on the propagation path, calculating the interface scattering intensity to generate scattering data; based on the scattering data and energy distribution, calculating the sound field intensity distribution to generate a prediction result;

[0062] S214. Based on the prediction results, the time-varying characteristics of the sound field parameters are calculated to generate time-varying data; based on the time-varying data, the environmental change law is extracted, the parameter update cycle is calculated, and the periodic data is generated; based on the periodic data, the parameter update threshold is constructed to generate the update parameter; based on the update parameter and the prediction results, the sound field characteristic quantity is calculated to generate the acoustic propagation model parameters, i.e., the environmental model parameters.

[0063] In one embodiment of the present application, the gradient field data is generated as follows: the sound velocity gradient equation c(z) = c0 +k·tanh[(z-z0) / h]; wherein c0 is the reference sound velocity; k is the sound velocity variation coefficient; z is the depth; z0 is the center depth of the thermocline; h is the characteristic parameter of the thermocline thickness; c(z) is the sound velocity at depth z; the sound line equation d 2 r / dz 2 = -[1 / c(z)]·[dc(z) / dz]·[1-(c0 / c(z)) 2 ·sin 2 θ0]; θ0 is the incident angle.

[0064] This embodiment solves the modeling problem of the complex acoustic environment in the ultra-long water transfer tunnel by establishing an accurate acoustic propagation model. First, by dividing the three-dimensional space into uniform grid units, the distribution characteristics of the sound speed in space can be accurately described. The sound speed distribution calculated by combining the water temperature and pressure data is used to establish a complete sound field model through spatial interpolation operations. In particular, when calculating the bending characteristics of the sound wave propagation path, the energy attenuation and path deflection of the sound wave during the propagation process are accurately predicted by analyzing the rate of change of the sound speed in each direction. When dealing with interface scattering problems, a complete multipath propagation model is established by calculating the reflection angle and scattering intensity of different medium interfaces. This embodiment not only improves the accuracy of acoustic positioning, but also can adapt to the complex and changeable hydrological environment inside the tunnel, and provides a reliable acoustic propagation reference model for the underwater positioning system.

[0065] In one embodiment of the present application, step S211 also includes: calculating the temperature gradient in the vertical direction based on the grid coordinate data and the sound speed distribution data, and generating temperature gradient data; performing threshold analysis on the temperature gradient data, identifying the location of the thermocline, and generating thermocline data; calculating the sound wave refraction angle based on the thermocline data, and generating refraction angle data; correcting the sound wave propagation path according to the refraction angle data, and generating corrected path data.

[0066] According to one aspect of the present application, step S22 is further:

[0067] S221. Read the image data in the preprocessing data stream, downsample the image according to different resolutions, calculate the feature points of each scale image, and generate a scale image; perform local area histogram equalization calculation on the scale image to improve the contrast of the reflective area and generate a feature area; calculate the pixel intensity distribution in the feature area, count the spatial distribution characteristics of the highlight area, and generate intensity data.

[0068] S222, read the intensity data, calculate the coordinate sequence of the boundary points of the highlight area, extract the contour feature parameters, and generate shape data; perform feature matching on the shape data and the preset identification plate template, calculate the similarity score, and generate contour data; calculate the rotation angle and inclination angle of the identification plate in the image plane based on the contour data, and generate posture parameters.

[0069] S223, read the posture parameters and contour data, calculate the transformation matrix from the image plane to the spatial coordinate system, and generate transformation data; use the transformation data to eliminate perspective deformation, calculate the actual size ratio of the identification plate, and generate correction data; calculate the position coordinates of the identification plate in the spatial coordinate system based on the correction data, and generate coordinate data.

[0070] S224, read the coordinate data, calculate the statistical distribution of the position of the identification point in multiple frames of images, and generate statistical data; calculate the standard deviation of the position measurement based on the statistical data, evaluate the measurement accuracy, and generate accuracy data; combine the accuracy data and statistical data to calculate the optimal position estimate and generate the identification point position data.

[0071] This embodiment aims at the problem of identifying reflective signs in ultra-long water transfer tunnels and realizes high-precision solution of the position of the marking points. Through multi-scale image analysis technology, feature points are first extracted at different resolutions, and the contrast of the reflective area is improved by local area histogram equalization. In the process of extracting the outline of the sign plate, the position and posture of the sign plate are accurately identified by calculating the coordinate sequence of the boundary points in the highlighted area and combining the preset template for feature matching. In particular, when dealing with perspective deformation problems, the actual position of the sign plate is accurately calculated by establishing a transformation matrix from the image plane to the spatial coordinate system. Through statistical analysis and accuracy evaluation of multiple frames of images, the reliability of position measurement is improved, and a high recognition accuracy rate can be maintained even in turbid water environments.

[0072] In one embodiment of the present application, it also includes:

[0073] S220a. Based on the image data in the preprocessed data stream, calculate the grayscale distribution histogram of the image to generate grayscale distribution data; perform statistical analysis on the grayscale distribution data, evaluate the turbidity level, and generate turbidity data; establish an image enhancement parameter model based on the turbidity data to generate enhancement parameter data; perform adaptive contrast adjustment according to the enhancement parameter data to generate enhanced image data.

[0074] S220b, based on the enhanced image data, perform Wiener filtering denoising operation to generate denoised data; perform wavelet transform on the denoised data to extract multi-scale features and generate feature data; calculate image quality score based on the feature data to generate quality score data; filter valid image frames according to the quality score data to generate filtered image data.

[0075] According to one aspect of the present application, step S23 is further:

[0076] S231. Read the acoustic propagation model parameters and the environmental sensor data in the preprocessed data stream, calculate the time series characteristics of the environmental parameters, and generate sequence data; perform wavelet decomposition on the sequence data, extract the long-term trend component and the short-term fluctuation component, and generate trend data; calculate the parameter change rate and acceleration based on the trend data, and generate prediction parameters.

