A method for identifying specific targets in passive magnetic detection based on discrete Mayer wavelet

Abstract

The present invention discloses a method for identifying specific targets by passive magnetic detection based on discrete Meyer wavelet. The method processes high-frequency noise signals through a low-pass filter circuit and obtains digital magnetic signals by sampling. The PCA algorithm is used to obtain an adaptive scaling function, and the magnetic signal data of ferromagnetic armored targets in the database are reduced in dimension to k The corresponding scale function size is obtained by using the dimensional data to realize the fast discrete wavelet transform of the Meyer wavelet multi-scale function; the signal characteristics of the magnetic signal in different frequency bands are extracted, and the bilateral filtering decision index allocation algorithm is designed to obtain the decision index of the magnetic signal characteristics in different frequency bands; a multi-signal fusion model is established, firstly, the decision indexes of the single-axis magnetic signals in different frequency bands are accumulated under the bilateral filtering algorithm to obtain the single-axis decision index, and then the decision indexes of the three-axis magnetic signals are accumulated to obtain the fusion decision index of the magnetic signal of this system; a fast discrimination criterion based on the fixed threshold method is used to quickly determine whether there is a ferromagnetic armored target within the detection range.

Description

Technical Field

[0001] The invention belongs to target recognition technology, and in particular relates to a passive magnetic detection specific target recognition method based on discrete Mayer wavelet. Background Art

[0002] In actual battlefields, most of the weapons and ammunition are aimed at ferromagnetic targets such as helicopters, tanks, and armored vehicles. In addition, the battlefield environment is complex, and traditional laser fuzes are easily affected by the ground, trees, clouds, fog, and smoke, and cannot accurately identify ferromagnetic armored targets. Therefore, if ferromagnetic armored targets can be effectively detected and identified, the detection performance of laser fuzes will be greatly improved.

[0003] The earth itself is a huge magnetic field. The ferromagnetic armored targets on the battlefield have a high magnetic permeability, which will cause local geomagnetic field deflection, that is, the intensity and direction of the geomagnetic field near the ferromagnetic armored targets will be distorted. Based on this characteristic, ferromagnetic metal targets within a certain range around the fuze can be effectively detected and identified.

[0004] Magnetic detection is divided into strong magnetic detection and weak magnetic detection. Battlefield ferromagnetic target detection is a typical weak magnetic detection. Currently, weak magnetic detection methods can be divided into total magnetic field intensity detection method, magnetic field scalar gradient detection method, magnetic field vector detection method, magnetic field vector gradient detection method, magnetic field gradient tensor detection method, etc.

[0005] Sheinker A et al. proposed a method for detecting magnetic anomaly signals based on a fixed AR whitening filter combined with an orthogonal function OBF. However, this method is only applicable to magnetic anomaly signals with a power density of 1 / f. α In order to prevent signal distortion, the orthogonal function also needs whitening filtering. The method is complex and cannot achieve fast and efficient judgment.

[0006] Ginzburg et al. proposed a gradient signal processing method based on the Gram-Schmidt algorithm to construct the MAD (magnetic anomaly detector) signal decomposition in the orthogonal function space, and used five independent orthogonal functions to describe the magnetic anomaly signal. Directly analyzing the magnetic signal without preprocessing the signal, the data is complex and the calculation is large, so it is impossible to achieve fast and efficient judgment.

[0007] Wang Deli from Nanjing University of Science and Technology used a three-axis magnetic anomaly signal detection and processing circuit, including multi-stage magnetic signal amplification, low-pass filtering and precision full-wave rectification hardware circuit. The three target identification characteristic values ​​of the magnetic signal amplitude, local change slope and local delay time were extracted. When the characteristic values ​​of the magnetic anomaly signals of more than two axes reached the threshold, it was judged that the real target was detected. The full-wave rectification hardware circuit reversely outputs the negative signal in the magnetic signal into a positive signal, resulting in the loss of magnetic signal vector information. The extracted magnetic signal features are all time domain features, and the signal features cannot be analyzed from the frequency domain perspective, and similar time domain signals cannot be distinguished. Zhao Chimeng and others from Northwestern Polytechnical University used a magnetic detection fuze target recognition algorithm based on dynamic threshold target detection and fuzzy reasoning. The algorithm collects magnetic signals that meet the requirements through dynamic thresholds, uses wavelet threshold method and sliding average filtering method to denoise the target signal, extracts the peak value, valley value, magnetic field change frequency and magnetic gradient of the target signal to be measured, compares the real target characteristic parameters, and uses fuzzy reasoning algorithm to discriminate the target feature quantity. Dynamic threshold acquisition will lead to signal loss and the inability to collect complete magnetic signals. The sliding average filter method will lead to the loss of characteristic signals in some frequency bands. When the signal is incomplete and the characteristic signals in some frequency bands are lost, the extracted magnetic signal features will have errors and the recognition accuracy will be low. Summary of the invention

