A target detection method based on magnetism multi-feature fusion

Abstract

The present application relates to the technical field of magnetic anomaly detection, and particularly relates to a target detection method based on magnetism multi-feature fusion; comprising: S1. removing environmental magnetic noise; S2. adopting a combination of a polynomial model and an autoregressive model to decompose the measured magnetic field into two parts; S3. based on the magnetic measurement data after the denoising processing of S2, constructing OBF features, ME features and HMS features of the target to comprehensively characterize the target from multiple dimensions; S4. adopting p-norm as a detection index after feature fusion; S5. calculating the cumulative distribution function of noise features by using kernel density estimation (KDE), determining the corresponding threshold according to the set false alarm rate (FAR), and outputting the detection result according to the threshold. The present application solves the problems of low detection probability and large false alarm rate of the existing detection method in the case of low signal-to-noise ratio and large environmental interference.

Description

Technical Field

[0001] This invention relates to the field of magnetic anomaly detection technology, specifically to a target detection method based on the fusion of multiple magnetic features. Background Technology

[0002] The Earth's magnetic field is its inherent vector field, composed of internal and external field sources. The internal field is a stable magnetic field, consisting of 95% core field, 4% crustal anomaly field, and less than 1% external magnetic field. It changes extremely slowly over time and is present throughout all time and regions. When magnetic targets (containing iron, nickel, etc.) are present in the measurement area, they are magnetized by the Earth's magnetic field, generating induced magnetic fields that, in turn, alter the original distribution of the Earth's magnetic field, thus creating magnetic anomalies. Magnetic detection technology analyzes the magnetic measurement data of targets to extract the characteristics of the magnetic anomalies they produce, thereby obtaining information about the targets.

[0003] Compared to the Earth's magnetic field, the magnetic anomaly signal of a target is very small. As the distance between the magnetic target and the sensor increases, the detected magnetic anomaly becomes smaller and smaller. This means that the magnetic anomaly signal generated by a distant target is usually buried deep in the background magnetic noise. Traditional magnetic anomaly detection techniques are mostly based on the detection of a single target feature. Due to problems such as low signal-to-noise ratio and large background magnetic interference, their applicable scenarios are limited, especially in complex environments, where the detection effect is very limited. Summary of the Invention

[0004] This invention provides a target detection method based on magnetic multi-feature fusion, which aims to detect and search for magnetic anomaly targets and solve the problems of low detection probability and high false alarm rate of existing detection methods under conditions of low signal-to-noise ratio and large environmental interference.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] This invention provides a target detection method based on magnetic multi-feature fusion, comprising the following steps: S1. Based on the sampling rate of the magnetic measurement data, determine the cutoff frequency and order of the Butterworth filter, and then filter the measured magnetic field data to remove environmental magnetic noise; S2. Further reduce the influence of background magnetic noise on the magnetic anomaly target signal by geomagnetic field reconstruction. The measured magnetic field is decomposed into two parts by a combination of a multinomial model and an autoregressive model: a deterministic trend term and a fluctuation term. S3. Based on the denoised magnetic measurement data from S2, construct the OBF, ME, and HMS features of the target to comprehensively characterize the target from multiple dimensions. S4. The p-norm is used as the detection metric after feature fusion:

[0007] in: These are the characteristics of magnetic anomalies (OBF characteristics, ME characteristics, and HHS characteristics). Let be a real number, and ; S5. Calculate the cumulative distribution function of the noise characteristics using kernel density estimation (KDE), determine the corresponding threshold based on the set false alarm rate (FAR), and output the detection results based on the threshold.

[0008] Furthermore, the specific method for decomposing the measured magnetic field into two parts as described in S2 is as follows: A polynomial model is used to fit the geomagnetic signal, and the long-term trend component is extracted; given a set of geomagnetic measurement data... ,in It is location data; the least squares method is used to fit a polynomial model to obtain the trend part of the signal. The residual is obtained by removing the trend component from the original signal.

[0009] An autoregressive model is used to fit the residuals, capturing local dependencies between spatial locations; for each location x i The neighborhood relationships are described by the adjacency matrix W; the autoregressive coefficients are estimated by fitting the maximum likelihood estimate. The spatial adjacency weight matrix W is used to extract the fluctuating magnetic anomaly signal using an autoregressive model.

