Method and system for adaptive detection of low signal-to-noise ratio vector magnetic anomaly of ocean floating platform

Abstract

The present application belongs to the technical field of vector magnetic detection, and relates to a low signal-to-noise ratio vector magnetic anomaly adaptive detection method and system for a marine floating platform. The method comprises the following steps: step one, synchronously collecting triaxial vector magnetic field data and floating platform attitude data; step two, constructing an attitude rotation matrix and performing geomagnetic projection compensation; step three, establishing a background state space model; step four, residual whitening and GLRT statistical quantity construction; and directly gating and fusing decision outputs of multi-scale self-resetting CUSUM. The present application also provides a system adapted to the method. Under the condition that the marine floating platform attitude is disturbed, non-stationary background noise and low signal-to-noise ratio weak magnetic targets coexist, the present application effectively suppresses the interference of geomagnetic projection fluctuation on detection performance, prevents the false absorption of the target magnetic anomaly by the adaptive filter, and improves the event-level detection probability and shortens the average detection delay while maintaining a low false alarm rate.

Description

Technical Field

[0001] This invention belongs to the field of vector magnetic detection technology, and relates to an adaptive detection method and system for low signal-to-noise ratio vector magnetic anomalies on marine floating platforms. Background Technology

[0002] Marine magnetic anomaly detection is an important tool in underwater target monitoring, marine environmental observation, marine resource exploration, and non-acoustic detection systems. Compared with active acoustic detection, magnetic anomaly detection operates in a passive receiving mode, offering advantages such as strong concealment and better resistance to acoustic interference, making it suitable for long-term monitoring in complex sea conditions. Floating marine platforms offer advantages such as flexible deployment, relatively low maintenance costs, long-term continuous observation capabilities, and ease of networking, making them ideal platforms for conducting marine magnetic field monitoring using high-precision three-axis vector magnetic sensors.

[0003] However, floating platforms operate in environments with multiple disturbances, including waves, swells, wind loads, and ocean currents. Continuous changes in roll, pitch, and heading cause significant fluctuations in the projection of the strong geomagnetic field onto the three-axis magnetic sensor coordinate system. The amplitude of this attitude-coupled projection term typically far exceeds that of weak target magnetic anomalies. Treating it merely as ordinary measurement noise will result in a non-stationary residual sequence and a significantly increased false alarm rate. Furthermore, the ocean background magnetic field and instrument measurement noise exhibit characteristics such as uncertainty, time-varying nature, correlation, and slow drift. Standard Kalman filtering, using fixed process noise covariance and fixed measurement noise covariance, struggles to maintain statistical consistency of innovation residuals under different sea states. While Sage-Husa adaptive Kalman filtering can estimate noise statistics online, continuously updating the measurement noise covariance and background state for all major innovations can lead to target components being misjudged as sudden changes in background noise and absorbed by the filter when target magnetic anomalies occur. This manifests as measurement noise covariance expansion, whitening residual energy attenuation, and a significant reduction in detection statistics.

[0004] Existing detection methods still suffer from problems such as unknown target orientation and single time scale. Adaptive disturbance suppression methods based on Normalized Least Mean Squares (NLMS) typically focus on eliminating disturbance components caused by the swaying of the floating platform, but do not clearly distinguish between background disturbances and suspected target magnetic anomalies. Although ordinary whitening energy detection or GLRT methods can handle targets with unknown orientations, their detection sensitivity will still decrease if their input residuals have been absorbed by adaptive filters or weakened by dilated covariance whitening. Single-scale CUSUM detectors are only sensitive to targets of a specific duration, and it is difficult to maintain good detection performance for both short-lived strong targets and persistent weak targets simultaneously. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention aims to effectively suppress the interference of geomagnetic projection fluctuations on detection performance under the conditions of attitude disturbances of marine floating platforms, non-stationary background noise, and weak magnetic targets with low signal-to-noise ratios, prevent the adaptive filter from falsely absorbing the magnetic anomalies of the target, and improve the event-level detection probability and shorten the average detection delay while maintaining a low false alarm rate.