[0008] The object of the present invention is to provide a passive magnetic detection specific target recognition method based on Discrete Meyer Wavelet (DMey), which can accurately detect whether there is a ferromagnetic armored target within the detection range.

[0009] The technical solution to achieve the purpose of the present invention is: a method for identifying specific targets in passive magnetic detection based on discrete Mayer wavelet, characterized in that:

[0010] Step 1: The magnetic sensor is powered on to detect the ferromagnetic armored target. The magnetic sensor on the carrier outputs magnetic signals of the magnetic field strength in the three axes x, y, and z of the carrier coordinate system.

[0011] Step 2: After the three-axis magnetic signals are DC-isolated, they are amplified. The amplified three-axis magnetic signals are low-pass filtered to filter out high-frequency noise signals. The cutoff frequency f c Set according to the empirical value; pull up the three-axis magnetic signal after low-pass filtering, and output the three-axis sampling signals x(t), y(t) and z(t) with a reference of 1.65V and an amplitude of 0V to 3.3V;

[0012] Step 3: Perform AD sampling on the three-axis magnetic signals x(t), y(t) and z(t) to obtain digital signals x(n), y(n) and z(n) respectively. According to the characteristic signals of the three axes x, y and z corresponding to the ferromagnetic armored targets in the typical battlefield ferromagnetic armored target database, the PCA algorithm is used to perform dimensionality reduction processing on the characteristic signals of the three axes of the ferromagnetic armored targets, and a k-dimensional data matrix after dimensionality reduction is obtained. and

[0013] Step 4: k-dimensional data matrix after dimensionality reduction optimization and Correspondingly obtain the characteristic signals of k frequency bands corresponding to the x, y, and z axes, and set the scale functions of the k discrete Mayer wavelet transforms corresponding to the x, y, and z axes according to the characteristic signals of the k frequency bands of each axis; perform discrete Mayer wavelet transforms of k multi-scale functions on the AD sampling signals x(n), y(n), and z(n) of the ferromagnetic armor target to obtain the matrix X after the discrete wavelet transformation of the ferromagnetic armor target MDey , Y MDey and Z MDey ;

[0014] Step 5: Matrix X after discrete wavelet transformation of ferromagnetic armor target MDey , Y MDey and Z MDey Perform bilateral filtering to obtain weight index processing and obtain the decision weight index of each frequency band of the x, y, and z axes and

[0015] Step 6: Decision weight index of each frequency band according to x, y, and z axes and The multi-signal fusion model is used to sum the weight indexes of the characteristic signals of different frequency bands, and the total decision indexes of the x-axis, y-axis, and z-axis are obtained as e x 、e y 、e z , then e x 、e y 、e z Perform a sum operation to obtain the total decision index e all ;

[0016] Step 7: The total decision index e all With the empirical threshold U tv Compare and determine whether there is a ferromagnetic armor target, as follows:

[0017] When e all≥U tv When , a ferromagnetic armored target appears within the effective detection range; when e all <U tv , it means that no ferromagnetic armored target appears within the effective detection range.