[0077] S232, read the prediction parameters, calculate the change amount of the environmental parameters at adjacent time points, extract the change rate sequence, and generate rate data; perform statistical analysis on the rate data, calculate the distribution characteristics of the change rate, and generate distribution data; identify the change points that exceed the normal range based on the distribution data, and generate mutation data; use the mutation data to determine the parameter fluctuation threshold and generate the fluctuation parameter.

[0078] S233, read the fluctuation parameters and prediction parameters, calculate the correlation coefficient matrix between the environmental parameters, and generate relevant data; use the relevant data to calculate the parameter state transition probability and generate probability data; calculate the trigger condition for state update based on the probability data and generate threshold data; calculate the parameter adjustment amount based on the threshold data and generate the adjustment parameter.

[0079] S234, read the adjustment parameters, calculate the stability index of the environmental parameters, and generate stable data; use the stable data to analyze the continuity characteristics of the parameter changes, calculate the credibility index, and generate credible data; update the environmental state components according to the credible data, calculate the values ​​of each component of the state vector, and generate the environmental state vector.

[0080] In one embodiment of the present application, the state transfer equation E(t+1) = A(t)·E(t) + B(t)·U(t) +W(t); where E(t) is the environment state vector [c(t), T(t), p(t)] T, respectively represent the speed of sound, temperature and pressure; A(t) is the state transfer matrix; B(t) is the control matrix; U(t) is the control input; W(t) is the process noise, W(t)~N(0,Q(t)); t is the time index; prediction uncertainty σE 2 = tr(A·P·A T + Q); P is the state covariance matrix, Q is the covariance matrix of process noise, and tr( ) is the trace of the matrix.

[0081] This embodiment realizes the dynamic update of the acoustic environment model by establishing the state transition relationship of the environmental parameters. First, the time series characteristics of the environmental parameters are analyzed by wavelet decomposition technology, and the long-term trend component and the short-term fluctuation component are extracted to accurately predict the parameter change trend. When dealing with the problem of sudden changes in environmental parameters, a complete parameter fluctuation monitoring mechanism is established by analyzing the distribution characteristics of the change rate and identifying the mutation point. By calculating the correlation coefficient matrix and state transition probability between environmental parameters, the intelligent update of the parameter state is realized. This embodiment not only improves the adaptability of the model to environmental changes, but also ensures the accuracy of acoustic positioning when the environmental parameters change drastically. By updating the environmental state vector in real time, an accurate description of the current acoustic environment can be maintained at all times.

[0082] like Figure 4 As shown, according to one aspect of the present application, step S3 is further:

[0083] S31, based on the preprocessed data stream, environmental model parameters, identification point location data and environmental state vector, evaluate the quality index and reliability index of each data source; based on the quality index and reliability index, calculate the dynamic weight coefficient and generate a data weight vector;

[0084] S32, based on the preprocessed data stream and the data weight vector, performing consistency evaluation operation on the multi-source data to generate consistency score data;

[0085] S33: Based on the preprocessed data stream, data weight vector and consistency score data, perform adaptive data fusion operation to generate a fused positioning result.

[0086] This embodiment aims at the fusion problem of multi-source data in the complex environment of ultra-long water transfer tunnels, and realizes the adaptive weight allocation and fusion optimization of data. By calculating the dynamic weight coefficient and generating the data weight vector, it is ensured that each data source is given an appropriate weight according to its quality and reliability during the fusion process, thereby improving the accuracy and reliability of data fusion; the dynamic weight calculation can adaptively adjust the influence of each data source and enhance the robustness of the system. By evaluating the consistency between each data source, identifying and processing conflicts and inconsistencies in the data, the generated consistency score data provides an important reference basis for subsequent data fusion, ensuring the reliability and consistency of the fusion results. Through adaptive data fusion, the data fusion strategy can be dynamically adjusted according to the real-time quality and consistency evaluation results, thereby improving the accuracy and stability of the positioning results. The fused positioning results combine the advantages of multi-source data and provide more accurate and reliable positioning information. This embodiment realizes the efficient evaluation and fusion of multi-source data, improves the accuracy and stability of the positioning results, and ultimately improves the overall performance and reliability of the system.

[0087] According to one aspect of the present application, step S31 is further:

[0088] S311, read the preprocessed data stream, acoustic propagation model parameters, identification point location data and environmental state vector, calculate the signal-to-noise ratio index and integrity index of each data source, and generate the quality index; perform standard deviation normalization on the quality index, calculate the standard score, and generate the score data; calculate the temporal continuity score and spatial consistency score based on the score data to generate the quality score.

[0089] S312, read the quality score, calculate the historical reliability statistics of the data source, and generate historical data; use the historical data to analyze the reliability change trend, calculate the stability coefficient, and generate stable data; calculate the predicted value of future reliability based on the stable data and generate a prediction result.

[0090] S313, read the quality score and prediction results, calculate the degree of complementarity between the data sources, and generate complementary data; use the complementary data to analyze the gain effect of data fusion, calculate the synergy coefficient, and generate gain data; calculate the optimal weight allocation plan based on the gain data and generate optimized weights.

[0091] S314, read the optimized weight, calculate the gradient value of the weight adjustment, and generate gradient data; use the gradient data to perform weight smoothing, calculate the transition coefficient, and generate transition data; calculate the final weight normalization value based on the transition data to generate a data weight vector.

[0092] This embodiment solves the weight optimization problem in multi-source data fusion through a dynamic weight allocation mechanism. First, by analyzing the signal-to-noise ratio index and integrity index of each data source, a scoring system based on data quality is established. In the weight calculation process, by analyzing the reliability statistics and change trends of historical data, the reliability of each data source can be accurately predicted. In particular, when calculating the degree of complementarity between data sources, the optimal weight allocation scheme is established by analyzing the gain effect of data fusion. This embodiment not only takes into account the real-time quality of the data, but also combines historical performance and predicted reliability. Through the smooth transition processing of weights, the positioning result jump caused by sudden changes in weights is avoided. The weights of each data source can be automatically adjusted according to changes in the tunnel environment, ensuring the stability and accuracy of the fused positioning results.