[0010] Furthermore, the specific methods for constructing the OBF features, ME features, and HMS features of the target in S3 are as follows: S31. OBF Feature Construction: First, construct a suitable OBF set using orthogonalization methods; second, convert the original scalar measured magnetic signal... Projecting onto the OBF space yields the corresponding coefficients; finally, the energy function is considered as a probe feature; ultimately, the OBF space consists of three typical functions, which are expressed as:

[0011] in: , , , Assuming the target is stationary, with the target as the origin, the sensor's motion direction is the X-axis, and the vertical upward direction is the Z-axis. D is the abscissa of the sensor in this coordinate system, and R0 is the minimum distance between the sensor and the target during the motion. Based on the relationship between the parameters, the magnetic anomaly is represented in the form of orthogonal basis functions:

[0012] In the formula Vacuum permeability, For the target magnetic moment, These are coefficients related to the geomagnetic vector and the target magnetic moment vector; the OBF characteristic of magnetic anomaly detection is:

[0013] S32. ME feature construction, the definition of magnetic entropy features is expressed as follows:

[0014] in Represents magnetic field The probability density function is calculated; the steps include: obtaining the residual magnetic field and estimating its probability density function PDF based on the background magnetic field constructed by the autoregressive model; and calculating the corresponding magnetic entropy characteristics based on the estimated probability density function.

[0015] S33. HMS Feature Construction: Utilizing Ensemble Empirical Mode Decomposition (EEMD) for... Perform adaptive decomposition to obtain multiple intrinsic mode function terms. and a residual term :

[0016] Where: N represents the number of levels in the empirical mode decomposition.

[0017] After obtaining the intrinsic mode functions of the magnetic signal, for each Perform Hilbert transform

[0018]

[0019] get Analyzed signal ,as follows:

[0020] in: Represents the intrinsic mode function The corresponding instantaneous amplitude, Represents the intrinsic mode function The corresponding instantaneous phase.

[0021] Based on instantaneous phase Calculate the corresponding instantaneous frequency ,as follows:

[0022] The amplitude of a magnetic signal can be expressed as a function of time and instantaneous frequency, thus obtaining the Hilbert spectrum of the magnetic signal amplitude. :

[0023] According to Hilbert's spectral Calculate the marginal spectrum of the magnetic signal ,as follows:

[0024] Furthermore, the specific method for outputting the detection results in S5 is as follows: Let the characteristic information of the background noise be The probability density function is estimated using KDE. It can be represented as:

[0025] Where: N is the amount of feature data; h is the bandwidth parameter, which controls the width of the kernel function; K is the kernel function, usually a Gaussian kernel is chosen; The probability density function obtained from the kernel density estimate can be further used to calculate the cumulative distribution function; the cumulative distribution function It is the integral of the probability density function, that is:

[0026] Within the framework of KDE, CDF is obtained by numerically integrating the estimated PDF; therefore, the CDF estimated by KDE is expressed as:

[0027] Where: H is the cumulative distribution function of the kernel function; When detecting magnetic anomalies, the false alarm rate should be set as needed. Using the cumulative distribution function (CDF) and false alarm rate It is possible to determine a suitable threshold size T such that the CDF value satisfies:

[0028] When the characteristic value of the magnetic field signal is greater than T, it indicates that a magnetic anomaly target has been detected; otherwise, it indicates that no magnetic anomaly target has been detected.

[0029] The beneficial effects achieved by this invention are as follows: Enhanced noise immunity: By using Butterworth filter for denoising and combining polynomial and autoregressive models for geomagnetic field reconstruction, background magnetic noise and environmental interference can be effectively removed, signal amplitude integrity can be preserved, and the signal-to-noise ratio of magnetic anomaly signals can be improved, laying a high-quality data foundation for subsequent feature extraction.

[0030] Multi-dimensional and accurate target characterization: Construct OBF features, ME features and HMS features to comprehensively characterize the target from different dimensions such as magnetic field energy distribution, noise mode changes and time and frequency information, breaking through the limitations of traditional single feature detection and improving the ability to capture weak magnetic anomaly targets.

[0031] Improve the balance of feature fusion: Using the p-norm as the feature fusion detection index can balance the contribution weight of features of different dimensions, avoid false detections caused by excessively large single feature values, enhance the robustness of features to noise, and make the detection results more robust and reliable.

[0032] Adaptive threshold optimization detection: The cumulative distribution function of noise features is calculated using kernel density estimation (KDE), and the threshold is adaptively determined based on the set false alarm rate (FAR), realizing dynamic optimization of the detection threshold, effectively reducing the false alarm rate, and improving the accuracy and reliability of target detection in complex environments.