[0018] Compared with the prior art, the present invention has the following significant advantages:

[0019] (1) The use of low-pass filter circuits to process high-frequency noise signals can reduce the wavelet decomposition level and the amount of data that needs to be processed during discrete wavelet transform. According to the signal characteristics of different ferromagnetic armored targets in the typical battlefield ferromagnetic armored target database, the PCA (Principal Component Analysis) algorithm is used to obtain an adaptive scaling function, which greatly reduces the number of scaling functions during discrete Meyer wavelet decomposition, thereby realizing fast digital magnetic signal data processing;

[0020] (2) The use of discrete Mayer wavelet function can realize fast discrete wavelet transform, which is faster than traditional wavelet transform operation and can distinguish similar time domain signals from the frequency domain perspective;

[0021] (3) The feature proportion function and feature discrimination function are used as the weight allocation criteria of the bilateral filtering algorithm, which respectively considers the characteristics of ferromagnetic armor targets in the time domain and frequency domain, making the discrimination of ferromagnetic armor targets more accurate;

[0022] (4) The use of low-pass filtering circuits to process high-frequency noise signals, PCA algorithm to perform data dimension reduction filtering, and discrete Mayer wavelet to distinguish ferromagnetic armor target signals and noise signals in the time domain and frequency domain can effectively reduce noise interference, improve the signal-to-noise ratio and the maximum detection distance. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The present invention is a flow chart of a method for identifying specific targets in passive magnetic detection based on discrete Mayer wavelet.

[0024] Figure 2 1 is a processing flow chart of the digital magnetic signals of each axis of the present invention (taking the x-axis as an example, the processing steps of the y-axis, the z-axis and the x-axis are the same). DETAILED DESCRIPTION

[0025] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0026] Combination Figure 1 and Figure 2 The method for identifying specific targets by passive magnetic detection based on discrete Mayer wavelet of the present invention comprises the following steps:

[0027] Step 1: The magnetic sensor is powered on to detect ferromagnetic armored targets. The magnetic sensor on the carrier outputs magnetic signals of the magnetic field strength in the three-axis directions of the carrier coordinate system x, y, and z.

[0028] Step 2: After the three-axis magnetic signals are DC-isolated, they are amplified. The amplified three-axis magnetic signals are low-pass filtered to filter out high-frequency noise signals. The cutoff frequency f c The three-axis magnetic signal after low-pass filtering is pulled up to output the three-axis sampling signals x(t), y(t) and z(t) with a reference of 1.65V and an amplitude of 0V to 3.3V.

[0029] Using low-pass filtering circuits to process high-frequency noise signals can reduce the wavelet decomposition level and the amount of data that needs to be processed during discrete wavelet transform, thereby increasing the speed of digital signal processing.

[0030] Among them, the calculation formula of the magnification G is:

[0031] G=3.3 / Out max

[0032] 3.3 is the supply voltage of the rail-to-rail amplifier circuit, Out max The maximum amplitude of the magnetic signal output by the magnetic sensor.

[0033] Step 3: Perform AD sampling on the three-axis magnetic signals x(t), y(t) and z(t) to obtain digital signals x(n), y(n) and z(n) respectively. According to the characteristic signals of the three axes x, y and z corresponding to the ferromagnetic armored targets in the typical battlefield ferromagnetic armored target database, the PCA algorithm is used to perform dimensionality reduction processing on the characteristic signals of the three axes of the ferromagnetic armored targets, and a k-dimensional data matrix after dimensionality reduction is obtained. and

[0034] The PCA algorithm is used to obtain an adaptive scaling function, which greatly reduces the number of scaling functions during discrete Mayer wavelet decomposition, thereby enabling fast digital magnetic signal data processing.

[0035] Taking the x-axis as an example, the data matrix X of the x-axis in the typical battlefield ferromagnetic armor target database is target As shown below:

[0036]

[0037] X target The row elements represent m data sampling time points, and the column elements represent the time from 0 to f at a sampling time point. c The n frequencies in the element x nm Represents the amplitude of the nth frequency at the mth time point.

[0038] Solving for X target The covariance matrix C of is:

[0039]

[0040] Diagonalize the covariance matrix C and solve for the eigenvector ξ of C i , and the eigenvector ξ i Sort the corresponding eigenvalues ​​from large to small and get the eigenvector matrix P:

[0041] P=[ξ1,…,ξ n ] T

[0042] Select the matrix consisting of the first k rows of the eigenvector matrix P and multiply it by the data matrix X on the x-axis target , and get a k (k<n) dimensional matrix

[0043]

[0044] where λ i is the eigenvector ξ of C after reordering from largest to smallest i The corresponding eigenvalues, λ1≥λ2≥…≥λ k …>>λ n .

