A composite target detection method

Abstract

The invention provides a target composite detection method, belonging to the technical field of target remote detection. The specific process of the method is as follows: firstly, a magnetic measurement system based on an unmanned aerial vehicle is used to carry out a large-scale full-measurement area test, and a characteristic vector x of the target is inverted by using a differential evolution algorithm using the measurement result of a magnetic field sensor; secondly, based on the position prior information in the characteristic vector x, local vehicle-mounted electromagnetic detection is carried out to collect electromagnetic data of a local small area, and the Levenberg-Marquardt algorithm is used according to the electromagnetic data, and the characteristic vector x is used as an inversion initial value to obtain a characteristic vector y; finally, the characteristic vector x and the characteristic vector y are fused to obtain the position, shape and posture characteristics of the target.

Description

Technical Field

[0001] The invention belongs to the technical field of target long-range detection, and in particular relates to a target composite detection method. Background Art

[0002] The magnetic measurement system based on drones has obvious advantages in achieving rapid measurement of large-scale measurement areas and detection efficiency, but it can only measure ferromagnetic targets and cannot measure non-magnetic metal targets, resulting in missed detection. At the same time, the inversion results based on magnetic field data can only provide the position and magnetic moment information of the target, and cannot obtain the posture or shape characteristics of the target, and cannot achieve target recognition and classification.

[0003] Based on the time domain electromagnetic detection method, all metal targets can be measured, and the target's posture, shape, material and other characteristics can be obtained according to the attenuation curve of the voltage signal at the receiving end. However, it has the problems of long processing time and short detection distance. Summary of the invention

[0004] The object of the present invention is to provide a target composite detection method, which can improve detection efficiency and achieve rapid target positioning and identification.

[0005] The technical solution for implementing the present invention is as follows:

[0006] A target composite detection method, the specific process is:

[0007] Step 1: Use magnetic method to determine the location of the target;

[0008] Suppose the eigenvector x=(x0,y0,z0,m x ,m y ,m z ) characterizes the magnetic dipole model;

[0009] The magnetic field sensor carried by the magnetic measurement system is used to measure the total magnetic field signal in the space, and the total magnetic field signal is preprocessed. Then, the magnetic anomaly signal ΔT of the target production is extracted from the measured total magnetic field signal. The magnetic anomaly signal can be expressed as:

[0010]

[0011] Where m=(m x ,m y ,m z ) is the magnetic moment of the target, r = (x-x0, y-y0, z-z0) is the distance vector, (x, y, z) is the position of the sensor, (x0, y0, z0) is the position of the target, B earth is the local geomagnetic field vector, where r is the modulus of r, μ0 is the vacuum permeability;

[0012] Using multiple observation points (x i ,y i ,z i ) at the magnetic anomaly data ΔT i , establish the objective function as shown in formula (2), and use the differential evolution algorithm to estimate the target's feature vector x, and use the differential evolution algorithm to estimate the target's feature vector x,

[0013]

[0014] Step 2: Determine the position of the target based on the magnetic method, and use the electromagnetic method to obtain the position, shape and posture characteristics of the target;

[0015] Assume the eigenvector y of electromagnetic data inversion:

[0016]

[0017] Among them, (x0, y0, z0) is the position of the target, are the yaw and pitch angles of the target, (k i ,β i ,γ i )i=1,2,3 are the three-dimensional magnetic dipole parameters;

[0018] Based on the target location determined in step 1, local electromagnetic detection is carried out to obtain multiple observation points (x i ,y i ,z i ), the induced voltage signal d of the receiving coil obsi ;

[0019] Using multiple observation points (x i ,y i ,z i ) receiving coil’s induced voltage signal d obsi , establish the objective function as shown in formula (5), take the position information in the eigenvector x as the initial value of the position information in the electromagnetic inversion eigenvector y, and use the local optimization algorithm Levenberg-Marquardt algorithm to estimate the target eigenvector y;

[0020]

[0021] Step 3: Fuse the feature vector x and the feature vector y to obtain the position, shape and posture features of the target.

[0022] Furthermore, the magnetic measurement system described in the present invention is carried on an unmanned aerial vehicle.

[0023] Furthermore, the preprocessing of the total magnetic field signal described in the present invention includes: filtering the total magnetic field signal using a low-pass filter, removing the DC component, and finally compensating for the interference magnetic field of the drone platform.

[0024] Furthermore, the local electromagnetic detection described in the present invention is achieved by a vehicle-mounted magnetic detection system.

[0025] Furthermore, the present invention utilizes more than 6 observation points (x i ,y i ,z i ) at the magnetic anomaly data ΔT i Establish the objective function.

