CN112904434A

CN112904434A - Magnetic anomaly signal detection method based on parameter optimization stochastic resonance

Info

Publication number: CN112904434A
Application number: CN202011532926.7A
Authority: CN
Inventors: 王成栋; 彭根斋; 杨冬辰; 闫欢; 陈勇
Original assignee: University of Electronic Science and Technology of China; CETC 9 Research Institute
Current assignee: University of Electronic Science and Technology of China; CETC 9 Research Institute
Priority date: 2020-12-22
Filing date: 2020-12-22
Publication date: 2021-06-04
Anticipated expiration: 2040-12-22
Also published as: CN112904434B

Abstract

The invention discloses a magnetic abnormal signal detection method based on parameter optimization stochastic resonance. According to the characteristics of the magnetic dipole of the magnetic abnormal signal and the characteristics of the stochastic resonance system, the invention constructs the influence factor index, quantifies the output response of the stochastic resonance system through the index and the output standard deviation, and adopts an intelligent optimization algorithm to find the most Finally, the collected magnetic signal is used as the input to obtain the output signal of the stochastic resonance system, and then according to the output signal, it is judged whether there is an abnormal signal caused by the ferromagnetic target.

Description

Magnetic anomaly signal detection method based on parameter optimization stochastic resonance

Technical Field

The invention relates to the technologies of signal processing, weak signal detection, machine learning and the like, and belongs to the field of weak signal detection.

Background

The magnetic abnormal signal belongs to one of weak signals, the magnetic field intensity of a magnetic field in nature is generally about 4 ten thousand nT, and the magnetic field intensity of the magnetic abnormal signal is often only dozens of nT or even nT. Therefore, the signal-to-noise ratio of the magnetic anomaly signal is very low, and the detection difficulty is also very high. Magnetic anomaly signal detection is widely used in various fields. In the national defense field, the magnetic anomaly signal detection is used for identifying ships and submarines at ports and invisible units in the air, and the identification mode has the advantages that the identification mode is passive and is not easy to find according to the ferromagnetic characteristic of a target; in life, magnetic anomaly signal detection is used for identification of vehicles coming and going, and the like. Therefore, the magnetic anomaly signal detection has high practical application value and theoretical research significance.

At present, the common methods for detecting the magnetic abnormal signals are an OBF method and a minimum entropy method. The drawback of the OBF method is that it is only suitable for detecting magnetic anomaly signals under the background of white gaussian noise, and the natural magnetic field is closer to colored noise. The minimum entropy method has the disadvantage that the low signal-to-noise ratio signal identification accuracy is not high. The stochastic resonance method does not have the disadvantages of the above two methods. Therefore, the stochastic resonance has certain research value in the aspect of magnetic anomaly signal detection.

Disclosure of Invention

The invention aims to provide a parameter-optimized stochastic resonance magnetic anomaly signal detection method which can accurately detect whether a target signal contains a magnetic anomaly signal.

In order to achieve the above purpose, the invention provides the following technical scheme:

a parameter optimized stochastic resonance magnetic anomaly signal detection method comprises the following steps:

the method comprises the following steps: collecting magnetic signals;

step two: normalizing the collected magnetic signals;

step three: constructing a nonlinear Langevin equation for describing stochastic resonance and a potential well function equation for describing particle motion according to a model of a magnetic dipole theory;

step four: calculating an influence factor gamma_CKDetermining the parameters of the stochastic resonance system by using an intelligent optimization algorithm;

step five, according to the parameters obtained in the step four, the acquired magnetic signals are used as input, and the output signal y of the stochastic resonance system is calculated by using a four-order Runge Kutta method;

step six: adding sliding window to signal y, and taking root mean square value of data in window as stable output value y of window intermediate point_optCalculating y corresponding to all the sampling points_opt；

Step seven: based on y_optGenerating a large number of simulation samples by adopting the parameters obtained in the step four and using a Monte Carlo method, and determining an optimal threshold value according to a Neyman-Pearson criterion;

step eight: taking the magnetic signal to be detected as input, and obtaining the corresponding y after the processing of the fifth step and the sixth step_optIf there is a part y_optIf the magnetic anomaly exceeds the threshold value, judging that the magnetic anomaly exists at the corresponding moment; otherwise, the magnetic signal is considered to be absent of a magnetic anomaly caused by a ferromagnetic target.

Drawings

FIG. 1 shows the influence factor gamma of the process of the invention_CKA flow chart of a method of parameter optimization.

FIG. 2 is a flow chart of a parameter-optimized stochastic resonance algorithm.

Detailed Description

The following describes the technical solution of the present invention in detail by taking a stochastic resonance algorithm for parameter optimization as an example with reference to the accompanying drawings and the detailed description.