[0006] The technical solution provided by this invention is: an adaptive detection method for low signal-to-noise ratio vector magnetic anomalies on marine floating platforms, comprising the following steps: Step 1: Simultaneously acquire three-axis vector magnetic field observation data of the marine floating platform and the corresponding carrier attitude data; Step 2: Construct a time-varying attitude rotation matrix based on the attitude data, and use the time-varying attitude rotation matrix to compensate for the projection of the geomagnetic field in the sensor coordinate system to obtain the compensated observation; Step 3: Construct a background state-space model based on the compensated observations, perform adaptive Kalman filtering, calculate innovation residuals, innovation covariance, and normalized innovation squares, and perform target protection gating based on the normalized innovation squares; Step 4: Under the target protection gating, the innovative residual is covariance whitened to construct the GLRT three-axis joint statistics under the conditions of unknown target magnetic anomaly amplitude and unknown direction, and the GLRT three-axis joint statistics are input into the direct gating branch and the multi-scale self-resetting CUSUM branch. Step 5: Perform a fusion decision based on the output results of the direct gating branch and the multi-scale self-resetting CUSUM branch. An alarm is triggered when at least two of the output results of the multi-scale self-resetting CUSUM branch are true, or an alarm is triggered based on the output result of the direct gating branch.

[0007] Furthermore, in step two, the geomagnetic field vector in the navigation coordinate system is obtained based on pre-deployment offline calibration, WMM geomagnetic model lookup table, or mean estimation of background segments without a target. And calculate the compensated observations The post-compensation observations mainly include ocean background disturbances, potential target magnetic anomalies, measurement noise, and attitude compensation residuals.

[0008] Furthermore, the compensated observation satisfies the following attitude-coupled magnetic field observation model: ; Or it can be expressed as: ; in, This is to compensate for the background magnetic field disturbance. The target magnetic anomaly vector, To measure noise, The background state vector, This is the observation matrix.

[0009] Furthermore, in step three, the background state-space model includes at least a three-axis slow random walk component and at least one set of three-axis periodic harmonic components, and satisfies: ; in, Here is the state transition matrix. The process noise is represented by the periodic harmonic components, which are described by the rotating state transition submatrix determined by the sampling period and the dominant wave frequency.

[0010] Furthermore, in step three, with As a weight for effective updating of new observations The trigger signal, where, To normalize the innovation square, Let be the dimension of the observation vector; when When within the background consistency range, slow adaptation of the measurement noise covariance is allowed; when When the target protection threshold is exceeded, Reduce to a preset lower limit or set to zero, where: .

[0011] Furthermore, when When the noise covariance is less than the background consistency threshold, the measurement noise covariance is updated online according to a Sage-Husa type restricted recursion; when When the noise level exceeds the target protection threshold, the noise covariance will be measured. Frozen as Fallback to nominal noise covariance with small weights Alternatively, limit it to a preset upper and lower bound, and set the effective Kalman gain to: ; in, This reduces the updating effect of current observations on the background state, so that the suspected target magnetic anomalies are mainly retained in the innovative residuals.

[0012] Furthermore, in step four, the nominal measurement noise covariance, the frozen measurement noise covariance, or the detection covariance constrained by eigenvalue upper and lower bounds are used. Whiten the innovation residuals, and satisfy the following conditions: ; in, The lower trigonometric factor of Cholesky. The normalized residual for whitening.

[0013] Furthermore, the GLRT triaxial joint statistic in step four is: ; in, This is the drift penalty parameter or offset penalty threshold.

[0014] Furthermore, the multi-scale self-resetting CUSUM branch includes a short-scale branch, a medium-scale branch, and a long-scale branch, and its recursive form is as follows: ; in, These are the short-scale branch, the medium-scale branch, and the long-scale branch, respectively. When the accumulated amount of any scale branch exceeds the corresponding alarm threshold, the scale branch outputs an alarm and immediately resets, and the reset does not change the accumulated state of other scale branches.

[0015] Furthermore, the present invention also provides a vector magnetic anomaly detection system for marine floating platforms. This system includes a triaxial magnetic sensor, an attitude measurement unit, a data synchronization acquisition module, a processor, an alarm output module, a memory / parameter configuration module, and a display / communication interface. The processor includes an attitude compensation module, an adaptive Kalman filter module, a NIS target protection gating module, a restricted covariance whitening module, a GLRT statistics construction module, and a direct gating and three-scale CUSUM fusion module. The processor is configured to execute the steps of the method.