[0026] Furthermore, the present invention utilizes more than 9 observation points (x i ,y i ,z i ) receiving coil’s induced voltage signal d obsi Establish the objective function.

[0027] Furthermore, after the characteristic vector y is calculated, the intrinsic dipole response curve L is calculated according to formula (4): i , according to the intrinsic dipole response curve L i , based on the response curve L i Get the shape of the target

[0028]

[0029] Furthermore, the present invention adopts a Bayesian estimation method to fuse the feature vector x and the feature vector y to obtain the position, shape and posture characteristics of the target.

[0030] Beneficial Effects

[0031] The target detection method combining magnetic method and electromagnetic method of the present invention uses magnetic method to quickly detect a large area, and uses electromagnetic method to perform accurate measurement based on the detection result of magnetic method. The present invention can quickly realize the detection of the target.

[0032] The present invention utilizes a target detection method combining magnetic and electromagnetic methods and a constructed objective function to quickly iterate and find the optimal feature vector, thereby accurately obtaining the position, shape and posture characteristics of the target. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0034] Figure 1 It is a processing flow chart of the combined magnetic method and electromagnetic method;

[0035] Figure 2 There are three targets to be tested (marked as 4#, 5#, 6#);

[0036] Figure 3 The magnetic survey data of the UAV (unit: nT); (a) magnetic field in the survey area, (b) magnetic anomaly generated by target 4#, (c) magnetic anomaly generated by target 5#, (d) magnetic anomaly generated by target 6#;

[0037] Figure 4 is the induced voltage of the receiving coil in the vehicle-mounted electromagnetic detection system; (a) the induced voltage of target 4#, (b) the induced voltage of target 5#, (c) the induced voltage of target 6#;

[0038] Figure 5 Dipole eigenresponse curves obtained by inverting electromagnetic data; (a) 4# target, (b) 5# target, (c) 6# target. DETAILED DESCRIPTION

[0039] The target positioning algorithm provided by the present invention is described below with reference to the accompanying drawings.

[0040] The design concept of the present invention is: Figure 1 As shown in the figure, firstly, the magnetic survey adopts the magnetic survey system based on drone to carry out large-scale test of the entire survey area, and uses the measurement results of the magnetic field sensor to invert the characteristic vector x of the target using the differential evolution (DE) algorithm; secondly, based on the position prior information in the characteristic vector x, local vehicle-mounted electromagnetic detection is carried out to collect electromagnetic data of a small local area, and the Levenberg-Marquardt algorithm is used based on the electromagnetic data. At the same time, the characteristic vector x is used as the inversion initial value to obtain the characteristic vector y. Finally, the characteristic vector x and the characteristic vector y are fused to obtain the position, shape and posture characteristics of the target.

[0041] The present application embodiment provides a target composite detection method, the specific process is as follows:

[0042] Step 1: Use magnetic method to determine the location of the target;

[0043] The target is approximated by the magnetic dipole model. In the rectangular coordinate system (northeast coordinate system), a magnetic dipole model can be represented by a characteristic vector x = (x0, y0, z0, m x ,m y ,m z ), the first three parameters are the position of the target, and the last three parameters are the magnetic moment of the target. )

[0044] The magnetic field sensor carried by the UAV magnetic measurement system is used to measure the total magnetic field signal in space. The total magnetic field signal is filtered using a low-pass filter, and then the DC component is removed. Finally, the interference magnetic field of the UAV platform is compensated, and the magnetic anomaly signal ΔT of the target production is extracted from the measured total magnetic field signal. The magnetic anomaly signal can be expressed as:

[0045]

[0046] Where m=(m x ,m y ,m z ) is the magnetic moment of the target, corresponding to the last three parameters of the eigenvector x; (x, y, z) is the position of the sensor, (x0, y0, z0) is the position of the target, r = (x-x0, y-y0, z-z0) is the distance vector, B earth is the local geomagnetic field vector, where r is the modulus of r and μ0 is the magnetic permeability of vacuum.

[0047] Using multiple observation points (x i ,y i ,z i ) at the magnetic anomaly data ΔT i (i=1,2,...N), N>6, establish the objective function as shown in formula (2), and use the differential evolution algorithm to estimate the characteristic vector x of the target, that is, by setting the initial value of x and the conditions for the end of iteration, the characteristic vector x of the target can be iteratively calculated based on formula (2);

[0048]

[0049] Step 2: Determine the position of the target based on the magnetic method, and use the electromagnetic method to obtain the position, shape and posture characteristics of the target.

