Cable positioning method, system, equipment and medium

Abstract

The invention relates to the technical field of cable positioning, and discloses a cable positioning method, system and device and a medium, and the method comprises the steps: collecting a magnetic field signal excited by a target cable; based on an information geometry theory, constructing a parameterized probability density model according to the magnetic field signal and a preset signal parameter space so as to determine a three-dimensional statistical manifold, and further generating a Fisher information matrix; projecting the three-dimensional statistical manifold by adopting an alternating optimization strategy to obtain a two-dimensional sub-manifold and a corresponding Fisher information matrix after projection; constructing a log-likelihood function according to the signal parameter space and the parameterized probability density model, and iteratively solving the log-likelihood function by adopting a natural gradient descent method based on the Fisher information matrix after projection to obtain an amplitude estimation result of the magnetic field signal so as to determine the position of the target cable; according to the method, the information geometry theory, the alternate optimization strategy and the natural gradient descent method are combined, and efficient and accurate positioning of the cable position is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cable positioning, in particular to a cable positioning method, system, device and medium. BACKGROUND

[0002] In the field of underground cable positioning, accurate detection of cable position highly depends on the accurate extraction of the amplitude of the alternating magnetic field signal around it. At present, the extraction of signal amplitude in strong noise background mostly adopts phase-locked amplification technology, which has certain noise suppression ability, but its essence is a filtering method based on the correlation between the reference signal and the input signal, which has inherent limitations: first, its optimization criterion is to minimize the mean square error, which is not optimal under non-Gaussian noise; second, its parameters are fixed and difficult to adapt to complex and changing noise statistical characteristics; third, the amplitude, phase and other parameters are estimated as independent variables in Euclidean space, ignoring the internal statistical coupling relationship between parameters, when the noise statistical characteristics are unknown or change, the estimation accuracy decreases rapidly, which leads to inaccurate cable positioning.

[0003] Therefore, how to provide an accurate amplitude extraction scheme for magnetic field signals in cable path detection to solve the problem of decreased cable positioning accuracy caused by inaccurate signal amplitude estimation in complex noise environment has become a technical problem to be solved by those skilled in the art. SUMMARY

[0004] The present application provides a cable positioning method, system, device and medium to solve the problem of decreased cable positioning accuracy caused by inaccurate signal amplitude estimation in complex noise environment.

[0005] To solve the above technical problems, the present application provides a cable positioning method in the first aspect, comprising: Collecting the magnetic field signal excited by the target cable; Based on information geometry theory, constructing a parameterized probability density model according to the magnetic field signal and a preset signal parameter space to determine a three-dimensional statistical manifold, and generating a Fisher information matrix according to the three-dimensional statistical manifold; Projecting the three-dimensional statistical manifold using an alternating optimization strategy to obtain a two-dimensional submanifold to update the Fisher information matrix and generate a projected Fisher information matrix; Constructing a log-likelihood function according to the signal parameter space and the parameterized probability density model, and based on the projected Fisher information matrix, iteratively solving the log-likelihood function using a natural gradient descent method to obtain an amplitude estimation result of the magnetic field signal to determine the position of the target cable.

[0006] As one of the preferred solutions, the parameterized probability density model is constructed according to the magnetic field signal and a preset signal parameter space, including: The magnetic field signal is normalized and discretized to obtain a normalized discrete signal sequence; Based on the signal parameter space, the product of the noise probability of each point in the normalized discrete signal sequence is quantified to construct the parameterized probability density model.

[0007] As one of the preferred solutions, the three-dimensional statistical manifold is determined, and the Fisher information matrix is generated according to the three-dimensional statistical manifold, including: Based on the parameterized probability density model, the probability distribution corresponding to each parameter combination determined according to the signal parameter space is taken as a point on the statistical manifold to construct the three-dimensional statistical manifold; The Fisher information matrix is determined according to the Riemannian geometric structure of the three-dimensional statistical manifold.

[0008] As one of the preferred solutions, the three-dimensional statistical manifold is projected to obtain a two-dimensional submanifold to update the Fisher information matrix to generate a projected Fisher information matrix by using an alternating optimization strategy, including: The frequency of the normalized discrete signal sequence is estimated by using a phase difference method to obtain an instantaneous angular frequency and determine a frequency drift amount in combination with the signal parameter space; Based on the frequency drift amount, the signal parameter space is reduced in dimension to obtain a reduced dimension parameter, and the three-dimensional statistical manifold is projected by using the reduced dimension parameter to obtain a two-dimensional submanifold; The Fisher information matrix is updated according to the two-dimensional submanifold to obtain the projected Fisher information matrix.

[0009] As one of the preferred solutions, the log-likelihood function is constructed according to the signal parameter space and the parameterized probability density model, and the log-likelihood function is iteratively solved by using a natural gradient descent method based on the projected Fisher information matrix to obtain an amplitude estimation result of the magnetic field signal, including: The reduced dimension parameter is initialized to obtain an initial parameter, and the parameterized probability density model is taken as a logarithm based on the initial parameter to obtain the log-likelihood function; Based on the initial parameter, the log-likelihood function is derived to obtain an ordinary gradient, which is combined with the projected Fisher information matrix to quantitatively obtain a natural gradient; updating the initial parameters along the direction of the natural gradient, iteratively performing the quantization step of the natural gradient according to the updated initial parameters until an iteration termination condition is reached, and generating an amplitude estimation result of the magnetic field signal based on the initial parameters at the final iteration.

[0010] As one of the preferred solutions, the generating of the amplitude estimation result of the magnetic field signal based on the initial parameters at the final iteration comprises: taking the initial parameters at the final iteration as a maximum likelihood estimation point on the two-dimensional submanifold, and determining a primary amplitude estimation result based on the maximum likelihood estimation point; performing adaptive adjustment processing on the frequency drift amount to obtain an adaptive frequency drift amount to update the projected Fisher information matrix and the log-likelihood function to obtain an updated Fisher information matrix and an updated log-likelihood function; re-executing the amplitude estimation step of the magnetic field signal based on the updated Fisher information matrix and the updated log-likelihood function to obtain a secondary amplitude calculation result; iteratively performing the adaptive processing process on the frequency drift amount to iteratively perform the amplitude estimation step of the magnetic field signal based on the re-obtained adaptive frequency drift amount until a preset iteration number is reached to obtain a plurality of amplitude calculation results; fusing the primary amplitude estimation result, the secondary amplitude calculation result and the plurality of amplitude calculation results to obtain an optimal amplitude estimation result of the magnetic field signal.