[0093] According to one aspect of the present application, step S32 is further:

[0094] S321. Based on the preprocessed data stream and the data weight vector, calculate the distribution characteristics of each data source in the spatial domain to generate spatial data; based on the spatial data, calculate the distance matrix and correlation coefficient between the data points to generate relevant data; based on the relevant data, identify spatial abnormal points and conflict points, calculate the conflict index, and generate the conflict parameter;

[0095] S322. Based on the conflict parameters and the relevant data, the autocorrelation function and the cross-correlation function of the data sequence are calculated to generate sequence data; based on the sequence data, the change law of the time dimension is extracted, the trend characteristics are calculated, and the trend data is generated; based on the trend data, the time series coordination index between the data is calculated to generate the coordination parameter;

[0096] S323. Based on the coordination parameters, calculate the information entropy and conditional entropy between the data sources to generate entropy value data; based on the entropy value data, analyze the information redundancy and complementarity, calculate the information gain, and generate gain data; based on the gain data, calculate the consistency quantitative index and generate the quantitative parameter;

[0097] S324. Based on the quantitative parameters, the comprehensive score of the multi-dimensional evaluation index is calculated to generate comprehensive data; based on the comprehensive data, a multi-level fuzzy evaluation is performed, the membership value is calculated, and the membership data is generated; based on the membership data, the final consistency evaluation result is calculated to generate consistency score data.

[0098] This embodiment implements all-round data quality verification for the consistency assessment problem of multi-source data. By analyzing the spatial distribution characteristics and correlation of data, spatial anomalies and conflict points can be accurately identified. In time series consistency analysis, the inherent laws of data changes are extracted by calculating the autocorrelation function and cross-correlation function of the data sequence. In particular, in the information entropy analysis link, a consistency evaluation system based on information theory is established by calculating the information redundancy and complementarity between data sources. This embodiment can not only discover potential conflicts between data, but also evaluate the effectiveness of data fusion. This embodiment provides a quantitative description of data consistency and provides reliable quality assurance for data fusion.

[0099] According to one aspect of the present application, step S33 is further:

[0100] S331, read the preprocessed data stream, data weight vector and consistency score data, calculate the kinematic characteristics of the state variables, and generate motion data; use the motion data to extract the acceleration and angular velocity components, calculate the state prediction value, and generate prediction data; calculate the propagation characteristics of the prediction error based on the prediction data and generate error data; calculate the prediction correction amount based on the error data and generate correction parameters.

[0101] S332, read correction parameters and prediction data, calculate the residual sequence between the observed data and the predicted value, and generate residual data; use the residual data to analyze the noise characteristics, calculate the variance and mean, and generate statistical data; calculate the optimal gain coefficient matrix based on the statistical data to generate gain parameters.

[0102] S333, read the gain parameters and statistical data, calculate the update amount of the state estimate, and generate the update data; use the updated data to correct the state quantity, calculate the corrected estimate value, and generate the estimated data; calculate the convergence criterion and iteration parameters based on the estimated data, and generate the optimization parameters.

[0103] S334, read the optimization parameters and estimated data, calculate the posterior probability distribution of the state estimation, and generate probability data; use the probability data to analyze the reliability of the estimation result, calculate the confidence index, and generate confidence data; calculate the final fusion positioning value based on the confidence data and generate the fusion positioning result.

[0104] In one embodiment of the present application, the state estimation equation X*(t) = ΣWi(t)·[X*i(t) + Ki(t)·(Zi(t) - Hi·X*i(t))]; wherein X*i(t) is the state prediction value of the i-th data source; Wi(t) is the dynamic weight; Ki(t) is the gain matrix, Ki(t) = Pi(t)·Hi T ·(Hi·Pi(t)·Hi T + Ri) -1; Zi(t) is the observation value; Hi is the observation matrix; Pi(t) is the prediction error covariance; Ri is the observation noise covariance; t is the time; i is the data source number.

[0105] The process of calculating the optimal gain coefficient matrix is: Gain matrix K(t) = P(t)·H(t) T ·[H(t)·P(t)·H(t) T + R(t) + Q(t)] -1 ; where P(t) is the prior error covariance; H(t) is the observation matrix; R(t) is the observation noise covariance; Q(t) is the process noise covariance, Q(t) = diag[q1·exp(-α1·SNR), q2·exp(-α2·v), q3·exp(-α3·θ)]; SNR is the signal-to-noise ratio; v is the motion speed; θ is the attitude angle; qi, αi are adjustment parameters; t is the current time.

[0106] This embodiment realizes adaptive state estimation and optimization in the process of multi-source data fusion. First, an accurate state prediction model is established by analyzing the kinematic characteristics of the state variables. When processing the observed data, the optimal gain coefficient is dynamically adjusted by analyzing the statistical characteristics of the residual sequence. In particular, in the state estimation update link, reliable state estimation is achieved by calculating the posterior probability distribution of the state quantity. This embodiment not only improves the accuracy of state estimation, but also enhances the system's adaptability to environmental changes through an adaptive mechanism. By dynamically evaluating the confidence of the fusion results, the fusion strategy can be adjusted in real time to ensure the reliability of the positioning results.

[0107] like Figure 5 As shown, according to one aspect of the present application, step S4 is further:

[0108] S41, based on the fusion positioning results, construct the pose nodes and constraint edge relationships, perform trajectory smoothing optimization operations, and generate an initial optimized trajectory;

[0109] S42, based on the position data of the identification points, using a constrained optimization algorithm to perform constraint processing on the initial optimization trajectory to generate a constrained optimization result;

[0110] S43. Based on the constraint optimization result, perform a backtracking calculation on the historical data within a preset time window to generate the historical data after the backtracking calculation; perform a trajectory correction process on the historical data after the backtracking calculation to generate optimized trajectory data.