[0033] Expanding Applicable Scenarios: This method significantly improves detection performance in complex environments such as low signal-to-noise ratio and strong background interference through multi-feature fusion and noise suppression technology, expanding the applicable scenarios of magnetic anomaly detection technology, especially suitable for detection and search tasks of long-distance weak magnetic targets. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0035] Figure 1 This is a flowchart of the target detection method based on magnetic multi-feature fusion of the present invention. Detailed Implementation

[0036] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0038] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, if the word "and / or" appears throughout the text, it means including three parallel solutions; for example, "A and / or B" includes solution A, solution B, or a solution that simultaneously satisfies A and B. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0039] This invention proposes a target detection method based on magnetic multi-feature fusion, mainly including environmental magnetic noise denoising, background geomagnetic field denoising, target multi-magnetic feature construction, multi-feature fusion, adaptive threshold determination, and output of detection results based on the threshold. This method reduces the influence of the background magnetic field on the target's magnetic anomaly signal through magnetic data preprocessing. It characterizes the target from multiple dimensions by constructing orthonormalized basis function (OBF) features, minimum entropy (ME) features, and Hilbert marginal spectrum (HMS) features. A multi-feature fusion processing algorithm is used to achieve the balance of multi-dimensional feature fusion, improve the relative weight of weak signals, and enhance the robustness of features to noise.

[0040] like Figure 1 As shown, the target detection method specifically includes the following steps: S1. Based on the sampling rate of the magnetic measurement data, the cutoff frequency and order of the Butterworth filter are determined. Then, the measured magnetic field data is filtered to remove environmental magnetic noise. Specifically, to better remove the interference of background magnetic noise on the magnetic anomaly target signal, a Butterworth filter is used, which can effectively preserve the amplitude integrity of the signal and is suitable for applications sensitive to signal distortion. Simultaneously, it achieves good transition performance between the passband and stopband and is easy to implement. The Butterworth filter is a classic linear filter widely used in signal processing, primarily for frequency-selective signal processing, such as filtering noise or extracting signals within a specific frequency range.

[0041] S2. Based on the filtered magnetic field data, the influence of background magnetic noise on the magnetic anomaly target signal is further reduced through geomagnetic field reconstruction. A combination of a polynomial model and an autoregressive model is used to decompose the measured magnetic field into two parts: a deterministic trend term (polynomial trend) and a fluctuation term. The deterministic trend term is used to eliminate background geomagnetic field noise, while the fluctuation term is used to further extract magnetic anomaly data.

[0042] S3. Based on the denoised magnetic measurement data from S2, construct the OBF, ME, and HMS features of the target to comprehensively characterize the target from multiple dimensions.

[0043] S4. The p-norm is used as the detection metric after feature fusion:

[0044] in: These are the characteristics of magnetic anomalies (OBF characteristics, ME characteristics, and HMS characteristics). Let be a real number, and ; This enhances the robustness of features to noise, balances smaller and larger features, and achieves equilibrium in multi-dimensional feature fusion. Specifically, OBF features characterize magnetic anomalies through the target magnetic field signal, ME features characterize magnetic anomalies through changes in magnetic noise patterns, and HMS features characterize magnetic anomalies through the time-frequency information of the magnetic signal. This allows for characterization of magnetic anomaly targets through different dimensions, improving detection performance. Due to the different dimensions, feature information values ​​may vary; for example, smaller features may carry important magnetic anomaly information.

[0045] This fusion strategy characterizes magnetic anomaly signals from different dimensions, increasing the contribution of smaller feature information to detection indicators and decreasing the contribution of larger feature values, thereby achieving a balance in multi-dimensional feature fusion. It effectively reduces false positives caused by excessively large single feature information, thus making the detection results more robust.

[0046] S5. Addressing the threshold determination problem in S4, kernel density estimation (KDE) is used to calculate the cumulative distribution function of the noise characteristics. The corresponding threshold is determined based on the set false alarm rate (FAR), and the detection result is output according to the threshold. KDE is a non-parametric estimation method used to estimate an unknown probability density function (PDF).

[0047] Furthermore, the specific method for decomposing the measured magnetic field into two parts as described in S2 is as follows: A polynomial model is used to fit the geomagnetic signal, and the long-term trend component is extracted. Assume there is a set of geomagnetic measurement data. ,in This is location data. The least squares method can be used to fit a polynomial model to obtain the trend component of the signal. The deterministic trend component is removed from the original signal to obtain the residual (i.e., the detrended volatility term):

[0048] An autoregressive model is used to fit the residuals, capturing local dependencies between spatial locations. For each location x i The neighborhood relationships are described by the adjacency matrix W. The autoregressive coefficients are estimated through maximum likelihood estimation. And the spatial adjacency weight matrix W. Using an autoregressive model, magnetic anomaly signals are extracted from the fluctuation term.