[0016] Compared with existing technologies, this invention has the following technical advantages: It employs a time-varying attitude rotation matrix to compensate for geomagnetic projection changes online, effectively suppressing non-stationary residuals caused by floating platform attitude disturbances; it constructs a target protection gating mechanism based on normalized innovative squares, constraining the recursive update of measurement noise covariance and reducing the effective Kalman gain when a suspected target appears, thereby preventing the adaptive filter from mistakenly absorbing target magnetic anomalies; it constructs a three-axis joint GLRT statistic using nominal or eigenvalue-constrained detection covariance, enabling vector magnetic anomaly detection even when the target amplitude and direction are unknown; and it combines direct gating with a multi-scale self-resetting CUSUM voting mechanism, giving the system both rapid response capability for short-term strong magnetic anomalies and cumulative detection capability for continuous weak magnetic anomalies. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall process of the vector magnetic anomaly detection method for marine floating platforms according to the present invention; Figure 2 This is a block diagram of the vector magnetic anomaly detection system for marine floating platforms according to the present invention; Figure 3 This is a schematic diagram of the NIS target protection gating and restricted covariance update process of the present invention; Figure 4This is a schematic diagram of the direct gating and three-scale self-resetting CUSUM voting fusion structure of the present invention; Figure 5 The following are event-level detection probability curves of the present invention under different signal-to-noise ratio conditions; wherein, (a) Pd-SNR curve; (b) Pfa-SNR curve; Figure 6 The ROC curves (MC=200) of the present invention and the comparative method are shown. Detailed Implementation

[0018] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The listed embodiments are intended to illustrate the implementation of the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0019] The flowchart of the adaptive detection method for low signal-to-noise ratio vector magnetic anomalies of marine floating platforms provided by this invention is as follows: Figure 1 As shown, at each sampling moment, three-axis vector magnetic field data and attitude data are acquired synchronously, and attitude compensation, background state prediction, NIS target protection gating, residual whitening, GLRT statistics construction, multi-scale sequential cumulative summation, and fusion decision are performed sequentially. Specifically, the following steps are included: Step 1: Synchronously acquire three-axis vector magnetic field data and floating platform attitude data. The three-axis magnetic sensor can be a fluxgate magnetometer, Overhauser magnetometer, optically pumped magnetometer, quantum magnetometer, or a combination thereof; the attitude measurement unit may include an IMU, electronic compass, gyroscope, and accelerometer. The data synchronization acquisition module times-aligns the magnetic field sampling time and the attitude sampling time to obtain three-axis vector magnetic field observation data at the same sampling time. Roll angle Pitch angle and heading angle .

[0020] The triaxial vector magnetic field observation data are represented as follows:

[0021] in, They are respectively the first one k Sensor coordinate system at each sampling time x axis, y shaft and z The magnetic field components of the axis.

[0022] Step 2: Construct the attitude rotation matrix and perform geomagnetic projection compensation. The processor constructs a time-varying attitude rotation matrix based on the attitude angles. : .

[0023] Then, based on offline calibration before deployment, WMM geomagnetic model lookup tables, or mean estimation of background segments without a target, the following results are obtained: And calculate the compensated observations : ; in, The geomagnetic field vector in the navigation coordinate system. For the post-compensation observation.

[0024] The compensated observations mainly include ocean background disturbances, potential target magnetic anomalies, measurement noise, and attitude compensation residuals. The compensated observations satisfy the following attitude-coupled magnetic field observation model: ; Or it can be expressed as: ; in, This is to compensate for the background magnetic field disturbance. The target magnetic anomaly vector, To measure noise, The background state vector, This is the observation matrix.

[0025] Step 3: Establish the background state space model. Background state vector It may include a three-axis slow random walk component and one or more sets of three-axis periodic harmonic components. The state transition submatrix of the periodic harmonic components can be determined by the sampling period and the dominant wave frequency. Determined, where the angular frequency is .

[0026] In one implementation, the background state model is as follows: ; ; in, Here is the state transition matrix. This is process noise.

[0027] The processor first predicts the background state and prediction error covariance based on the state transition matrix, and then calculates the innovation residual based on the compensated observations. and innovation covariance And calculate the normalized square of innovation. It is used for filtering consistency judgment.