[0011] As one of the preferred solutions, the determining of the position of the target cable comprises: constructing a magnetic field depth relationship model of the target cable based on electromagnetic field theory; inputting the optimal amplitude estimation result into the magnetic field depth relationship model for processing to obtain a preliminary depth of the target cable; performing error correction on the preliminary depth, and determining the position of the target cable based on the corrected preliminary depth.

[0012] The second aspect of the application provides a cable positioning system, comprising: a data acquisition module configured to acquire a magnetic field signal excited by a target cable; a matrix generation module configured to construct a parameterized probability density model based on information geometry theory according to the magnetic field signal and a preset signal parameter space to determine a three-dimensional statistical manifold, and generate a Fisher information matrix according to the three-dimensional statistical manifold; A manifold projection module is configured to project the three-dimensional statistical manifold to obtain a two-dimensional sub-manifold by using an alternating optimization strategy, update the Fisher information matrix, and generate a projected Fisher information matrix; A position determination module is configured to construct a log-likelihood function according to the signal parameter space and the parameterized probability density model, and iteratively solve the log-likelihood function by using a natural gradient descent method based on the projected Fisher information matrix to obtain an amplitude estimation result of the magnetic field signal and determine the position of the target cable.

[0013] The third aspect of the present application provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the cable positioning method as described above when executing the computer program.

[0014] The fourth aspect of the present application provides a computer-readable storage medium, comprising a stored computer program, wherein the device where the computer-readable storage medium is located implements the cable positioning method as described above when executing the computer program.

[0015] Compared with the prior art, the beneficial effects of the embodiments of the present application are as follows: Based on the information geometry theory, the three-dimensional statistical manifold is constructed to map the magnetic field signal to a high-dimensional geometric space, which can capture the statistical characteristics of the signal, avoid the strong assumption of the signal form in the traditional method, and improve the adaptability of the model to the complex electromagnetic environment. By using the alternating optimization strategy, the three-dimensional statistical manifold is projected into a two-dimensional sub-manifold, which significantly reduces the computational complexity, retains the key geometric information, reduces the redundant dimensions, makes the update of the Fisher information matrix more efficient, and focuses on the core change direction of the signal parameter after the projection of the Fisher information matrix, highlights the geometric features sensitive to the amplitude estimation, and improves the pertinence of the subsequent gradient calculation. The natural gradient descent method used in the present application adjusts the gradient direction based on the geometric structure of the statistical manifold, overcomes the deviation problem of the traditional Euclidean gradient in the nonlinear space, ensures that each iteration is optimized along the steepest descent direction, accelerates the convergence, and avoids local optimization. Through the joint optimization of the log-likelihood function and the natural gradient descent, the magnetic field amplitude and the cable position are further associated, the error accumulation of the intermediate variable transmission is reduced, and the positioning accuracy is improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the present application, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0017] Figure 1 is a flow chart of a cable positioning method provided by some embodiments of the present application; Figure 2 is a structural diagram of a cable positioning system provided by some embodiments of the present application; Figure 3 is a structural diagram of an electronic device provided by some embodiments of the present application; Reference signs: Wherein, 10, a data acquisition module; 20, a matrix generation module; 30, a manifold projection module; 40, a position determination module; 5000, an electronic device; 5001, a processor; 5002, a bus; 5003, a memory; 5004, a transceiver. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings and embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments of the present application. The purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0019] In the description of the present application, the terms "first", "second", "third" and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second", "third" and the like can explicitly or implicitly include one or more of the features. In the description of the present application, unless otherwise specified, the meaning of "multiple" is two or more. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0020] In the description of the present application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present application have the same meaning as generally understood by those skilled in the art. The terms used in the description of the present application are only used to describe specific embodiments, and are not intended to limit the present application. For those of ordinary skill in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0021] In an embodiment, as shown in Figure 1 The first aspect of the present application provides a cable positioning method, comprising: S1. Acquire the magnetic field signal excited by the target cable; specifically, in cable path detection, inject a frequency of [frequency missing] into the target cable. (angular frequency) ), amplitude stabilized sinusoidal current signal According to the principle of electromagnetic induction, this current will induce an alternating magnetic field of the same frequency in the space surrounding the target cable. This invention uses a positioning antenna to receive the vertical component of the magnetic field directly above the target cable as the magnetic field signal. Considering the frequency drift of ±0.1% to ±1% when injecting current into the target cable under actual operating conditions, its signal model... Represented as: In the formula, This refers to the amplitude of the magnetic field strength, which is directly related to the cable burial depth. , Angular frequency, The nominal frequency of the injected current is typically 50-60Hz. For the frequency drift, satisfying ; For time; This is the initial phase.

[0022] The amplified expression of the antenna received signal is: In the formula, It is a magnetic field signal; The sensor system gain is obtained through factory calibration or field calibration. The system and environmental noise have unknown but bounded statistical characteristics.

[0023] S2. Based on information geometry theory, a parameterized probability density model is constructed according to the magnetic field signal and the preset signal parameter space to determine the three-dimensional statistical manifold, and a Fisher information matrix is ​​generated according to the three-dimensional statistical manifold. The present application establishes a parameterized probability model containing amplitude, phase and noise statistical characteristics for the sinusoidal magnetic field signal generated by the cable, and uses information geometry theory to regard the probability distribution corresponding to each possible parameter combination as a point, and the set of all these points constitutes a statistical manifold which has a Riemannian geometric structure defined by the Fisher information matrix and can accurately reflect the difficulty of parameter estimation and the coupling relationship between parameters, and then converts the signal amplitude extraction problem into an optimization problem on the statistical manifold. The Riemannian manifold method based on information geometry used in the present application can essentially consider the statistical coupling relationship between parameters through the geometric structure of the manifold, and has strong robustness to model mismatch (including the uncertainty of noise distribution). When the noise statistical characteristics are known, the method can approach the Cramer-Rao lower bound; even if the noise characteristics are unknown, the geometric structure can provide better estimation performance than the traditional Euclidean space method, so that the amplitude extraction accuracy far exceeds that of the traditional method at very low signal-to-noise ratio, thereby laying a solid foundation for subsequent accurate cable positioning based on the magnetic field decay law.