[0045] The k-dimensional data matrix after dimensionality reduction corresponding to the y-axis and z-axis The solution process is the same as

[0046] Step 4: The k-dimensional data matrix after dimensionality reduction optimization obtained in step 3 and Obtain the characteristic signals of the k frequency bands corresponding to the x, y, and z axes, and set the scale functions of the k discrete Mayer wavelet transforms corresponding to the x, y, and z axes according to the characteristic signals of the k frequency bands of each axis. Perform the discrete Mayer wavelet transform of k multi-scale functions on the AD sampling signals x(n), y(n), and z(n) of the ferromagnetic armor target to obtain the matrix X after the discrete wavelet transformation of the ferromagnetic armor target. MDey , Y MDey and Z MDey .

[0047] The use of discrete Mayer wavelet function can realize fast discrete wavelet transform. Compared with traditional wavelet transform operation, it has faster calculation speed and can distinguish similar time domain signals from the frequency domain perspective.

[0048] Taking the x-axis as an example, the sampling signal x(n) of the ferromagnetic armor target is transformed by the discrete Mayer wavelet transform of k multi-scale functions to obtain the matrix X after the discrete wavelet transformation of the ferromagnetic armor target. MDey :

[0049]

[0050]

[0051] ω i ∈{ω1,…,ω k}

[0052] Ψ M (ω i ) is the discrete Mayer wavelet function; ω1~ω k is the frequency of the characteristic signal of the k frequency bands after dimension reduction corresponding to the ferromagnetic armor target, v(x) is the auxiliary function for constructing discrete Mayer wavelet, and j is the complex number symbol;

[0053]

[0054] The matrix X after discrete wavelet transformation of ferromagnetic armor target MDey The k-th row represents the amplitude of the frequency signal corresponding to the discrete Mayer wavelet transform of the AD sampling signal x(n) on the x-axis under the k-th scaling function at each sampling time.

[0055] The matrix Y after the discrete wavelet changes corresponding to the y-axis and z-axis MDey , Z MDey With X MDey The solution method is the same.

[0056] Among them, the typical battlefield ferromagnetic armor target database includes target magnetic signal data of various ferromagnetic armors (for example, tanks, infantry fighting vehicles, radar vehicles, armored reconnaissance vehicles and transport vehicles, etc.), as well as noise magnetic signal data under environments such as vibration of detection device carriers (for example, drones, helicopters, etc.), interference fields of detection device carriers and geomagnetic background fields.

[0057] Step 5: Matrix X after discrete wavelet transformation of ferromagnetic armor target MDey , Y MDey and Z MDey , perform bilateral filtering to obtain weight index processing to obtain the decision weight index of each frequency band of the x, y, and z axes and

[0058] The feature proportion function and feature discrimination function are used as the weight distribution criteria of the bilateral filtering algorithm, and the characteristics of ferromagnetic armor targets in the time domain and frequency domain are considered respectively, which makes the discrimination of ferromagnetic armor targets more accurate.

[0059] Taking the x-axis as an example, the matrix X after discrete wavelet transformation MDey Obtain weight index through bilateral filtering:

[0060]

[0061] in is the weight index of the characteristic signal of the ith frequency band (0≤i≤k) of the x-axis magnetic signal; I(x i ) is the amplitude of the characteristic signal of the i-th frequency band of the input magnetic signal; T(x i ) is the characteristic signal duration of the i-th frequency band of the input magnetic signal; T i (x) is the characteristic signal duration of the i-th frequency band corresponding to the input magnetic signal of a specific ferromagnetic armor target in the typical battlefield ferromagnetic armor target database; i is the characteristic signal of the i-th frequency band of the input magnetic signal; W p is the regularization factor; Ω is the characteristic signal domain of the input magnetic signal; x is the characteristic information of a specific ferromagnetic armored target in the typical battlefield ferromagnetic armored target database.