[0028] Innovation residuals, innovation covariance, and normalized squared innovation are calculated according to the following formulas: ; ; ; in, For the prediction error covariance matrix, To measure the noise covariance matrix; in the case of no target and consistent filtering, Right now (Normalized innovation square) approximately follows a chi-square distribution with degrees of freedom equal to the three-axis observation dimension.

[0029] when When the background is within a consistent range, the processor assumes the current observations primarily reflect normal background disturbances, allowing the measurement noise covariance to update slowly. For example, the following restricted recursion can be used: ; when When the target protection threshold is exceeded, the processor considers that the current residual may contain the target magnetic anomaly, and at this time... Set to zero or limit to a preset lower limit, freeze or limit the measurement noise covariance update, and set the effective Kalman gain to zero. Perform a state update. This process ensures that suspected target magnetic anomalies do not enter or are minimally entered into the measurement noise covariance recursion term, while reducing the absorption of target components by the background state.

[0030] In one implementation, Depend on Triggered and satisfied: ; ; in, To adjust the rate parameter, and These are the effective lower and upper limits for updating weights, respectively. The specific process is as follows: Figure 3 As shown.

[0031] Step 4: Residual whitening and GLRT statistic construction. The processor uses nominal or bounded detection covariance. Instead of whitening the covariance directly inflated by major innovations during the target period, whitening is performed. The whitening residual is , The lower triangular factor of Cholesky. This represents the whitening normalized residual. In practice, it can also be solved by solving linear equations. get To improve numerical stability.

[0032] Constructing a triaxial joint statistic of GLRT based on whitening residuals : ; in, For drift penalty parameters or offset penalty thresholds This statistic does not depend on prior knowledge of the target magnetic anomaly direction and is applicable to the detection of vector magnetic anomalies in unknown directions.

[0033] Step 5: Fusion decision-making using direct gating and three-scale self-resetting CUSUM voting. (For example...) Figure 4 As shown, the direct gating branch will The alarm result is obtained by comparing it with the direct gating threshold. The three-scale CUSUM branches correspond to short-scale, medium-scale, and long-scale operations, respectively, and their recursive forms are as follows: ; in, These are the short-scale branch, the medium-scale branch, and the long-scale branch.

[0034] When any scale Cumulative CUSUM Exceeding the corresponding alarm threshold At that time, the scale branch outputs an alarm and immediately... Reset to 0; this reset only applies to the current scale branch and does not change the cumulative state of other scale branches.

[0035] The final fusion decision was: ; in, For the final alarm result, express , and At least two of them are true. , and Alarms are output for short-scale branches, medium-scale branches, and long-scale branches, respectively.

[0036] Thresholds can be determined by empirical quantiles of the targetless calibration segment, Bonferroni false alarm rate allocation, or Monte Carlo calibration methods to meet system-level constant false alarm rate constraints. Direct gating thresholds are used for rapid responses to significant anomalies, short-scale branches for responding to short-duration strong magnetic anomalies, mesoscale branches for responding to magnetic anomalies of general duration, and long-scale branches for accumulating persistent weak magnetic anomalies.

[0037] To achieve the above method, the present invention also provides a vector magnetic anomaly detection system for marine floating platforms adapted thereto, such as... Figure 2 As shown, it includes: a triaxial magnetic sensor, an attitude measurement unit, a data synchronization acquisition module, a processor and an alarm output module, a memory / parameter configuration module, and a display / communication interface.

[0038] The processor includes an attitude compensation module, an adaptive Kalman filter module, an NIS target protection gating module, a restricted covariance whitening module, a GLRT statistics construction module, and a direct gating and three-scale CUSUM fusion module.

[0039] The function of the attitude compensation module is to eliminate the interference of carrier attitude changes on magnetic field observation and to convert the magnetic field data of the sensor system to the geographic coordinate system.

[0040] The adaptive Kalman filter module is designed to perform optimal estimation of the magnetic field data after attitude compensation, while adaptively adjusting the system / observation noise covariance to suppress noise and track changes in the real magnetic field.

[0041] The NIS target protection gating module uses Normalized Innovative Square (NIS) to detect whether the innovative residual exceeds the normal statistical range, thereby achieving anomaly protection and coarse gating for Kalman filtering.

[0042] The function of the constrained covariance whitening module is to perform covariance whitening on the innovative residuals of the Kalman filter output, eliminate the correlation between components and normalize the variance, and generate whitened residuals with a standard Gaussian distribution.