[0111] This embodiment effectively solves the problem of trajectory accuracy of the underwater robot in the tunnel by performing trajectory optimization and constraint processing on the fusion positioning results. First, by establishing the relationship between pose nodes and constraint edges, abnormal jitters and jumps in the trajectory can be identified and corrected. In particular, in the optimization of identification point constraints, the precise position information of known identification points is used to establish a strict spatial constraint relationship through the constraint optimization algorithm, which not only improves the smoothness of the trajectory, but also ensures the consistency of the trajectory with the actual tunnel contour. By performing historical data backtracking calculations, the historical trajectory can be effectively corrected. This embodiment improves the stability and reliability of the underwater positioning system during long-term operation.

[0112] According to one aspect of the present application, step S41 is further:

[0113] S411. Read the fusion positioning result, calculate the time series relationship of the posture data, and generate time series data; use the time series data to extract the spatial displacement and posture change between adjacent posture points to generate change data; calculate the connection relationship and weight coefficient between nodes based on the change data to generate node data; calculate the attribute parameters of the spatial constraint in combination with the node data to generate constraint parameters.

[0114] S412, read constraint parameters and node data, calculate motion speed and acceleration constraints, and generate speed data; use speed data to analyze motion continuity characteristics, calculate physical feasibility indicators, and generate feasible data; construct a coefficient matrix of the optimization objective function based on the feasible data and generate an optimization function.

[0115] S413, read the optimization function, calculate the weight coefficient of the error term, and generate weight data; use the weight data to perform outlier robustness processing, calculate the kernel function parameters, and generate kernel function data; calculate the step size and termination condition of iterative optimization based on the kernel function data, and generate solution parameters.

[0116] S414, reading the solution parameters and kernel function data, calculating the gradient value of the state increment, and generating gradient data; using the gradient data to perform iterative optimization calculations, analyzing the convergence performance, and generating convergence data; calculating the final smooth trajectory coordinates based on the convergence data, and generating an initial optimization trajectory.

[0117] In one embodiment of the present application, the optimization objective function J = Σ||ei,j|| 2 Ωi,j +Σ||ck|| 2 Λk; where ei,j is the error term between adjacent pose nodes, ei,j = log(T(i-1)·Ti,j·Tj); Ti,j is the observed relative pose; Ωi,j is the information matrix; ck is the constraint term; Λk is the constraint weight matrix; i,j is the node number; k is the constraint number; log(·) is the Lie algebra mapping.

[0118] This embodiment implements trajectory smoothing based on graph optimization for underwater robot trajectory optimization problems. By analyzing the temporal relationship of posture data, the spatial displacement and posture change between adjacent posture points are accurately extracted, and a complete node constraint network is established. In motion constraint processing, by analyzing the velocity and acceleration characteristics, an optimization objective function that conforms to physical laws is established. Especially in the outlier processing link, the robustness of the optimization algorithm is enhanced by the kernel function technology, and the influence of abnormal data on trajectory optimization is effectively suppressed. This embodiment not only ensures the geometric continuity of the trajectory, but also achieves the global optimal solution through iterative optimization. Through adaptive step size control and convergence analysis, it is possible to improve the computational efficiency while ensuring the optimization effect, providing reliable technical support for real-time trajectory optimization.

[0119] According to one aspect of the present application, step S42 is further:

[0120] S421, read the initial optimized trajectory and the identification point position data, calculate the spatial correspondence between the trajectory points and the identification points, and generate corresponding data; use the corresponding data to perform nearest neighbor search and distance calculation, extract matching point pairs, and generate matching data; calculate the strength coefficient of the position constraint based on the matching data to generate constraint data.

[0121] S422, read the constraint data, calculate the deformation amount of the local trajectory segment, and generate deformation data; use the deformation data to analyze the trajectory bending degree and stress distribution, calculate the deformation energy, and generate stress data; calculate the deformation compensation coefficient based on the stress data and generate compensation parameters.

[0122] S423, read compensation parameters and constraint data, calculate the spatial range of constraint influence, and generate range data; use range data to analyze the attenuation law of constraint force, calculate weight attenuation coefficient, and generate attenuation data; calculate the intensity distribution of constraint application based on attenuation data, and generate optimization parameters.

[0123] S424, read the optimization parameters, calculate the displacement of the trajectory correction, and generate displacement data; use the displacement data to evaluate the curve smoothness, calculate the continuity index, and generate continuous data; calculate the final corrected trajectory coordinates based on the continuous data and generate the constrained optimization result.

[0124] In one embodiment of the present application, the identification point constraint algorithm: the constraint optimization function L(X) = ||X - X0|| 2 W+ Σλm·||f(X,Pm)|| 2Σm; X is the optimized trajectory point set; X0 is the initial trajectory point set; W is the trajectory smoothing weight matrix; f(X, Pm) is the constraint function from the trajectory point to the marker point Pm; λm is the constraint strength coefficient of the mth marker point, λm= exp(-d 2 m / σ 2 ), dm is the distance to the marker point, Σm is the position covariance of the marker point, m is the marker point number, and σ is the influence radius parameter.

[0125] This embodiment improves the absolute positioning accuracy of the trajectory by optimizing the identification point constraints. First, by analyzing the spatial correspondence between the trajectory points and the identification points, an accurate pair of matching points is established. During the deformation processing, the trajectory correction based on the physical model is realized by analyzing the degree of curvature and stress distribution of the local trajectory segment. In particular, during the constraint propagation process, a reasonable constraint influence range is established by analyzing the spatial attenuation law of the constraint force. This embodiment not only corrects the cumulative error of the trajectory, but also avoids sudden changes in the trajectory through soft propagation of constraints. By comprehensively evaluating the continuity and smoothness of the trajectory, a natural transition of the trajectory is achieved while ensuring accuracy.

[0126] According to one aspect of the present application, step S43 is further:

[0127] S431, read the constraint optimization result, calculate the segmentation points of the time series, and generate segmentation data; use the segmentation data to analyze the overlapping area of ​​adjacent windows, calculate the cross-correlation coefficient, and generate correlation data; calculate the transition coefficient between windows based on the correlation data and generate transition parameters.