[0050] In electromagnetic data processing, when the distance between the sensor and the target is greater than three times the target size, the magnetic dipole model can be used for approximation. However, unlike magnetic field data processing, the magnetic dipole in electromagnetic inversion changes with time. The characteristic vector y of electromagnetic data inversion is established:

[0051]

[0052] Among them, (x0, y0, z0) is the three-dimensional position of the target, are the yaw and pitch angles of the target, (k i ,β i ,γ i )i=1,2,3 are the three-dimensional magnetic dipole parameters, and the magnetic dipole response curve L in the three directions of the target i The relationship is shown in formula (4).

[0053] Based on the target location determined in step 1, local vehicle-mounted electromagnetic detection is carried out to obtain multiple observation points (x i ,y i ,z i ), the induced voltage signal d of the receiving coil obsi , i=1,2,...N;

[0054] Using multiple observation points (x i ,y i ,z i ) receiving coil’s induced voltage signal d obsi , i=1,2,...N, N>9, according to Faraday's law of electromagnetic induction, the induced voltage of the receiving coil is the time derivative of the magnetic field generated by the target. Therefore, the objective function is established as shown in formula (5), where the position information in the characteristic vector x of the magnetic survey inversion result is used as the initial value of the position information in the electromagnetic inversion characteristic vector y, and then the initial values ​​of the remaining parameters in the characteristic vector y and the iteration end condition are set, and the local optimization algorithm Levenberg-Marquardt algorithm is used to estimate the characteristic vector y of the target; in formula (5), f i () is the same as in formulas (1) and (2), and is the magnetic field signal generated by the dipole.

[0055]

[0056] Among them, the parameters in the eigenvector y of the electromagnetic data inversion are The magnetic moment m of the target can be calculated. After calculating m, it can be substituted into formula (1) to calculate f i (y), t represents time.

[0057] After the characteristic vector y is calculated, the intrinsic dipole response curve L is calculated according to formula (4): i , according to the intrinsic dipole response curve L i , based on the response curve L i The shape of the target can be obtained.

[0058]

[0059] Step 3: Fuse the feature vector x and the feature vector y to obtain the position, shape and posture features of the target.

[0060] Generally speaking, the larger the magnetic moment, the larger the target. Therefore, the size of the target can be inferred based on the magnetic moment in the eigenvector. At the same time, the shape of the target can be inferred from the intrinsic response curve. Therefore, the eigenvector x and the eigenvector y are fused to obtain the position, shape and posture characteristics of the target.

[0061] The feature vector calculation result obtained by a single detection method is inaccurate. The Bayesian estimation method is used to fuse the feature vector x and the feature vector y to obtain a more accurate target position and shape, the posterior probability of the posture feature, and realize information fusion.

[0062] For example, according to Bayesian theory, the process of updating the magnetic moment m in the feature vector x is as follows:

[0063] Calculate the posterior probability of the magnetic moment p(m|ΔT)

[0064]

[0065] The probability distribution p(m) is the prior information. The ability to reproduce the observed magnetic anomaly data ΔT is the conditional probability density p(ΔT|m). The probability distribution p(ΔT) represents the reliability of the observed magnetic anomaly data, which is related to the measurement noise signal and can be evaluated by the signal-to-noise ratio.

[0066] By multiplying the calculated posterior probability p(m|ΔT) of the magnetic moment by the magnetic moment m, the magnetic moment m in the feature vector x is updated using the Bayesian theory.

[0067] Similarly, for the remaining parameters in the feature vector x and the feature vector y, their corresponding posterior probabilities are also calculated and updated using the calculated posterior probabilities.

[0068] Since both the feature vector x and the feature vector y contain position information, according to the Bayesian theory, the probability distribution of the position information of the fused target satisfies:

[0069]

[0070] P Location is the probability of the target position after fusion, and are the probability estimates (i.e., posterior probabilities) of the target location for magnetic and electromagnetic detection, respectively.

[0071] Figure 2 There are three targets to be measured. Table 1 gives the position and attitude information of the targets. The positions are collected in advance by the differential GPS system.

[0072] Table 1 Position and posture characteristics of underground targets

[0073]

[0074]

[0075] Figure 3 It is the UAV magnetic survey data, which is obtained through data preprocessing based on the UAV magnetic survey system. Figure 3(bd) is the magnetic anomaly field generated by a single target in the survey area. According to the differential evolution algorithm, the characteristic vectors x of the three targets are determined respectively, as shown in Table 2. The table provides the position and magnetic moment information of the target, and the fitting degree is R 2 -square, evaluates the fitting characteristics of the dipole model.

[0076] Table 2 Inversion eigenvectors of magnetic field data

[0077]

[0078] Figure 4 is the induced voltage of the receiving coil in the vehicle-mounted electromagnetic detection system, and the characteristic vectors y of the three targets are inverted respectively, as shown in Table 3. The table gives the position and posture characteristics of the target, and the fitting degree is R 2 -square, evaluates the fitting characteristics of the dipole model.