As shown in fig. 1 and 2, the method comprises the following specific steps:

the method comprises the following steps: collecting magnetic signals;

step two: normalizing the collected magnetic signals;

step six: applying a sliding window to the signal y and averaging the data in the windowRoot value as stable output value y of window middle point_optCalculating y corresponding to all the sampling points_opt；

As shown in fig. 1, the influence factor γ_CKIs calculated as follows, γ_CKThe kurtosis K and the peak factor C are jointly determined, the physical characteristic of whether the shock exists or not is detected according to the reaction of the kurtosis K to the shock characteristic of the vibration signal and the peak factor, and the magnetic abnormal signal belongs to the non-periodic shock signal, so that the magnetic abnormal signal and the non-periodic shock signal are combined to detect the existence of the magnetic abnormal signal in the target signal.

Wherein x_peakIs the peak value of the signal, x_rmsIs the signal root mean square value. Obtaining gamma by simulated annealing algorithm_CKWhen the maximum value is taken, the corresponding parameters of potential well function a, b, step length h and the like.

As shown in fig. 1, the flow of the intelligent algorithm simulated annealing algorithm is as follows: setting iteration times T and rate alpha, generating a standard energy function p (x) according to an initialized calculation objective function equation model, calculating the difference delta p between the standard energy function p (x) and the previous standard energy function p (x), if the delta p is less than or equal to 0, accepting p (x), if not less than 0, accepting p (x) according to the Metropolis criterion:

until reaching the iteration number, obtainingAnd (4) parameters. If the parameters do not meet the termination condition, slowly reducing the temperature, resetting the iteration times, and repeating the iteration until the maximum value gamma is obtained_CK. By maximum gamma_CKAnd solving the corresponding potential well function parameters a and b and the step length h of the Runge Kutta equation.

As shown in fig. 2, the signal preprocessing is followed by constructing a stochastic resonance system model, which is constructed from langevin's equations describing stochastic resonance:

x′(t)＝-V′(x)+s(t)+ξ(t)

x (t) is SR system response, s (t) is magnetic anomaly signal, ξ (t) is geomagnetic noise signal. V (x) is a nonlinear potential well function equation:

through the above two equations, the magnetic anomaly signal and the parameters of the stochastic resonance potential well function can be correlated.

As shown in fig. 2, the stochastic resonance system output can be solved by a fourth-order lattice tower method, wherein the formula of the fourth-order lattice tower method is as follows:

by solving the fourth order equation, the system output x (t) corresponding to the parameters of the potential well function a and b and the step length h can be obtained.

As shown in fig. 2, the stable output y of the system output_optThe output can be optimized to indicate the degree of signal fluctuation in the vicinity. The threshold q can be determined by the Neyman-Pearson criterion. y is_optThe calculation method of (2) is as follows:

y_opt＝MSE(x(n-N+1：n))

MSE represents the root mean square value of x between N and N-N +1, and the calculated y_optThe curve determines a threshold q according to the Neyman-Pearson criterion to determine whether the signal contains a magnetic anomaly signal. If there is a moiety y_optExceeding the thresholdIf so, judging that magnetic anomaly exists at the corresponding moment; otherwise, the magnetic signal is considered to be absent of a magnetic anomaly caused by a ferromagnetic target.

It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A magnetic anomaly signal detection method based on parameter optimization stochastic resonance is characterized by comprising the following steps:

the method comprises the following steps: collecting magnetic signals;

step two: normalizing the collected magnetic signals;

step eight: taking the magnetic signal to be detected as input, after the processing of the fifth step and the sixth step,get the corresponding y_optIf there is a part y_optIf the magnetic anomaly exceeds the threshold value, judging that the magnetic anomaly exists at the corresponding moment; otherwise, the magnetic signal is considered to be absent of a magnetic anomaly caused by a ferromagnetic target.

2. The method for detecting magnetic anomaly signals based on parametric optimization stochastic resonance as claimed in claim 1, wherein the influencing factor γ in step four_CKThe kurtosis K and the peak factor C are jointly determined, and parameters in the random resonance system are determined by using an intelligent optimization algorithm, wherein the parameters comprise optimal potential well function parameters a and b and a step size h.

3. The method for detecting magnetic anomaly signals based on parametric optimization stochastic resonance as claimed in claim 1, wherein the computing system output y in step six_optThe calculation formula of (2) is as follows:

y_opt＝MSE(x(n-N+1：n))

where MSE represents the root mean square value of x between N and N-N +1, and N represents the length of the time window, the window sliding on the time axis of the sampled data as N varies.

4. The magnetic anomaly signal detection method based on parameter-optimized stochastic resonance as claimed in claim 1, wherein the optimal threshold value described in step seven is determined by: the original signal samples are subjected to quantity expansion by using a Monte Carlo method, random noise is changed under the condition that parameters a and b of a potential well function and a step length h are kept unchanged, a large number of simulation samples are obtained, and then a threshold value of the samples under the condition that the false alarm rate is small is determined according to a Neyman-Pearson criterion.