[0062] f r (||T(x i )-T i (x)||) is the feature proportion function. According to the different feature proportions of the characteristic signals of different frequency bands in the entire ferromagnetic armor target magnetic signal, the weight of the proportion is assigned. The larger the feature proportion, the greater the assigned weight. The weight index distribution expression is as follows:

[0063]

[0064] x′ i is the characteristic information of the i-th frequency band of a specific ferromagnetic armor target in the typical battlefield ferromagnetic armor target database; sign(x′ i ) is x′ i The sign function; sign(x i ) is x′ i The symbol function of .

[0065] g s (||x i-x||) is the feature discrimination function. According to the inherent magnetic characteristic signal of the ferromagnetic armor target, this inherent and unchanging feature can distinguish the magnetic signal of this frequency band from the noise or similar interference signals in the time domain described in the typical battlefield ferromagnetic armor target database. Therefore, the signal of this frequency band is assigned a larger weight. The weight index distribution expression is as follows:

[0066]

[0067] x noise is the characteristic information of noise in the database of typical battlefield ferromagnetic armor targets; f(x i ) is the judgment function of the discrimination degree between the ith frequency band signal of the ferromagnetic armor target and the noise in the database of typical battlefield ferromagnetic armor targets.

[0068] Decision weight index of each frequency band on the y and z axes and The solution method is the same.

[0069] Step 6: Decision weight index of each frequency band of x, y, and z axes obtained in step 5 and The multi-signal fusion model is used to sum the weight indexes of the characteristic signals of different frequency bands, and the total decision indexes of the x-axis, y-axis, and z-axis are obtained as e x 、e y 、e z , then e x 、e y 、e z Perform a sum operation to obtain the total decision index e all .

[0070] Among them, the multi-signal fusion model expression is as follows:

[0071]

[0072] Step 7: Obtain the total decision index e from step 6 all With the empirical threshold U tv Compare and determine whether there is a ferromagnetic armor target, as follows:

[0073] When e all ≥U tv When , a ferromagnetic armored target appears within the effective detection range; when e all <U tv , it means that no ferromagnetic armored target appears within the effective detection range.

[0074] Example 1

[0075] Taking the detection of tanks as an example, the three-axis magnetic sensor outputs magnetic signals, which are amplified, filtered, and pulled up before AD sampling, converting the analog signals into digital signals to obtain sampling signals x(n), y(n), and z(n). Based on the magnetic signal data of tanks without interference in the typical battlefield ferromagnetic armor target database, the PCA adaptive algorithm is used to reduce the dimension of the data to obtain a k-dimensional data matrix. and (accounting for 80% of the tank magnetic signal data information). k-dimensional data matrix and The corresponding frequency band is converted into the scale function of discrete Mayer wavelet transform, and the input sampling signals x(n), y(n) and z(n) are processed by multi-window discrete Mayer wavelet transform to obtain the matrix X after discrete Mayer wavelet transform MDey , Y MDey and Z MDey , transform the matrix X after discrete Mayer wavelet transformation MDey , Y MDey and Z MDey The data of ferromagnetic armor targets and noise in the typical battlefield ferromagnetic armor target database are compared, and the weight index is obtained through bilateral filtering. The weight index of each frequency band and each axis is obtained through a multi-signal fusion model to obtain the total decision index e all , and finally the total judgment index e all With the empirical threshold U tv By comparison, it is determined whether a tank appears within the effective detection range.