[0049] Furthermore, the specific methods for constructing the OBF features, ME features, and HMS features of the target in S3 are as follows: S31. OBF Feature Construction: First, a suitable set of OBFs is constructed using orthogonalization methods. Second, the original scalar measured magnetic signal is... Projecting onto the OBF space yields the corresponding coefficients. Finally, the energy function is considered as a probe feature. Ultimately, the OBF space consists of three typical functions, which can be expressed as:

[0050] in: , , , Assuming the target is stationary, with the target as the origin, the sensor's motion direction is the X-axis, and the vertical upward direction is the Z-axis. D is the abscissa of the sensor in the coordinate system, and R0 is the minimum distance between the sensor and the target during the motion. Based on the relationship between the parameters, the magnetic anomaly is represented in the form of orthogonal basis functions:

[0051] In the formula Vacuum permeability, For the target magnetic moment, These are coefficients related to the geomagnetic vector and the target magnetic moment vector. The OBF characteristic of magnetic anomaly detection is:

[0052] S32. ME Feature Construction: The presence of magnetic targets alters the distribution of background magnetic field noise. Magnetic entropy can be used to detect changes in the distribution of magnetic noise, making it a feature applicable to the detection of weak magnetic anomalies. The definition of magnetic entropy features can be expressed as:

[0053] in Represents magnetic field The probability density function is calculated. Construction steps: Based on the background magnetic field constructed from the autoregressive model, obtain the residual magnetic field and estimate its probability density function (PDF); based on the estimated probability density function, calculate the corresponding ME features.

[0054] S33. HMS Feature Construction: Utilizing Ensemble Empirical Mode Decomposition (EEMD) for... Perform adaptive decomposition to obtain multiple intrinsic mode function terms. and a residual term :

[0055] Where: N represents the number of levels in the empirical mode decomposition.

[0056] After obtaining the intrinsic mode functions of the magnetic signal, for each Perform Hilbert transform

[0057]

[0058] get Analyzed signal ,as follows:

[0059] in: Represents the intrinsic mode function The corresponding instantaneous amplitude, Represents the intrinsic mode function The corresponding instantaneous phase.

[0060] Based on instantaneous phase Calculate the corresponding instantaneous frequency ,as follows:

[0061] The amplitude of a magnetic signal can be expressed as a function of time and instantaneous frequency, thus obtaining the Hilbert spectrum of the magnetic signal amplitude. :

[0062] According to Hilbert's spectral Calculate the marginal spectrum of the magnetic signal ,as follows:

[0063] Furthermore, the specific method for outputting the detection results in S5 is as follows: Assume the characteristic information of the background noise is The probability density function is estimated through kernel density estimation (KDE). It can be represented as:

[0064] Where: N is the amount of feature data; h is the bandwidth parameter, which controls the width of the kernel function; K is the kernel function, usually a Gaussian kernel is chosen.

[0065] The probability density function obtained from the kernel density estimate can be further used to calculate the cumulative distribution function (CDF). It is the integral of the probability density function, that is:

[0066] Within the framework of KDE, the cumulative distribution function (CDF) is obtained by numerically integrating the estimated probability density function (PDF). Therefore, the CDF estimated by KDE can be expressed as:

[0067] Where H is the cumulative distribution function of the kernel function.

[0068] When detecting magnetic anomalies, the false alarm rate should be set as needed. Using the cumulative distribution function (CDF) and false alarm rate It is possible to determine a suitable threshold size T such that the CDF value satisfies:

[0069] When the characteristic value of the magnetic field signal is greater than T, it indicates that a magnetic anomaly target has been detected; otherwise, it indicates that no magnetic anomaly target has been detected.