[0128] S432, read transition parameters and segmented data, calculate the correlation characteristics of historical states, and generate correlation data; use the correlation data to analyze the continuity of state transfer, calculate the rate of change of state quantities, and generate change data; calculate the correction amount of state reconstruction based on the change data and generate backtracking parameters.

[0129] S433, read the backtracking parameters, calculate the propagation characteristics of the error accumulation, and generate the accumulated data; use the accumulated data to analyze the error distribution law, calculate the compensation coefficient, and generate the compensation data; calculate the allocation weight of the error correction according to the compensation data and generate the allocation parameter.

[0130] S434, read the step allocation parameters and compensation data, calculate the correction amount of the global trajectory, and generate correction data; use the correction data to calculate the trajectory smoothness, evaluate the correction effect, and generate evaluation data; calculate the final optimized trajectory coordinates based on the evaluation data and generate optimized trajectory data.

[0131] In one embodiment of the present application, the process of generating backtracking parameters is specifically as follows: historical state correction equation X"(t-τ) = X(t-τ) + G(τ)·[X'(t) - X(t)]; wherein X(t-τ) is the historical state τ time ago; X'(t) is the current optimization state; G(τ) is the backtracking gain function, G(τ) = exp(-λ·τ); λ is the attenuation coefficient; τ is the backtracking time interval; t is the current moment; X"(t-τ) is the corrected historical state.

[0132] This embodiment implements trajectory optimization based on time windows in the historical trajectory backtracking correction. By analyzing the segmentation characteristics of the time series, a smooth transition mechanism between adjacent windows is established. In the state backtracking process, continuous state reconstruction is achieved by analyzing the correlation characteristics of historical states. In particular, in the error processing link, a reasonable error compensation strategy is established by analyzing the propagation characteristics of error accumulation. This embodiment not only corrects the accumulated errors in the historical trajectory, but also ensures the overall consistency of the trajectory through global optimization. By dynamically evaluating the correction effect, the optimization strategy can be adaptively adjusted to ensure the reliability and stability of the trajectory correction.

[0133] like Figure 6 As shown, according to one aspect of the present application, step S5 is further:

[0134] S51, based on the optimized trajectory data and the fused positioning result, perform position consistency calculation, trajectory smoothness calculation and positioning accuracy calculation to generate verification result data;

[0135] S52, calculating uncertainty parameters of the positioning result based on the optimized trajectory data and the verification result data, and generating uncertainty assessment data;

[0136] S53, based on the optimized trajectory data, verification result data and uncertainty assessment data, a comprehensive evaluation calculation is performed to generate a final positioning result including three-dimensional coordinates, confidence and accuracy assessment.

[0137] This embodiment realizes multi-dimensional verification and uncertainty assessment of positioning results. By calculating the spatial consistency of the optimized trajectory and the fused positioning results, the trajectory smoothness is evaluated and a complete quality assessment system is established. In particular, in the positioning accuracy analysis, the uncertainty of the positioning results is accurately quantified. In the long-term operation process, through comprehensive evaluation calculations, the output positioning results not only contain accurate three-dimensional coordinates, but also come with detailed confidence and accuracy assessment information, providing a reliable location reference for tunnel defect detection. This embodiment improves the credibility of the positioning results and provides strong support for subsequent defect diagnosis and maintenance decisions.

[0138] According to one aspect of the present application, step S51 is further:

[0139] S511, read the optimized trajectory data and the fused positioning results, calculate the spatial corresponding points of the two sets of trajectories, and generate corresponding data; use the corresponding data to calculate the deviation value of the three-dimensional coordinates, analyze the spatial distribution characteristics, and generate deviation data; calculate the quantitative index of position consistency based on the deviation data and generate consistency parameters.

[0140] S512, reading the optimized trajectory data, calculating the tangent vector and curvature value of the trajectory line, and generating curvature data; using the curvature data to analyze the geometric continuity of the trajectory, calculating the smoothness index, and generating smoothing data; calculating the trajectory quality score according to the smoothing data, and generating a smoothing parameter.

[0141] S513, read consistency parameters and smoothing parameters, calculate the error distribution of positioning accuracy, and generate error data; use the error data to analyze the error propagation chain, calculate the uncertainty components, and generate component data; calculate the comprehensive accuracy score based on the component data and generate accuracy parameters.

[0142] S514, read the accuracy parameters, calculate the combined weights of multiple indicators, and generate weight data; use the weight data to perform multi-criteria evaluation calculations, obtain a comprehensive score, and generate scoring data; calculate the final verification conclusion based on the scoring data and generate verification result data.

[0143] This embodiment achieves a comprehensive quality assessment of the positioning results through a multiple verification mechanism. First, by comparing the optimized trajectory and fusing the positioning results, a complete spatial consistency assessment system is established. In the trajectory evaluation process, the quantitative evaluation of the trajectory smoothness is achieved by analyzing the geometric features of the trajectory line (such as tangent vectors and curvature). In particular, in the accuracy assessment link, a complete accuracy assessment model is established by analyzing the error propagation chain and uncertainty components. This embodiment can not only detect potential positioning anomalies, but also provide comprehensive quality assessment results through multi-criteria evaluation. By establishing a scientific scoring system, the reliability of the positioning results can be objectively evaluated, providing a credibility basis for subsequent data applications.

[0144] According to one aspect of the present application, step S52 is further:

[0145] S521, read the optimized trajectory data and verification result data, extract the position coordinate sequence and the timestamp sequence, construct the state sampling matrix, and generate sampling data; calculate the probability distribution characteristics of each sampling point, extract the statistical moment, and generate statistical data; calculate the covariance matrix of the state estimation based on the statistical data and generate the covariance parameters.

[0146] S522. Read covariance parameters and statistical data, calculate the variance contribution of each measurement component, and generate variance data; decompose the systematic error and random error components of the measurement noise, evaluate the weight of each error source, and generate error data; calculate the sensitivity coefficient of the error propagation chain based on the error data and generate sensitivity parameters.