Claims

1. A method for identifying specific targets in passive magnetic detection based on discrete Mayer wavelet, characterized in that: Step 1: The magnetic sensor is powered on to detect the ferromagnetic armored target. The magnetic sensor on the carrier outputs magnetic signals of the magnetic field strength in the three-axis directions of the carrier coordinate system x, y, and z; Step 2: After the three-axis magnetic signals are DC-isolated, they are amplified. The amplified three-axis magnetic signals are low-pass filtered to filter out high-frequency noise signals. The cutoff frequency f c Set according to the empirical value; pull up the three-axis magnetic signal after low-pass filtering, and output the three-axis sampling signals x(t), y(t) and z(t) with a reference of 1.65V and an amplitude of 0V to 3.3V; Step 3: Perform AD sampling on the three-axis magnetic signals x(t), y(t) and z(t) to obtain digital signals x(n), Y(n) and z(n) respectively. According to the characteristic signals of the three axes x, y and z corresponding to the ferromagnetic armored targets in the typical battlefield ferromagnetic armored target database, the PCA algorithm is used to reduce the dimension of the characteristic signals of the three axes of the ferromagnetic armored targets to obtain a k-dimensional data matrix after dimension reduction. and Step 4: k-dimensional data matrix after dimensionality reduction optimization and Correspondingly obtain the characteristic signals of k frequency bands corresponding to the x, y, and z axes, and set the scale functions of the k discrete Mayer wavelet transforms corresponding to the x, y, and z axes according to the characteristic signals of the k frequency bands of each axis; perform discrete Mayer wavelet transforms of k multi-scale functions on the AD sampling signals x(n), y(n), and z(n) of the ferromagnetic armor target to obtain the matrix X after the discrete wavelet transformation of the ferromagnetic armor target MDey , Y MDey and Z MDey ; Step 5: Matrix X after discrete wavelet transformation of ferromagnetic armor target MDey , Y MDey and Z MDey Perform bilateral filtering to obtain weight index processing and obtain the decision weight index of each frequency band of the x, y, and z axes and Step 6: Decision weight index of each frequency band according to x, y, and z axes and The multi-signal fusion model is used to sum the weight indexes of the characteristic signals of different frequency bands, and the total decision indexes of the x-axis, y-axis, and z-axis are obtained as e x 、e y 、e z , then e x 、e y 、e z Perform sum operation to obtain the total decision index e all ; Step 7: The total decision index e all With the empirical threshold U tυ Compare and determine whether there is a ferromagnetic armor target, as follows: When e all ≥U tυ When , a ferromagnetic armored target appears within the effective detection range; when e all <U tυ , it means that no ferromagnetic armored target appears within the effective detection range.

2. The method for identifying specific targets by passive magnetic detection based on discrete Mayer wavelet according to claim 1 is characterized in that: In step 2, the three-axis magnetic signals are amplified after being DC-isolated. The calculation formula of the amplification factor G is: G=3.3 / Out max 3.3 is the supply voltage of the rail-to-rail amplifier circuit, Out max The maximum amplitude of the magnetic signal output by the magnetic sensor.

3. The method for identifying specific targets by passive magnetic detection based on discrete Meyer wavelet according to claim 1, characterized in that: In step 3, AD sampling is performed on x(t) to obtain a digital signal x(n). According to the characteristic signal of the x-axis corresponding to the ferromagnetic armor target in the typical battlefield ferromagnetic armor target database, the PCA algorithm is used to reduce the dimension of the characteristic signal of the ferromagnetic armor target to obtain a k-dimensional data matrix after dimension reduction. The details are as follows: The data matrix X of the x-axis in the database of typical battlefield ferromagnetic armor targets target As shown below: X target The row elements represent m data sampling time points, and the column elements represent the time from 0 to f at a sampling time point. c The n frequencies in the element x nm Represents the amplitude of the nth frequency at the mth time point; Solving for X target The covariance matrix C of is: Diagonalize the covariance matrix C and solve for the eigenvector of C And the feature vector Sort the corresponding eigenvalues ​​from large to small and get the eigenvector matrix P: Select the matrix consisting of the first k rows of the eigenvector matrix P and multiply it by the data matrix X on the x-axis target , where k < n, we get a k-dimensional matrix λ i is the eigenvector of C after reordering from largest to smallest The corresponding eigenvalues, λ1≥λ2≥…≥λ k …>>λ n .