[0024] In an embodiment, the parameterized probability density model is constructed according to the magnetic field signal and a preset signal parameter space, comprising: The magnetic field signal is normalized and discretized to obtain a normalized discrete signal sequence; Based on the signal parameter space, the product of the noise probability of each point in the normalized discrete signal sequence is quantized to construct the parameterized probability density model.

[0025] Specifically, to achieve accurate amplitude extraction, the observed magnetic field signal is divided by a gain to achieve normalization of the magnetic field signal and obtain a normalized signal, which is as follows: In the formula, is the normalized signal; is the normalized noise.

[0026] The continuous signal is discretized at a sampling rate (typically 1000 Hz) to obtain an observation sequence of points is the signal length, typically 1024, balancing accuracy and computational load. The discrete signal model is: In the formula, is the sampling period, .

[0027] To achieve accurate amplitude extraction based on information geometry, a parameterized probability model of the signal needs to be established. It is assumed that in a given signal parameter space Under these conditions, normalized observation signal data The probability density function is derived from noise. Since the distribution is determined by the signal parameter space, the product of the noise probabilities at each point in the normalized discrete signal sequence can be used as the parameterized probability density model. It is expressed by the following formula: In the formula, Let be the probability density function of the noise, initially set to a Gaussian distribution (for non-Gaussian scenarios, it can be replaced with a corresponding distribution, such as a Laplace distribution, etc.). For noise residuals, i.e. The observed values.

[0028] This invention unifies the signal scale and simplifies calculations through normalization and discretization preprocessing, and constructs a probability density model with clear physical meaning through parameterized noise probability quantization. Ultimately, it achieves an accurate description of the magnetic field signal and explicit modeling of noise, providing a highly reliable and adaptable basic model for subsequent manifold optimization and gradient descent, and significantly improving the accuracy and robustness of the cable positioning method in complex environments.

[0029] In one embodiment, determining a three-dimensional statistical manifold and generating a Fisher information matrix based on the three-dimensional statistical manifold includes: Based on the parameterized probability density model, the probability distribution corresponding to each parameter combination determined according to the signal parameter space is used as a point on the statistical manifold to construct the three-dimensional statistical manifold. The Fisher information matrix is ​​determined based on the Riemannian geometry of the three-dimensional statistical manifold.

[0030] Specifically, this invention is based on a parameterized probability density model, which considers each possible combination of parameters... The corresponding probability distribution Treat as a statistical manifold A point on the graph, and the set of all such points constitutes a three-dimensional statistical manifold (parameter). (where the coordinates are local coordinates of the manifold), which has a Riemannian geometric structure defined by the Fisher information matrix to characterize the difficulty of parameter estimation and coupling relationships; that is, the Fisher information matrix of a three-dimensional statistical manifold. This can be expressed by the following formula: In the formula, Parameters corresponding to the signal parameter space ; is the mathematical expectation (approximated by sample mean in practice). In this formula, the diagonal elements respectively reflect the estimation difficulty (the larger the value, the higher the theoretical estimation accuracy of the parameter) of ; the non-diagonal elements reflect the statistical coupling relationship between parameters (such as is not zero, indicating and are coupled, the change will affect the estimation of ).

[0031] The present application constructs a three-dimensional statistical manifold by taking the probability distribution corresponding to each parameter combination in the signal parameter space as a point on the statistical manifold, comprehensively and meticulously characterizes the features of the magnetic field signal from multiple dimensions, integrates various parameter information of the signal and their mutual relationship into the manifold structure, and can more completely reflect the essential characteristics of the signal compared with the traditional single-dimensional analysis method. The Fisher information matrix is determined according to the Riemannian geometry structure of the three-dimensional statistical manifold, which can accurately quantify the information content contained in different parameters in the signal, and provide optimal direction and step information for parameter estimation, improve the efficiency and accuracy of parameter estimation, reduce estimation error, and make the estimation result closer to the true value.

[0032] S3, projecting the three-dimensional statistical manifold by using an alternating optimization strategy to obtain a two-dimensional submanifold to update the Fisher information matrix and generate a post-projection Fisher information matrix; In an embodiment, step S3 comprises: using a phase difference method to estimate the frequency of the normalized discrete signal sequence to obtain an instantaneous angular frequency and combining the signal parameter space to determine a frequency drift amount; performing dimensionality reduction on the signal parameter space based on the frequency drift amount to obtain a reduced dimension parameter, and projecting the three-dimensional statistical manifold by using the reduced dimension parameter to obtain a two-dimensional submanifold; updating the Fisher information matrix according to the two-dimensional submanifold to obtain the post-projection Fisher information matrix.

[0033] Specifically, the present application takes into account that the frequency drift range is limited in the cable positioning scene and the main target is to accurately extract the amplitude, and therefore uses an alternating optimization strategy to process the frequency drift, that is, the secondary parameter frequency drift in the signal parameter space is fixed first, and then the optimization of the core parameters amplitude and phase in the signal parameter space is focused on, so as to project the three-dimensional statistical manifold into a two-dimensional submanifold.