[0147] S523. Read sensitivity parameters, calculate the cumulative effect of uncertainty in the state transfer process, and generate cumulative data; analyze the error evolution law under different time scales, extract characteristic patterns, and generate evolution data; calculate the confidence interval boundaries based on the evolution data and generate boundary parameters.

[0148] S524, read boundary parameters and accumulated data, calculate the credibility index of the positioning result, and generate credible data; evaluate the spatiotemporal distribution characteristics of uncertainty, extract key influencing factors, and generate characteristic data; comprehensively calculate the final uncertainty index based on the characteristic data to generate uncertainty assessment results.

[0149] In one embodiment of the present application, the uncertainty assessment method is: comprehensive uncertainty U = k·sqrt[Σ(ci·ui) 2 + 2·Σ(ρij·ui·uj)]; where ui is the standard uncertainty of the ith error component; ci is the sensitivity coefficient; ρij is the correlation coefficient between the ith and jth error components; k is the coverage factor (confidence level is 95% when k=2); the error propagation equation σt 2 = σ0 2 + v·t·σr 2 + (a·t 2 )·σb 2 ; σ0 is the initial position error; σr is the random walk error; σb is the zero bias instability; t is the time; v is the velocity; a is the acceleration.

[0150] This embodiment implements systematic error analysis and quantification for the evaluation of positioning uncertainty. A complete state sampling matrix is ​​established by extracting the position coordinate sequence and the timestamp sequence. In the error analysis process, the weight contribution of each error source is accurately evaluated by decomposing the systematic error and random error components of the measurement noise. In particular, in the uncertainty propagation analysis, a reliable error prediction model is established by studying the error evolution law at different time scales. This embodiment not only quantifies the credibility of the positioning results, but also identifies the key influencing factors through error sensitivity analysis. By establishing a reasonable confidence interval, a reliable error range estimate is provided for the application of positioning results.

[0151] According to one aspect of the present application, step S53 is further:

[0152] S531. Read the optimized trajectory data, verification result data and uncertainty assessment results, extract the position coordinates and attitude angle data at each moment, calculate the statistical characteristics of the state quantity, and generate characteristic data; analyze the time correlation of the state quantity, calculate the autocorrelation coefficient, and generate correlation data; calculate the stability index of the state quantity based on the correlation data and generate a stable parameter.

[0153] S532, read the stable parameters and feature data, calculate the confidence index of the positioning result, and generate confidence data; evaluate the weight coefficients of various evaluation indicators, construct a comprehensive scoring function, and generate scoring data; calculate the level division threshold of the positioning quality based on the scoring data and generate the level parameter.

[0154] S533, read the level parameters and confidence data, calculate the spatiotemporal distribution characteristics of the positioning accuracy, and generate accuracy data; extract the accuracy mutation points at critical moments, evaluate the mutation causes, and generate mutation data; calculate the accuracy assurance probability based on the mutation data and generate probability parameters.

[0155] S534, read probability parameters and accuracy data, calculate the optimal estimate of three-dimensional coordinates, and generate coordinate data; evaluate the reliability index of the positioning result, extract quality features, and generate quality data; based on the comprehensive calculation of quality data and coordinate data, generate the final positioning result including three-dimensional coordinates, confidence and accuracy evaluation.

[0156] This embodiment achieves the final quality judgment of the positioning results through a comprehensive evaluation mechanism. First, a complete state evaluation model is established by analyzing the time correlation of the state quantity. In the quality rating process, a scientific comprehensive scoring system is established by evaluating the weight coefficients of various indicators. In particular, in the accuracy assurance analysis, reliable quality grade division is achieved by studying the spatiotemporal distribution characteristics and mutation characteristics of the positioning accuracy. This embodiment not only provides the precise three-dimensional coordinates of the positioning results, but also provides complete quality assurance information through confidence and accuracy evaluation. By establishing strict evaluation standards, the reliability and practicality of the output results are ensured, providing high-quality location reference data for defect detection in ultra-long water transfer tunnels.

[0157] In another embodiment of the present application, a method for underwater positioning of defects in an ultra-long water transfer tunnel specifically includes:

[0158] (1) A joint solution method for the underwater inertial navigation device timing positioning data and the underwater long baseline positioning device and ultra-short baseline positioning device timing positioning data is used. The update frequency of the underwater inertial navigation device timing positioning data is set to 10 Hz, and the update frequency of the ultra-short baseline positioning device timing positioning data is set to 0.1 Hz. The long baseline positioning device timing positioning data is used to solve the underwater positioning data when the distance to the transponder beacon carried by the underwater submersible is less than 1000 m and the visibility conditions are good.

[0159] (2) Optical image recognition method for highly reflective marking devices. When an underwater vehicle collects optical image data along the central axis of an ultra-long water diversion tunnel and finds a local highlight mark generated by a highly reflective marking device, the underwater vehicle approaches within 1 m to collect optical image data and recognizes tunnel mileage information based on the YOLOv8 basic network to form time-series positioning data.

[0160] (3) Multi-source underwater positioning data weight correction method. The positioning of underwater submersibles is completed by the underwater inertial navigation device timing positioning data, the ultra-short baseline positioning device timing positioning data, the long baseline positioning device timing positioning data, and the strong reflective marking device timing positioning data. The confidence weight priority of each data is from high to low: strong reflective marking device timing positioning data, long baseline positioning device timing positioning data, ultra-short baseline positioning device timing positioning data, and underwater inertial navigation device timing positioning data. When an underwater submersible uses high-priority positioning data, it should be compared with the underwater inertial navigation device timing positioning data to give quantitative error data to monitor the operating status of the underwater positioning device.