4. The method for identifying specific targets by passive magnetic detection based on discrete Mayer wavelet according to claim 3 is characterized in that: In step 3, the y-axis and z-axis correspond to the k-dimensional data matrix after dimensionality reduction The solution process is the same as 5. The method for identifying specific targets by passive magnetic detection based on discrete Meyer wavelet according to claim 4 is characterized in that: In step 4, according to the k-dimensional data matrix after dimensionality reduction optimization Acquire the characteristic signals of the k frequency bands corresponding to the x-axis, and set the scale functions of the k discrete Mayer wavelet transforms corresponding to the x-axis according to the characteristic signals of the k frequency bands of the x-axis; perform the discrete Mayer wavelet transform of k multi-scale functions on the AD sampling signal x(n) of the ferromagnetic armor target to obtain the matrix X after the discrete wavelet transformation of the ferromagnetic armor target MDey , as follows: After the sampling signal x(n) of the ferromagnetic armor target is transformed by the discrete Mayer wavelet transform of k multi-scale functions, the discrete wavelet transformed matrix X of the ferromagnetic armor target is obtained. MDey : oh i ∈{ω1,…,ω k } Ψ M (ω i ) is the discrete Mayer wavelet function; ω1~ω k is the frequency of the characteristic signal of the k frequency bands after dimension reduction corresponding to the ferromagnetic armor target, v(x) is the auxiliary function for constructing discrete Mayer wavelet, and j is the complex number symbol; The matrix X after discrete wavelet transformation of ferromagnetic armor target MDey The k-th row represents the amplitude of the frequency signal corresponding to the discrete Mayer wavelet transform of the AD sampling signal x(n) on the x-axis under the k-th scaling function at each sampling time.

6. The method for identifying specific targets by passive magnetic detection based on discrete Meyer wavelet according to claim 5, characterized in that: The matrix Y after the discrete wavelet changes corresponding to the y-axis and z-axis MDey , Z MDey With X MDey The solution method is the same.

7. The method for identifying specific targets by passive magnetic detection based on discrete Mayer wavelet according to claim 5, characterized in that: The typical battlefield ferromagnetic armor target database includes magnetic signal data of various ferromagnetic armor targets, as well as the vibration of the detection device carrier, the interference field of the detection device carrier and the noise magnetic signal data under the geomagnetic background field environment.

8. The method for identifying specific targets by passive magnetic detection based on discrete Meyer wavelet according to claim 6, characterized in that: In step 5, the matrix X after discrete wavelet transformation of the ferromagnetic armor target is MDey Perform bilateral filtering to obtain weight index processing to obtain the decision weight index of each frequency band on the x-axis The details are as follows: The matrix X after discrete wavelet transformation MDey Obtain weight index through bilateral filtering: in is the weight index of the characteristic signal of the ith frequency band of the x-axis magnetic signal, 0≤i≤k; I(x i ) is the amplitude of the characteristic signal of the i-th frequency band of the input magnetic signal; T(x i ) is the characteristic signal duration of the i-th frequency band of the input magnetic signal; T i (x) is the characteristic signal duration of the i-th frequency band corresponding to the input magnetic signal of a specific ferromagnetic armor target in the typical battlefield ferromagnetic armor target database; i is the characteristic signal of the i-th frequency band of the input magnetic signal; W p is the regularization factor; Ω is the characteristic signal domain of the input magnetic signal; x is the characteristic information of a specific ferromagnetic armored target in the database of typical battlefield ferromagnetic armored targets; f r (||T(x i )-T i (x)||) is the feature proportion function. According to the different feature proportions of the characteristic signals of different frequency bands in the entire ferromagnetic armor target magnetic signal, the weight of the proportion is assigned. The larger the feature proportion, the greater the assigned weight. The weight index distribution expression is as follows: x′ i is the characteristic information of the i-th frequency band of a specific ferromagnetic armor target in the typical battlefield ferromagnetic armor target database; sign(x′ i ) is x′ i The sign function; sign(x i ) is x′ i The symbolic function of g s (||x i -x||) is the feature discrimination function. According to the inherent magnetic characteristic signal of the ferromagnetic armor target, this inherent and unchanging feature can distinguish the magnetic signal of this frequency band from the noise or similar interference signals in the time domain described in the typical battlefield ferromagnetic armor target database. Therefore, the signal of this frequency band is assigned a larger weight. The weight index distribution expression is as follows: x noise is the characteristic information of noise in the database of typical battlefield ferromagnetic armor targets; f(x i ) is the judgment function of the discrimination degree between the ith frequency band signal of the ferromagnetic armor target and the noise in the database of typical battlefield ferromagnetic armor targets.

9. The method for identifying specific targets by passive magnetic detection based on discrete Mayer wavelet according to claim 8, characterized in that: Decision weight index of each frequency band on the y and z axes and The solution method is the same.

10. The method for identifying specific targets by passive magnetic detection based on discrete Mayer wavelet according to claim 8, characterized in that: In step 6, the multi-signal fusion model expression is as follows:

Images