[0034] Frequency drift is a small frequency deviation of the injected current in a cable. Its range of variation is limited, and its impact on the estimation accuracy of amplitude is far less than the statistical coupling relationship between amplitude and phase. Directly optimizing the three-dimensional parameters in the signal parameter space would lead to a significant increase in computational complexity due to the high dimensionality, making it unsuitable for the real-time requirements of on-site positioning. Frequency drift estimation needs to meet both sufficient accuracy and computational efficiency. In each iteration, this invention estimates the frequency drift based on the current normalized discrete signal sequence using the phase difference method (or short-time Fourier transform, with the phase difference method preferred): First, the normalized discrete signal sequence is subjected to Hilbert transformation... The transform is used to obtain a complex analytic signal, and the instantaneous phase reflecting the frequency change is extracted from the complex analytic signal. The essence of frequency drift is the deviation between the instantaneous frequency and the nominal frequency. Since the instantaneous frequency is proportional to the rate of change of the phase over time, the phase difference between adjacent points is quantified based on the calculated instantaneous phase. At the same time, since the phase difference may have a 2π jump (periodicity of the phase function), the phase difference between adjacent points needs to be unwrapped (such as correcting the jump point through linear interpolation). The corrected phase difference is obtained and its average value is calculated to obtain the instantaneous angular frequency. The difference between the instantaneous angular frequency and the nominal angular frequency is then used as the frequency drift.

[0035] In the signal parameter space By fixing the frequency drift to a known parameter—the calculated frequency shift—the dimensionality of the signal parameter space can be reduced, transforming it into a dimensionality-reduced parameter space. With these dimensionality reduction parameters, a three-dimensional statistical manifold can be projected to obtain a two-dimensional submanifold, where each point on the manifold corresponds to a set of dimensionality reduction parameters and their corresponding probability distribution. ( (For fixed known quantities).

[0036] By updating the Fisher information matrix based on the two-dimensional submanifold, the projected Fisher information matrix can be obtained, which is equivalent to eliminating... right , The coupling effect is described only. and The two-dimensional Fisher information matrix of statistical relationships requires the derivation of the Fisher information matrix to eliminate... Zhongyu Related coupling terms ( , , , , ), resulting in only a description and The two-dimensional Fisher information matrix of the relationship, i.e., the projected Fisher information matrix. It is expressed by the following formula: wherein, correspond to the parameters , ; is the position element of the Fisher information matrix , to reflect the estimated accuracy weight of itself.

[0037] The geodesic distance between any two points and on the two-dimensional submanifold is: wherein, is the path connecting and ; , is the tangent vector and radial vector component of the path.

[0038] In the actual system, the geodesic distance is calculated by using the discretization approximation: wherein, is the midpoint of the two points; is the Fisher information metric at the midpoint.

[0039] Based on this, the amplitude extraction problem of the magnetic field signal is converted into finding the maximum likelihood estimation point on the two-dimensional submanifold of the statistical manifold: wherein, is the maximum likelihood estimation point.

[0040] In addition, it should be noted that the alternating iteration alternating optimization adopted in the present application is not fixed and projected at one time, but is embedded into the subsequent amplitude extraction iteration process, and the two-dimensional submanifold is re-estimated and updated before each natural gradient optimization, to ensure that the dynamic change of does not affect the estimation accuracy of the amplitude, and the iteration triggering time is that in the amplitude extraction step, each iteration optimization and needs to re-execute the estimation step and the manifold projection step, and then through the estimation, manifold projection, / ​​The optimized cyclical alternation ensures both the computational efficiency of 2D optimization and dynamic adaptation. These subtle changes prevent accuracy loss due to model mismatch, ultimately enabling stable extraction even in frequency-drift scenarios. The goal is to simplify the problem's complexity through alternating optimization without losing the coupling relationships of core parameters. This solves both the real-time performance problem of traditional 3D optimization and the accuracy loss problem of traditional Euclidean dimensionality reduction, thus paving the way for subsequent natural gradient extraction. Precision laid the foundation for geometry.

[0041] This invention employs the phase difference method to estimate the frequency of a normalized discrete signal sequence, enabling more accurate acquisition of the instantaneous angular frequency of the signal and effectively capturing subtle changes in signal frequency, thus reflecting the dynamic characteristics of the signal more accurately. Dimensionality reduction of the signal parameter space based on frequency drift yields dimensionality-reduced parameters, effectively reducing the number of parameters to consider. Projecting the dimensionality-reduced parameters onto a three-dimensional statistical manifold yields a two-dimensional submanifold, simplifying the data while preserving key information in the signal to the greatest extent possible. The updating of the Fisher information matrix accurately reflects the changes in the signal's information structure after dimensionality reduction, allowing for a more reasonable assessment of the reliability and accuracy of parameters in the low-dimensional space.

[0042] S4. Construct a log-likelihood function based on the signal parameter space and the parameterized probability density model, and iteratively solve the log-likelihood function using the natural gradient descent method based on the projected Fisher information matrix to obtain the amplitude estimation result of the magnetic field signal and determine the position of the target cable. In one embodiment, the step of constructing a log-likelihood function based on the signal parameter space and the parameterized probability density model, and iteratively solving the log-likelihood function using the natural gradient descent method based on the projected Fisher information matrix to obtain the amplitude estimation result of the magnetic field signal includes: The dimensionality reduction parameters are initialized to obtain initial parameters, and based on the initial parameters, the logarithm of the parameterized probability density model is taken to obtain the log-likelihood function. Based on the initial parameters, the derivative of the log-likelihood function is obtained to get the ordinary gradient, which is then combined with the projected Fisher information matrix to quantize and obtain the natural gradient. The initial parameters are updated along the direction of the natural gradient. The quantization steps of the natural gradient are iterated according to the updated initial parameters until the iteration termination condition is met. The amplitude estimation result of the magnetic field signal is generated based on the initial parameters at the time of the final iteration.

[0043] Specifically, after the two-dimensional sub-manifold is constructed, an efficient optimization algorithm needs to be designed to find the maximum likelihood estimation point on the manifold to realize the accurate estimation of the weak signal amplitude in the noise background. The amplitude optimal extraction algorithm based on Riemann distance is proposed, which fully utilizes the Riemann geometric structure of the manifold, and is particularly optimized for the ± 0.1%-± 1% frequency drift problem existing in the cable positioning scene.