[0161] According to one aspect of the present application, the underwater positioning device for defects in ultra-long water transfer tunnels is mainly composed of an inertial navigation positioning device, a long baseline positioning device, an ultra-short baseline positioning device, and a strong reflective marking device. The inertial navigation positioning device is carried by an underwater submersible and installed at the bottom of the submersible. The device is composed of an attitude sensor, a Doppler bottom tracking current meter, an altimeter, and a sonic velocity meter. Among them, the attitude sensor and the Doppler bottom tracking current meter need to be rigidly fixed and installed, and the altimeter and sonic velocity meter system are integrated to realize the real-time positioning function of the underwater submersible in the process of being able to identify the bottom topography. The long baseline positioning device and the ultra-short baseline positioning device, the transponder beacon is carried by the underwater submersible and installed on the top of the submersible. The device consists of an underwater long baseline positioning device and a surface ultra-short baseline positioning device. The long baseline positioning device adopts a floating beacon array and is placed at the water inlet of the long water transfer tunnel and along the tunnel. The buoy at the water inlet introduces the GNSS satellite positioning signal into the positioning system; the ultra-short baseline positioning device adopts an unmanned ship or an autonomous underwater submersible and is placed along the water transfer tunnel, such as the water surface in the ventilation shaft. The strong reflective identification device uses the emptying process of the water transfer tunnel to set a stainless steel identification plate with a length of 42cm and a width of 32cm. The surface is marked with characteristic text, characteristic points, and characteristic lines by using fluorescent agents.

[0162] According to one aspect of the present application, an underwater defect positioning device for an ultra-long water conveyance tunnel comprises:

[0163] at least one processor; and,

[0164] a memory communicatively connected to at least one of the processors; wherein,

[0165] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the underwater positioning method for defects in an ultra-long water transfer tunnel described in any of the above embodiments.

[0166] The present invention comprehensively solves the technical difficulties of underwater positioning in ultra-long water transfer tunnels through multi-source data fusion and multi-level optimization strategies. At the data acquisition level, the inherent limitations of a single sensor are overcome through a multi-sensor collaborative working mechanism; at the environmental modeling level, dynamic acoustic models and identification recognition are used to adapt to complex underwater environments; at the data fusion level, the optimal combination of various types of data is achieved through adaptive weight distribution; at the trajectory optimization level, the accuracy of the positioning trajectory is ensured through multiple constraint processing; at the result verification level, the reliability of the positioning results is guaranteed through uncertainty assessment. The present invention not only achieves sub-meter positioning accuracy in ultra-long water transfer tunnels, but also has strong environmental adaptability and stability, providing reliable technical support for tunnel defect detection and maintenance.

[0167] The preferred embodiments of the present invention are described in detail above; however, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. A method for underwater positioning of defects in an ultra-long water transfer tunnel, characterized in that: The steps include: S1. Acquire multi-source sensor data of the underwater vehicle, perform time synchronization processing on the multi-source sensor data, and generate a synchronized data stream; perform signal preprocessing on the synchronized data stream, and generate a preprocessed data stream; The multi-source sensor data includes inertial navigation data, acoustic positioning data, environmental sensor data and strong reflective sign image data; S2. Based on the preprocessed data stream, build an acoustic environment feature model, identify strong reflective sign information, update environmental parameters, and generate environmental model parameters and sign point location data; S3. Based on the preprocessed data stream, environmental model parameters and identification point location data, the dynamic weight of each data source is calculated, and data consistency check is performed to obtain consistency score data; Based on the consistency score data, perform multi-source data fusion operations to generate fusion positioning results; S4, based on the fusion positioning result, perform trajectory optimization calculation to obtain the initial optimized trajectory; Based on the initial optimization trajectory and the location data of the marker points, the marker point constraint processing is performed to generate the constraint optimization result; Based on the constraint optimization results, the historical trajectory is backtracked and corrected to generate optimized trajectory data; S5. Based on the optimized trajectory data and the fused positioning results, multiple verification operations are performed to conduct uncertainty assessment and generate the final positioning result. Step S2 is further as follows: S21, constructing an acoustic propagation model based on the environmental sensor data in the preprocessed data stream, and generating acoustic propagation model parameters, i.e., environmental model parameters; S22, based on the strong reflective mark image data in the preprocessed data stream, image feature extraction and mark position calculation are performed to generate mark point position data; S23, based on the acoustic propagation model parameters and the environmental sensor data in the preprocessed data stream, construct an environmental parameter state transfer relationship and generate an environmental state vector; Step S3 is further as follows: S31, based on the preprocessed data stream, environmental model parameters, identification point location data and environmental state vector, evaluate the quality index and reliability index of each data source; based on the quality index and reliability index, calculate the dynamic weight coefficient and generate a data weight vector; S32, based on the preprocessed data stream and the data weight vector, performing consistency evaluation operation on the multi-source data to generate consistency score data; S33, performing adaptive data fusion operation based on the preprocessed data stream, the data weight vector and the consistency score data to generate a fusion positioning result; Step S4 is further as follows: S41, based on the fusion positioning results, construct the pose nodes and constraint edge relationships, perform trajectory smoothing optimization operations, and generate an initial optimized trajectory; S42, based on the position data of the identification points, using a constrained optimization algorithm to perform constraint processing on the initial optimization trajectory to generate a constrained optimization result; S43, based on the constraint optimization result, perform a backtracking calculation of historical data within a preset time window to generate the historical data after the backtracking calculation; perform trajectory correction processing on the historical data after the backtracking calculation to generate optimized trajectory data; Step S5 is further as follows: S51, based on the optimized trajectory data and the fused positioning result, perform position consistency calculation, trajectory smoothness calculation and positioning accuracy calculation to generate verification result data; S52, calculating uncertainty parameters of the positioning result based on the optimized trajectory data and the verification result data, and generating uncertainty assessment data; S53, based on the optimized trajectory data, verification result data and uncertainty assessment data, a comprehensive evaluation calculation is performed to generate a final positioning result including three-dimensional coordinates, confidence and accuracy assessment.