[0070] The above description is merely an optional embodiment of the present invention and does not limit the patent scope of the present invention. All equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A target detection method based on magnetic multi-feature fusion, characterized in that, Includes the following steps: S1. Based on the sampling rate of the magnetic measurement data, determine the cutoff frequency and order of the Butterworth filter, and then filter the measured magnetic field data to remove environmental magnetic noise; S2. Further reduce the influence of background magnetic noise on the magnetic anomaly target signal by geomagnetic field reconstruction. The measured magnetic field is decomposed into two parts by a combination of a multinomial model and an autoregressive model: a deterministic trend term and a fluctuation term. S3. Based on the denoised magnetic measurement data from S2, construct the OBF, ME, and HMS features of the target to comprehensively characterize the target from multiple dimensions. S4. The p-norm is used as the detection metric after feature fusion: in: These are characteristics of magnetic anomalies. Let be a real number, and ; S5. Calculate the cumulative distribution function of the noise characteristics using kernel density estimation, determine the corresponding threshold based on the set false alarm rate, and output the detection results based on the threshold. The specific methods for constructing the OBF, ME, and HMS features of the target in S3 are as follows: S31. OBF Feature Construction: First, construct a suitable OBF set using orthogonalization methods; second, convert the original scalar measured magnetic signal... Projecting onto the OBF space yields the corresponding coefficients; finally, the energy function is considered as a probe feature; ultimately, the OBF space consists of three typical functions, which are expressed as: in: , , , Assuming the target is stationary, with the target as the origin, the sensor's motion direction is the X-axis, and the vertical upward direction is the Z-axis. D is the abscissa of the sensor in this coordinate system, and R0 is the minimum distance between the sensor and the target during the motion. Based on the relationship between the parameters, the magnetic anomaly is represented in the form of orthogonal basis functions: In the formula Vacuum permeability, For the target magnetic moment, These are coefficients related to the geomagnetic vector and the target magnetic moment vector; the OBF characteristic of magnetic anomaly detection is: S32. ME feature construction, the definition of magnetic entropy features is expressed as follows: in Represents magnetic field The probability density function is calculated; the steps include: obtaining the residual magnetic field and estimating its probability density function PDF based on the background magnetic field constructed by the autoregressive model; and calculating the corresponding magnetic entropy characteristics based on the estimated probability density function. S33. HMS Feature Construction: Utilizing Ensemble Empirical Mode Decomposition to... Perform adaptive decomposition to obtain multiple intrinsic mode function terms. and a residual term : In the formula, N represents the number of levels in the empirical mode decomposition; After obtaining the intrinsic mode functions of the magnetic signal, for each Perform the Hilbert transform: get Analyzed signal ,as follows: in: Represents the intrinsic mode function The corresponding instantaneous amplitude, Represents the intrinsic mode function The corresponding instantaneous phase; Based on instantaneous phase Calculate the corresponding instantaneous frequency ,as follows: The amplitude of a magnetic signal can be expressed as a function of time and instantaneous frequency, thus obtaining the Hilbert spectrum of the magnetic signal amplitude. : According to Hilbert's spectral Calculate the marginal spectrum of the magnetic signal ,as follows: 。 2. The target detection method based on magnetic multi-feature fusion according to claim 1, characterized in that, The specific method for decomposing the measured magnetic field into two parts as described in S2 is as follows: A polynomial model is used to fit the geomagnetic signal, and the long-term trend component is extracted; given a set of geomagnetic measurement data... ,in It is location data; the least squares method is used to fit a polynomial model to obtain the trend part of the signal. The residual is obtained by removing the trend component from the original signal. An autoregressive model is used to fit the residuals, capturing local dependencies between spatial locations; for each location x i The neighborhood relationships are described by the adjacency matrix W; the autoregressive coefficients are estimated by fitting the maximum likelihood estimate. And the spatial adjacency weight matrix W; An autoregressive model was used to extract magnetic anomaly signals from the fluctuation term.

3. The target detection method based on magnetic multi-feature fusion according to claim 1, characterized in that, The specific method for outputting detection results in S5 is as follows: Let the characteristic information of the background noise be The probability density function is estimated through kernel density estimation. Represented as: Where: N is the amount of feature data; h is the bandwidth parameter, which controls the width of the kernel function; K is the kernel function, usually a Gaussian kernel is chosen; The probability density function obtained from the kernel density estimate can be further used to calculate the cumulative distribution function; the cumulative distribution function It is the integral of the probability density function, that is: Within the framework of KDE, the cumulative distribution function is obtained by numerically integrating the estimated probability density function; therefore, the CDF estimated by KDE is expressed as: Where: H is the cumulative distribution function of the kernel function; When detecting magnetic anomalies, the false alarm rate should be set as needed. Using the cumulative distribution function (CDF) and false alarm rate It is possible to determine a suitable threshold size T such that the CDF value satisfies: When the characteristic value of the magnetic field signal is greater than T, it indicates that a magnetic anomaly target has been detected; otherwise, it indicates that no magnetic anomaly target has been detected.

Images