[0044] The point discrete magnetic field signal sequence collected on the target cable site is regarded as a sample from the parameterized probability density model , wherein is a point on the manifold, and then the amplitude optimal extraction problem is converted into solving the maximum likelihood estimation on the statistical manifold : In order to efficiently solve the optimization problem, the natural gradient descent method is used to solve the problem, which utilizes the Riemann geometric structure of the manifold, and the update direction is the fastest direction of the likelihood function growth on the manifold (i.e. the steepest ascent direction), thereby overcoming the low efficiency problem of the traditional gradient method in the curved parameter space; the core of the method is that on the manifold with Riemann metric, the gradient direction of the ordinary Euclidean space is not the steepest ascent direction of the objective function. The true steepest ascent direction is given by the natural gradient, which is defined as the product of the ordinary gradient and the inverse of the Fisher information matrix (i.e. the Riemann metric tensor): In the formula, is the projection of the Fisher information matrix after projection on the two-dimensional sub-manifold ; is the ordinary gradient; is the natural gradient, which is also the steepest ascent direction under the constraint of the local distance (defined by the Riemann distance) on the manifold. Therefore, the optimization process is equivalent to searching along the geodesic direction on the Riemann manifold, thereby essentially considering the statistical coupling relationship between parameters.

[0045] Then, the specific implementation steps of the amplitude optimal extraction algorithm based on Riemann distance are as follows: Initialize the dimension reduction parameters, i.e. determine the initial frequency based on the FFT frequency estimation, set the initial amplitude , set the initial phase (the phase deviation will be quickly corrected in the subsequent iteration), to form the initial parameter vector (corresponding to the starting point on the two-dimensional sub-manifold): and based on the initial parameter vector, take the logarithm of the parameterized probability density model to obtain the log-likelihood function , which is used to quantify the matching degree of the current parameters and the observed signal, which is represented by the following formula: where, is the noise variance, which can be estimated from the background noise segment.

[0046] Then the log-likelihood function of the observation data and its ordinary gradient are calculated at the current parameter point , that is, the partial derivatives of the log-likelihood function with respect to the amplitude and phase in are taken to obtain the ordinary gradient reflecting the rate of change of the likelihood function in the Euclidean space; subsequently, the optimization direction (through the natural gradient) is guided by the geometry of the Riemannian manifold, rather than relying entirely on specific noise distribution assumptions. This is because the manifold structure itself has certain inclusiveness for changes in noise distribution. For a known non-Gaussian noise, the noise distribution model can be replaced and the Fisher information matrix is recalculated; that is, the natural gradient is calculated using the Fisher information matrix (i.e., the Riemannian metric) at the current parameter point , that is, the ordinary gradient is converted into the natural gradient on the manifold by projecting the inverse matrix of the Fisher information matrix after the projection: where, is the inverse matrix of the projected Fisher information matrix at the current parameter .

[0047] The parameter is updated along the natural gradient direction (the steepest ascent direction on the manifold, i.e., the direction in which the likelihood function grows fastest under the constraint of the Riemannian distance), and the step size adopts an adaptive strategy (to avoid oscillation or slow convergence), that is: where, is the updated initial parameter; is the adaptive step size (initially 0.01, halved every 10 iterations); if higher precision is required, an exponential mapping operation can be replaced to ensure that the updated strictly falls on the two-dimensional sub-manifold (to avoid model mismatch caused by the parameter point deviating from the manifold).

[0048] Since the parameter space can be locally considered as a Euclidean space, and the step size is small, the above Euclidean update can approximate the geodesic movement on the manifold, and for higher precision requirements, an exponential mapping or retraction operation can be used to ensure that the point strictly remains on the manifold.

[0049] Finally, the quantization steps of the natural gradient are iterated according to the updated initial parameter until the iteration termination condition is reached: wherein, is the iteration number.

[0050] The present application closely links the parameters in the signal parameter space with the probability distribution of the magnetic field signal by constructing the log-likelihood function, describes the possibility of observing the current magnetic field signal under a given parameter from the probability perspective, provides a solid theoretical foundation for subsequent parameter estimation, and simplifies the subsequent calculation process; the natural gradient descent method used in the present application corrects the gradient according to the Riemannian geometry structure of the manifold, so that the direction of parameter update is more consistent with the real geometry of the parameter space, thereby improving the efficiency and accuracy of the optimization process and overcoming the low efficiency problem of the traditional gradient method in the curved parameter space.

[0051] In an embodiment, the generating the amplitude estimation result of the magnetic field signal based on the initial parameter at the final iteration comprises: taking the initial parameter at the final iteration as a maximum likelihood estimation point on the two-dimensional submanifold, and determining a primary amplitude estimation result based on the maximum likelihood estimation point; performing adaptive adjustment processing on the frequency drift amount to obtain an adaptive frequency drift amount, and updating the projected Fisher information matrix and the log-likelihood function based on the adaptive frequency drift amount to obtain an updated Fisher information matrix and an updated log-likelihood function; re-executing the amplitude estimation step of the magnetic field signal based on the updated Fisher information matrix and the updated log-likelihood function to obtain a secondary amplitude calculation result; iteratively performing the adaptive processing process on the frequency drift amount to iteratively execute the amplitude estimation step of the magnetic field signal based on the re-obtained adaptive frequency drift amount until a preset iteration number is reached, and obtaining a plurality of amplitude calculation results; fusing the primary amplitude estimation result, the secondary amplitude calculation result and the plurality of amplitude calculation results to obtain an optimal amplitude estimation result of the magnetic field signal.

[0052] Specifically, the present application takes the reduced dimension parameter at the final iteration when the iteration termination condition is first reached, that is, the maximum likelihood estimation point on the two-dimensional submanifold, as the convergence point , and the in the point is the primary amplitude estimation result, that is, the single-frame amplitude.