2. The method for underwater positioning of defects in ultra-long water transfer tunnels according to claim 1, characterized in that: Step S1 is further as follows: S11, using a sensor collector to respectively acquire inertial navigation data, acoustic positioning data, environmental sensor data and strong reflective sign image data of the underwater submersible, and combine them to form multi-source sensor data; wherein the inertial navigation data includes acceleration data and angular velocity data, the acoustic positioning data includes ultra-short baseline positioning data and long baseline positioning data, the environmental sensor data includes water temperature data, pressure data and sound speed data, and the strong reflective sign image data includes image data of the sign plate; S12, based on multi-source sensor data, perform time alignment operations on data with different sampling frequencies, build a unified time base, and generate a synchronous data stream; S13. Perform outlier detection and data noise reduction processing on the synchronous data stream to generate a preprocessed data stream.

3. The method for underwater positioning of defects in an ultra-long water transfer tunnel according to claim 2 is characterized in that: Step S11 is further as follows: S111, collecting inertial navigation data from an inertial navigation sensor, compensating and correcting the inertial navigation data through a dynamic calibration coefficient to generate a compensation signal; performing zero bias estimation and temperature drift correction on the compensation signal to generate a correction signal; based on the correction signal, calculating static and dynamic response characteristics, judging data validity, and generating effective inertial navigation data; S112, receiving acoustic positioning data from the acoustic positioning system, calculating the signal arrival time difference and the sound wave propagation path, and generating path data; performing multipath component decomposition on the path data, identifying direct wave and reflected wave components, and generating decomposition data; Based on the decomposed data, a sound wave propagation probability model is constructed to screen the optimal propagation path and generate the optimized path; Based on the optimized path, signal strength calibration and phase compensation are performed to generate effective acoustic data; S113, collecting environmental sensor data from environmental sensors, building an environmental parameter spatial distribution model, and generating distribution data; Perform gradient calculation and outlier identification on the distribution data to generate verification data; Based on the verification data, calculate the parameter change trend, judge the rationality of the data, and generate effective environmental data; S114, acquiring image data of a strong reflective sign from an image acquisition system, performing adaptive histogram equalization processing, and generating an enhanced image; Calculate edge strength and evaluate contrast of the enhanced image to generate feature data; Based on the feature data, the image clarity index and the illumination uniformity index are calculated to generate evaluation data; Based on the evaluation data, an image quality scoring model is constructed to screen valid images and generate valid image data; S115, constructing a space-time correspondence matrix based on the effective inertial navigation data, the effective acoustic data, the effective environmental data and the effective image data, and generating corresponding data; Perform data integrity verification and time consistency check on the corresponding data to generate verification data; Based on the verification data, a multi-source data association index is constructed to generate multi-source sensor data.

4. The method for underwater positioning of defects in an ultra-long water transfer tunnel according to claim 2 is characterized in that: Step S21 is further as follows: S211, based on the environmental sensor data in the preprocessed data stream, dividing the three-dimensional space into uniform grid units, calculating the center coordinates of each grid unit, and generating grid coordinate data; Based on the environmental sensor data of each grid unit, the corresponding sound speed value is calculated to generate sound speed distribution data; based on the grid coordinate data and the sound speed distribution data, a spatial interpolation operation is performed to generate node data; Based on the node data, the sound velocity change rate between adjacent grid points is calculated to generate gradient field data; S212, based on the gradient field data, calculating the rate of change of the speed of sound in each direction, extracting the curvature characteristics of the sound wave propagation path, and generating curvature data; Based on the curvature data, the energy attenuation coefficient of the sound wave during propagation is calculated to generate attenuation data; Based on the attenuation data, the energy loss at different propagation distances is calculated to generate energy distribution; S213, calculating the reflection angle of the sound wave at different medium interfaces based on the gradient field data and energy distribution, and generating reflection data; Based on the reflection data, the multipath propagation path is analyzed, the delay difference of each path is calculated, and the propagation path is generated; Based on the propagation path, the interface scattering intensity is calculated to generate scattering data; Based on the scattering data and energy distribution, the sound field intensity distribution is calculated and the prediction results are generated; S214, based on the prediction results, calculating the time-varying characteristics of the sound field parameters to generate time-varying data; Based on time-varying data, extract the law of environmental changes, calculate the parameter update cycle, and generate periodic data; Based on the periodic data, a parameter update threshold is constructed to generate update parameters; based on the update parameters and the prediction results, the sound field characteristics are calculated to generate acoustic propagation model parameters, namely, environmental model parameters.

5. The method for underwater positioning of defects in an ultra-long water transfer tunnel according to claim 2, characterized in that: Step S32 is further as follows: S321, based on the preprocessed data stream and the data weight vector, calculating the distribution characteristics of each data source in the spatial domain, and generating spatial data; Based on spatial data, the distance matrix and correlation coefficient between data points are calculated to generate relevant data; Based on relevant data, identify spatial anomalies and conflict points, calculate conflict indicators, and generate conflict parameters; S322, calculating the autocorrelation function and the cross-correlation function of the data sequence based on the conflict parameter and the related data, and generating sequence data; Based on the sequence data, extract the changing rules of the time dimension, calculate the trend characteristics, and generate trend data; Based on trend data, calculate the temporal coordination index between data and generate coordination parameters; S323. Based on the coordination parameters, calculate the information entropy and conditional entropy between the data sources to generate entropy value data; Based on entropy data, analyze information redundancy and complementarity, calculate information gain, and generate gain data; Based on the gain data, the consistency quantitative index is calculated to generate the quantitative parameters; S324, calculating the comprehensive score of the multi-dimensional evaluation index based on the quantitative parameters to generate comprehensive data; Based on the comprehensive data, multi-level fuzzy evaluation is carried out to calculate the membership value and generate membership data; Based on the affiliation data, the final consistency evaluation result is calculated to generate consistency score data.

6. Ultra-long water tunnel defect underwater positioning device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the underwater positioning method for defects in an ultra-long water transfer tunnel as described in any one of claims 1 to 5.

Images