[0043] The GLRT statistic construction module is designed to construct a generalized likelihood ratio test (GLRT) statistic based on whitened residuals, enabling optimal detection of magnetic field abrupt changes.

[0044] The direct gating and three-scale CUSUM fusion module integrates the detection results of direct gating (such as GLRT, NIS) with multi-scale (three-scale) self-resetting CUSUM, and outputs the results through voting decision-making to achieve highly robust detection of magnetic field mutations / anomalies.

[0045] In a set of simulation embodiments, a target magnetic anomaly time series was constructed using a publicly available dataset of ferromagnetic ship magnetic feature reproduction. Multi-frequency harmonic background, AR(1) related noise, slow random walk, measurement noise, and attitude compensation residuals were superimposed on this target magnetic anomaly sequence. Detection evaluation used an event-level detection probability Pd, meaning that a successful detection was determined by triggering at least one alarm during the target's passage. The background false alarm rate Pfa used the false alarm rate of background sampling points outside the target interval and its buffer zone. Monte Carlo iterations were performed 200 times under each SNR (signal-to-noise ratio) condition. The threshold was determined by the empirical quantile of the target-free calibration segment, and the GLRT drift penalty parameter was used. Take 1.0.

[0046] Table 1. Comparison of event-level detection performance under different SNR conditions (MC=200) .

[0047] Table 1 and Figure 5 It can be seen that in the low signal-to-noise ratio range of SNR of 0.6 to 1.7, the event-level detection probabilities of the method of the present invention reach 0.80, 0.86 and 0.91 respectively, which are all better than the standard KF single-scale CUSUM and the ordinary adaptive KF single-scale CUSUM; the background false alarm rate is stably maintained in the range of 0.03 to 0.04. Figure 6 The ROC curves also show that, under the constraint of similar false alarm rates, the method of the present invention has a higher detection probability at the low SNR operating point.

[0048] Table 2. Comparison of overall performance of different methods under the condition of SNR≈3.2 (MC=200) .

[0049] As shown in Table 2, under the condition that the SNR is about 3.2, the average Pd value of the method of the present invention is 0.970, the average background false alarm rate Pfa value is 0.031, and the average detection delay is 191.9 steps. Compared with the standard KF single-scale method, the detection probability is increased by about 6.0 percentage points and the average delay is shortened by about 54.4 steps. Compared with the ordinary adaptive KF single-scale method, the detection probability is increased by about 2.0 percentage points, and the average detection delay is reduced under the condition of similar false alarm rate.

[0050] This invention can also be used to construct a vector magnetic anomaly detection system for marine floating platforms. For example... Figure 2 As shown, the system consists of a three-axis vector magnetic sensor, an attitude measurement unit, a data synchronization acquisition module, a processor, and an alarm output module. The processor is responsible for performing attitude compensation, adaptive filtering, NIS target protection gating, residual whitening, GLRT statistics construction, multi-scale CUSUM accumulation, and fusion decision; the alarm output module is responsible for outputting alarm flags, alarm times, detection delays, and statistical information.

[0051] The processor can be deployed on a local edge computing terminal of a floating platform or on a shore-based server. The processor can be an embedded CPU, DSP, FPGA, GPU, microcontroller, industrial computer, or a combination thereof. This system can be used in scenarios such as port surveillance, long-term marine environmental monitoring, underwater target non-acoustic detection, and marine resource exploration. This invention is not limited to the above embodiments. The three-scale CUSUM can be expanded to two-scale, four-scale, or more scales; the fusion rules can employ arbitrary scale triggering, two-out-of-three voting, three-out-of-four voting, weighted voting, persistent constraint voting, or cost-sensitive fusion; the NIS threshold can employ a theoretical chi-square threshold, an empirical quantile threshold, or an online adaptive threshold; the background state model can incorporate multiple harmonic components of different dominant frequencies, temperature drift state, sensor zero-bias state, or platform magnetic interference state according to sea state; the whitening covariance can employ innovative covariance, nominal measurement noise covariance, or adaptive covariance with upper and lower bound constraints. All equivalent substitutions and modifications described above, without departing from the core concept of this invention, should fall within the scope of protection of this invention.