[0053] Since the cable injection current has a frequency drift of ±0.1%-±1%, if the frequency is fixed, it will cause the signal model to be mismatched, therefore, after each iteration (i.e. the convergence point and the corresponding amplitude calculation result are obtained), the frequency drift is updated according to the alternating optimization strategy, that is, the frequency drift is adaptively adjusted: that is, after the magnetic field signal is re-acquired and normalized and discretized, the phase difference method is used to re-estimate the frequency drift, and the sum of the frequency drift in the last iteration is taken as the adaptive frequency drift. Since the change of the frequency drift will cause the model expression of the target cable magnetic field signal and the Riemann geometry structure of the statistical manifold to change, the discrete signal model of the observation sequence is recalculated based on the adaptive frequency drift (to ensure that it matches the current frequency) to update the log-likelihood function and the projected Fisher information matrix (to ensure that the manifold structure is accurate), and the updated Fisher information matrix and the updated log-likelihood function are obtained (for details of the updating process, refer to the aforementioned log-likelihood function construction steps and projected Fisher information matrix generation steps, which will not be described here).

[0054] Then, the amplitude estimation step of the magnetic field signal is re-executed based on the updated Fisher information matrix and the updated log-likelihood function to obtain a secondary amplitude calculation result; the generation process of the secondary amplitude calculation result is the same as that of the initial amplitude estimation result, which will not be described here. Then, the adaptive processing process of the frequency drift is iteratively executed to iteratively execute the amplitude estimation step of the magnetic field signal based on the re-obtained adaptive frequency drift until a preset iteration number (preferably 8) is reached, and multiple amplitude calculation results including three times, four times, five times, …, etc. are obtained; to improve the robustness of the amplitude estimation, the multiple independent estimation results obtained by multiple times of calculation are fused by using a sliding window method to obtain an optimal amplitude estimation result of the magnetic field signal: wherein, is the optimal amplitude estimation result; is the fusion frame number; is the amplitude estimation result of the i-th frame data obtained by using the algorithm. m

[0055] In addition, the inverse matrix of the Fisher information matrix at the convergence point can also be used to calculate the approximate Cramer-Rao lower bound of the amplitude estimation to quantify the estimation accuracy: .

[0056] When the Cramer-Rao lower bound, that is, the estimation variance, exceeds the preset variance threshold, the system automatically prolongs the data acquisition time or increases the fusion frame number to improve the estimation accuracy.

[0057] ​The application is based on maximum likelihood estimation point to determine the initial amplitude estimation result, and determines the maximum likelihood estimation point on a two-dimensional submanifold, can better utilize the geometric structure of the manifold, makes the estimation result more in line with the distribution characteristics of the signal in the low-dimensional space, and improves the estimation accuracy; performs adaptive adjustment processing on the frequency drift amount, can dynamically correct the frequency drift amount according to the actual change of the magnetic field signal, makes the estimation process better adapt to the dynamic characteristics of the signal, and thus improves the estimation precision; estimates the amplitude based on the updated matrix and function, can constantly utilize new information to optimize and adjust the estimation result, gradually improves the precision of the amplitude estimation, obtains the optimal estimation by fusing the amplitude calculation results of multiple times, can comprehensively utilize the estimation information under different stages and different conditions, gives full play to the advantages of each result, makes up for the deficiency of a single estimation result, and thus improves the accuracy and comprehensiveness of the final estimation result.

[0058] In an embodiment, the position of the target cable is determined, comprising: A magnetic field depth relationship model of the target cable is constructed based on electromagnetic field theory; The optimal amplitude estimation result is input into the magnetic field depth relationship model for processing to obtain a preliminary depth of the target cable; The preliminary depth is error corrected, and the position of the target cable is determined based on the corrected preliminary depth.

[0059] Specifically, the application is based on electromagnetic field theory, and the cable depth is calculated according to the function relationship between the magnetic field strength amplitude of two points directly above the cable and the interval and the depth of the two points: According to electromagnetic field theory, the magnetic field strength generated by an infinite straight conductor is inversely proportional to the distance. For a buried cable, it can be approximated as an infinite straight conductor, so the magnetic field strength amplitude of a point on the ground satisfies the relationship with the cable depth h, that is, the magnetic field depth relationship model of the target cable is represented by the following formula: In the formula, is the vacuum permeability; I is the amplitude of the current injected into the target cable.

[0060] However, the on-site current I may be unknown or fluctuating, and directly using the single-point amplitude to calculate the depth will introduce errors. Therefore, the application uses the magnetic field amplitude ratio of two measurement points directly above the cable to eliminate the influence of I . It is assumed that the two measurement points are both located directly above the cable, the horizontal interval distance is , point 1 is located directly above the cable, and point 2 is offset from point 1 by a distance in the vertical direction of the cable, the magnetic field amplitude of point 1 is The magnetic field amplitude of point 2 is According to the magnetic field attenuation law: .

[0061] The optimal amplitude estimation results corresponding to the two points are input into the magnetic field depth relationship model for processing, and the preliminary depth of the target cable can be obtained: .

[0062] If there is uneven ground or cable bending on the spot, the preliminary depth needs to be error corrected, the precision can be improved by taking the average value through multiple measurements or using curve fitting, and finally the corrected preliminary depth is taken as the position of the target cable.

[0063] The signal parameters are converted into geometric objects in the application, the optimization problem under strong noise is solved by using Riemann geometry, and the magnetic field depth relationship model is constructed based on electromagnetic field theory, the characteristics of the target cable and the actual environmental factors are fully considered, the applicability and pertinence of the model to the target cable are improved, and the actual depth of the target cable is more accurately reflected; the cable position is determined by error correction of the preliminary depth, the applicability and robustness of the whole cable position determination method in complex environment are improved, and the position of the target cable can be accurately determined under different conditions.

[0064] The application embodiment designs a cable positioning method based on the problem of cable positioning precision decline caused by inaccurate signal amplitude estimation in complex noise environment, constructs a three-dimensional statistical manifold based on information geometry theory, maps the magnetic field signal to a high-dimensional geometric space, can capture the statistical characteristics of the signal, avoids the strong hypothesis of the signal form in the traditional method, and improves the adaptability of the model to complex electromagnetic environment; the three-dimensional statistical manifold is projected into a two-dimensional submanifold through an alternating optimization strategy, the calculation complexity is significantly reduced, the key geometric information is retained, the redundant dimension is reduced, the update of the Fisher information matrix is more efficient, and the Fisher information matrix after projection focuses on the core change direction of the signal parameter, highlights the geometric features sensitive to amplitude estimation, and improves the pertinence of subsequent gradient calculation; the natural gradient descent method used adjusts the gradient direction based on the geometric structure of the statistical manifold, overcomes the deviation problem of the traditional Euclidean gradient in nonlinear space, ensures that each iteration optimizes along the steepest descent direction, accelerates convergence and avoids local optimum; through joint optimization of the log-likelihood function and the natural gradient descent, the magnetic field amplitude and the cable position are associated, the error accumulation of intermediate variable transmission is reduced, and the positioning precision is improved.