[0079] Table 3. Electromagnetic data inversion eigenvectors

[0080]

[0081] Comparing Table 1, Table 2 and Table 3, Table 3 is the target position obtained by joint inversion. Compared with the actual position of the target, the positioning accuracy is higher, and the positioning error is <10cm. At the same time, the result of magnetic survey inversion cannot obtain the attitude information of the target. The result of joint inversion includes the attitude information of the target. The attitude angle error of the inversion of 4# and 5# targets is <10°, and the attitude angle error of the inversion of 6# target is larger, which is related to the irregular shape of 6# target.

[0082] Figure 5 The dipole intrinsic response curve of the target is obtained based on the inversion of electromagnetic data. Figure 5 The three intrinsic response curves of (ab) have similar characteristics. The dipole response along the long principal axis is much larger than the dipole responses in the other two directions, and decays more slowly with time. The two short principal axes of the dipole intrinsic curves in the directions basically coincide, indicating that targets 4# and 5# may be columnar and have axisymmetric characteristics. Figure 5 (c) The dipole intrinsic response curves in the three directions also decay over time, but the decay curves and amplitudes are completely different, which is consistent with the characteristics of irregular clutter. The shape characteristics of the target are obtained based on the dipole intrinsic response curves in the inversion results, and the target is classified and identified.

[0083] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A target composite detection method, characterized in that: The specific process is: Step 1: Use magnetic method to determine the location of the target; Suppose the eigenvector x=(x0,y0,z0,m x , m y , m z ) characterizes the magnetic dipole model; The magnetic field sensor carried by the magnetic measurement system is used to measure the total magnetic field signal in the space, and the total magnetic field signal is preprocessed. Then, the magnetic anomaly signal ΔT of the target production is extracted from the measured total magnetic field signal. The magnetic anomaly signal can be expressed as: Where m=(m x , m y , m z ) is the magnetic moment of the target, r = (x-x0, y-y0, z-z0) is the distance vector, (x, y, z) is the position of the sensor, (x0, y0, z0) is the position of the target, B earth is the local geomagnetic field vector, where r is the modulus of r and μ0 is the vacuum permeability; Using multiple observation points (x i ,y i ,z i ) magnetic anomaly data ΔT i , establish the objective function as shown in formula (2), and use the differential evolution algorithm to estimate the target's feature vector x, and use the differential evolution algorithm to estimate the target's feature vector x, Step 2: Determine the position of the target based on the magnetic method, and use the electromagnetic method to obtain the position, shape and posture characteristics of the target; Assume the eigenvector y of electromagnetic data inversion: Among them, (x0, y0, z0) is the position of the target, are the yaw and pitch angles of the target, (k i ,β i ,γ i )i=1,2,3 are the three-dimensional magnetic dipole parameters; Based on the target location determined in step 1, local electromagnetic detection is carried out to obtain multiple observation points (x i ,y i ,z i ), the induced voltage signal d of the receiving coil obsi ; Using multiple observation points (x i ,y i ,z i ) receiving coil’s induced voltage signal d obsi , establish the objective function as shown in formula (5), take the position information in the eigenvector x as the initial value of the position information in the electromagnetic inversion eigenvector y, and use the local optimization algorithm Levenberg-Marquardt algorithm to estimate the target eigenvector y; Step 3: Fuse the feature vector x and the feature vector y to obtain the position, shape and posture features of the target.

2. The target composite detection method according to claim 1, characterized in that: The magnetic measurement system is carried on a UAV.

3. The target composite detection method according to claim 2, characterized in that: The preprocessing of the total magnetic field signal includes: filtering the total magnetic field signal using a low-pass filter, removing the DC component, and finally compensating for the interference magnetic field of the UAV platform.

4. The target composite detection method according to claim 1, characterized in that: The local electromagnetic detection is achieved by a vehicle-mounted magnetic detection system.

5. The target composite detection method according to claim 1, characterized in that: Using more than 6 observation points (x i ,y i ,z i ) magnetic anomaly data ΔT i Establish the objective function.

6. The target composite detection method according to claim 1, characterized in that: Using more than 9 observation points (x i ,y i ,z i ) receiving coil’s induced voltage signal d obsi Establish the objective function.

7. The target composite detection method according to claim 1, characterized in that: After the characteristic vector y is calculated, the intrinsic dipole response curve L is calculated according to formula (4): i , according to the intrinsic dipole response curve L i , based on the response curve L i Get the shape of the target 8. The target composite detection method according to claim 1, characterized in that: The Bayesian estimation method is used to fuse the feature vector x and the feature vector y to obtain the position, shape and posture characteristics of the target.

Images