Claims

1. An adaptive detection method for low signal-to-noise ratio vector magnetic anomalies on marine floating platforms, characterized in that, Includes the following steps: Step 1: Simultaneously acquire three-axis vector magnetic field observation data of the marine floating platform and the corresponding carrier attitude data; Step 2: Construct a time-varying attitude rotation matrix based on the attitude data, and use the time-varying attitude rotation matrix to compensate for the projection of the geomagnetic field in the sensor coordinate system to obtain the compensated observation; Step 3: Construct a background state-space model based on the compensated observations, perform adaptive Kalman filtering, calculate innovation residuals, innovation covariance, and normalized innovation squares, and perform target protection gating based on the normalized innovation squares; specifically: using As a weight for effective updating of new observations The trigger signal, where, To normalize the innovation square, Let be the dimension of the observation vector; when When within the background consistency range, slow adaptation of the measurement noise covariance is allowed; when When the target protection threshold is exceeded, Reduce to a preset lower limit or set to zero, where: ； when When the noise covariance is less than the background consistency threshold, the measurement noise covariance is updated online according to a Sage-Husa type restricted recursion; when When the noise level exceeds the target protection threshold, the noise covariance will be measured. Frozen as Regress to nominal covariance with small weights Alternatively, limit it to a preset upper and lower bound, and set the effective Kalman gain to: ； in, This is to weaken the updating effect of current observations on the background state, so that the suspected target magnetic anomaly is retained in the innovative residual; Step 4: Under the target protection gating, the innovative residual is covariance whitened to construct the GLRT three-axis joint statistics under the conditions of unknown target magnetic anomaly amplitude and unknown direction, and the GLRT three-axis joint statistics are input into the direct gating branch and the multi-scale self-resetting CUSUM branch. Step 5: Perform a fusion decision based on the output results of the direct gating branch and the multi-scale self-resetting CUSUM branch. An alarm is triggered when at least two of the output results of the multi-scale self-resetting CUSUM branch are true, or an alarm is triggered based on the output result of the direct gating branch.

2. The method according to claim 1, characterized in that, In step two, the geomagnetic field vector in the navigation coordinate system is obtained based on offline calibration before deployment, table lookup of the WMM geomagnetic model, or mean estimation of the background segment without a target. And calculate the compensated observations The post-compensation observations mainly include ocean background disturbances, potential target magnetic anomalies, measurement noise, and attitude compensation residuals.

3. The method according to claim 2, characterized in that, The compensated observations satisfy the following attitude-coupled magnetic field observation model: ； Or it can be expressed as: ; in, This is to compensate for the background magnetic field disturbance. The target magnetic anomaly vector, To measure noise, The background state vector, This is the observation matrix.

4. The method according to claim 1, characterized in that, In step three, the background state-space model includes at least a three-axis slow random walk component and at least one set of three-axis periodic harmonic components, and satisfies: ； in, Here is the state transition matrix. The process noise is represented by the periodic harmonic components, which are described by the rotating state transition submatrix determined by the sampling period and the dominant wave frequency.

5. The method according to claim 1, characterized in that, In step four, the nominal measurement noise covariance, the frozen measurement noise covariance, or the detection covariance constrained by eigenvalue upper and lower bounds are used. Whiten the innovation residuals, and satisfy the following conditions: ； in, The lower triangular factor of Cholesky. For whitening normalization residuals; To innovate residuals.

6. The method according to claim 5, characterized in that, GLRT triaxial joint statistics in step four for: ； in, This is the drift penalty parameter or offset penalty threshold.

7. The method according to claim 6, characterized in that, The multi-scale self-resetting CUSUM branch includes a short-scale branch, a medium-scale branch, and a long-scale branch, and its recursive form is as follows: ; in, These are the short-scale branch, the medium-scale branch, and the long-scale branch, respectively. The cumulative CUSUM value at the (k-1)th sampling time; When the accumulated amount of any scale branch exceeds the corresponding alarm threshold, the scale branch outputs an alarm and immediately resets, and the reset does not change the accumulated state of other scale branches.

8. A vector magnetic anomaly detection system for a marine floating platform, characterized in that, It includes a triaxial magnetic sensor, an attitude measurement unit, a data synchronization acquisition module, a processor, and an alarm output module; the processor is configured to perform the method according to any one of claims 1 to 7.

Images