[0065] It should be noted that although each step in the above flowchart is shown in sequence according to the direction of the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, there is no strict order limitation for the execution of these steps, and these steps can be executed in other orders.

[0066] In another embodiment, as shown in Figure 1 The second aspect of the present application provides a cable positioning system, comprising: a data acquisition module 10 configured to acquire a magnetic field signal excited by a target cable; a matrix generation module 20 configured to construct a parameterized probability density model according to the magnetic field signal and a preset signal parameter space based on information geometry theory, to determine a three-dimensional statistical manifold, and to generate a Fisher information matrix according to the three-dimensional statistical manifold; a manifold projection module 30 configured to project the three-dimensional statistical manifold to obtain a two-dimensional submanifold to update the Fisher information matrix, and to generate a projected Fisher information matrix by using an alternating optimization strategy; a position determination module 40 configured to construct a log-likelihood function according to the signal parameter space and the parameterized probability density model, and to obtain an amplitude estimation result of the magnetic field signal to determine the position of the target cable by iteratively solving the log-likelihood function based on the projected Fisher information matrix using a natural gradient descent method.

[0067] It should be noted that each module in the above cable positioning system can be implemented in whole or in part by software, hardware, and combinations thereof. Each module described above can be embedded in or independent of a processor in a computer device in hardware form, or can be stored in a memory in a computer device in software form, so as to be called and executed by a processor to perform the operations corresponding to each module. For specific limitations of a cable positioning system, refer to the limitations of a cable positioning method described above, both of which have the same functions and effects, and will not be described here.

[0068] The third aspect of the present application provides an electronic device, comprising: a processor, a memory, and a bus; the bus is configured to connect the processor and the memory; the memory is configured to store operation instructions; the processor is configured to execute the operations corresponding to the cable positioning method of the first aspect of the present application by calling the operation instructions.

[0069] In an optional embodiment, an electronic device is provided, as shown in Figure 3 , Figure 3The electronic device 5000 shown includes a processor 5001 and a memory 5003. The processor 5001 and the memory 5003 are connected, for example, via a bus 5002. Optionally, the electronic device 5000 can also include a transceiver 5004. It should be noted that the transceiver 5004 is not limited to one in actual applications, and the structure of the electronic device 5000 does not constitute a limitation on the embodiments of the present application.

[0070] The processor 5001 can be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in combination with the present disclosure. The processor 5001 can also be a combination that implements a computing function, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like.

[0071] The bus 5002 can include a path for transmitting information between the above-mentioned components. The bus 5002 can be a PCI bus or an EISA bus, etc. The bus 5002 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 In the figure, only one thick line is used to represent the bus, but it does not mean that there is only one bus or only one type of bus.

[0072] The memory 5003 can be a ROM or other type of static storage device that can store static information and instructions, a RAM or other type of dynamic storage device that can store information and instructions, an EEPROM, a CD-ROM or other optical disk storage, an optical disk storage (including a compact disk, a laser disk, an optical disk, a digital versatile disk, a Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto.

[0073] The memory 5003 is used to store application program code for executing the scheme of the present application, and is controlled by the processor 5001 to execute. The processor 5001 is used to execute the application program code stored in the memory 5003 to realize the content shown in any of the preceding method embodiments.

[0074] The electronic device includes, but is not limited to, a mobile terminal such as a mobile phone, a notebook computer, a digital broadcast receiver, a PDA (Personal Digital Assistant), a PAD (Tablet Personal Computer), a PMP (Portable Multimedia Player), a vehicle terminal (such as a vehicle navigation terminal), and the like, and a fixed terminal such as a digital TV, a desktop computer, and the like.

[0075] The fourth aspect of the present application provides a computer readable storage medium, and the computer readable storage medium stores a computer program, and the program is executed by a processor to realize the cable positioning method in the first aspect of the present application.

[0076] Another embodiment of the present application provides a computer readable storage medium, and the computer readable storage medium stores a computer program, and when the computer program is executed on a computer, the computer can execute the corresponding content in the foregoing method embodiments.

[0077] In addition, the embodiments of the present application also provide a computer readable storage medium, and the computer readable storage medium stores a computer program, and the program is executed by a processor to realize the steps of the foregoing method.

[0078] In summary, the present application relates to the technical field of cable positioning, and discloses a cable positioning method, system, device and medium, the method collects the magnetic field signal excited by the target cable; based on the information geometry theory, a parameterized probability density model is constructed according to the magnetic field signal and a preset signal parameter space to determine a three-dimensional statistical manifold, and then a Fisher information matrix is generated; an alternating optimization strategy is used to project the three-dimensional statistical manifold to obtain a two-dimensional submanifold and the corresponding projected Fisher information matrix; a log-likelihood function is constructed according to the signal parameter space and the parameterized probability density model, and based on the projected Fisher information matrix, a natural gradient descent method is used to iteratively solve the log-likelihood function to obtain an amplitude estimation result of the magnetic field signal to determine the position of the target cable; the present application combines the information geometry theory, the alternating optimization strategy and the natural gradient descent method to realize efficient and accurate positioning of the cable position.

[0079] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts of each embodiment can be referred to each other, and each embodiment mainly describes the difference from other embodiments. Especially, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the related parts can be referred to the part of the method embodiment. It should be noted that, each technical feature of the above-mentioned embodiments can be combined arbitrarily, in order to make the description simple, each technical feature of the above-mentioned embodiments is not described all possible combinations, however, as long as the combination of these technical features does not exist contradictory, it should be considered that it is within the scope of the present application.

[0080] The above embodiments only express several preferred embodiments of the present application, which are described in more detail and in more detail, but cannot be understood as limiting the scope of the patent. It should be noted that for ordinary skilled in the art, several improvements and replacements can be made without departing from the technical principles of the present application, and these improvements and replacements should also be considered as the protection scope of the present application. Therefore, the protection scope of the present application patent should be subject to the protection scope of the claims.

Claims

1. A cable positioning method, characterized in that, include: Collect the magnetic field signal excited by the target cable; Based on information geometry theory, a parameterized probability density model is constructed according to the magnetic field signal and a preset signal parameter space to determine the three-dimensional statistical manifold, and a Fisher information matrix is ​​generated according to the three-dimensional statistical manifold. An alternating optimization strategy is used to project the three-dimensional statistical manifold to obtain a two-dimensional submanifold, which is then used to update the Fisher information matrix, generating the projected Fisher information matrix. A log-likelihood function is constructed based on the signal parameter space and the parameterized probability density model. Based on the projected Fisher information matrix, the log-likelihood function is iteratively solved using the natural gradient descent method to obtain the amplitude estimation result of the magnetic field signal and determine the location of the target cable.

2. The cable positioning method according to claim 1, characterized in that, The step of constructing a parameterized probability density model based on the magnetic field signal and a preset signal parameter space includes: The magnetic field signal is normalized and discretized to obtain a normalized discrete signal sequence; Based on the signal parameter space, the product of noise probabilities at each point in the normalized discrete signal sequence is quantized to construct the parameterized probability density model.

3. The cable positioning method according to claim 1, characterized in that, The step of determining a three-dimensional statistical manifold and generating a Fisher information matrix based on the three-dimensional statistical manifold includes: Based on the parameterized probability density model, the probability distribution corresponding to each parameter combination determined according to the signal parameter space is used as a point on the statistical manifold to construct the three-dimensional statistical manifold. The Fisher information matrix is ​​determined based on the Riemannian geometry of the three-dimensional statistical manifold.

4. The cable positioning method according to claim 2, characterized in that, The method of projecting the three-dimensional statistical manifold using an alternating optimization strategy to obtain a two-dimensional submanifold for updating the Fisher information matrix, and generating the projected Fisher information matrix, includes: The frequency of the normalized discrete signal sequence is estimated using the phase difference method to obtain the instantaneous angular frequency, and the frequency drift is determined by combining the signal parameter space. The signal parameter space is reduced in dimension based on the frequency drift to obtain the dimension reduction parameters, and the three-dimensional statistical manifold is projected onto the dimension reduction parameters to obtain the two-dimensional submanifold. The Fisher information matrix is ​​updated based on the two-dimensional submanifold to obtain the projected Fisher information matrix.

5. A cable positioning method according to claim 4, characterized in that, The step of constructing a log-likelihood function based on the signal parameter space and the parameterized probability density model, and iteratively solving the log-likelihood function using the natural gradient descent method based on the projected Fisher information matrix to obtain the amplitude estimation result of the magnetic field signal includes: The dimensionality reduction parameters are initialized to obtain initial parameters, and based on the initial parameters, the logarithm of the parameterized probability density model is taken to obtain the log-likelihood function. Based on the initial parameters, the derivative of the log-likelihood function is obtained to get the ordinary gradient, which is then combined with the projected Fisher information matrix to quantize and obtain the natural gradient. The initial parameters are updated along the direction of the natural gradient. The quantization steps of the natural gradient are iterated according to the updated initial parameters until the iteration termination condition is met. The amplitude estimation result of the magnetic field signal is generated based on the initial parameters at the time of the final iteration.

6. A cable positioning method according to claim 5, characterized in that, The generation of amplitude estimation results for the magnetic field signal based on the initial parameters at the final iteration includes: The initial parameters at the final iteration are used as the maximum likelihood estimation point on the two-dimensional submanifold, and the initial magnitude estimation result is determined based on the maximum likelihood estimation point. The frequency drift is adaptively adjusted to obtain an adaptive frequency drift, which is then used to update the projected Fisher information matrix and the log-likelihood function, resulting in an updated Fisher information matrix and an updated log-likelihood function. Based on the updated Fisher information matrix and the updated log-likelihood function, the amplitude estimation step for the magnetic field signal is re-executed to obtain the secondary amplitude calculation result; The adaptive processing of the frequency drift is performed iteratively, and the amplitude estimation step of the magnetic field signal is performed iteratively based on the newly obtained adaptive frequency drift until the preset number of iterations is reached, and multiple amplitude calculation results are obtained. The initial amplitude estimation result, the secondary amplitude calculation result, and the multiple amplitude calculation results are fused to obtain the optimal amplitude estimation result for the magnetic field signal.

7. A cable positioning method according to claim 6, characterized in that, The step of determining the location of the target cable includes: Based on electromagnetic field theory, a magnetic field burial depth relationship model for the target cable is constructed; The optimal amplitude estimation result is input into the magnetic field burial depth relationship model for processing to obtain the preliminary burial depth of the target cable; The initial burial depth is corrected for errors, and the location of the target cable is determined based on the corrected initial burial depth.

8. A cable positioning system, characterized in that, include: The data acquisition module is used to acquire the magnetic field signal excited by the target cable; The matrix generation module is used to construct a parameterized probability density model based on the magnetic field signal and a preset signal parameter space, according to the information geometry theory, to determine the three-dimensional statistical manifold, and to generate the Fisher information matrix based on the three-dimensional statistical manifold. The manifold projection module is used to project the three-dimensional statistical manifold using an alternating optimization strategy to obtain a two-dimensional submanifold to update the Fisher information matrix and generate the projected Fisher information matrix. The location determination module is used to construct a log-likelihood function based on the signal parameter space and the parameterized probability density model, and to iteratively solve the log-likelihood function using the natural gradient descent method based on the projected Fisher information matrix, so as to obtain the amplitude estimation result of the magnetic field signal and determine the location of the target cable.

9. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the cable positioning method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein when the device containing the computer-readable storage medium executes the computer program, it implements the cable positioning method as described in any one of claims 1 to 7.

Images