CN106707352B - A kind of minimizing technology of the aeromagnetic interference based on ε-SVR algorithms - Google Patents

Abstract

A method for removing aeromagnetic gradient interference based on the ε-SVR algorithm, comprising: obtaining a compensation model for the interference of the aircraft itself based on the ε-SVR algorithm; obtaining aircraft survey data; and using the compensation model to compensate the survey data, to Remove the influence of the aircraft's interference magnetic field.

Description

A Removal Method of Aeromagnetic Gradient Interference Based on ε-SVR Algorithm

technical field

The invention relates to the field of geophysical aeromagnetic prospecting, in particular to a method for removing aeromagnetic gradient interference of an unmanned aerial vehicle based on an ε-SVR algorithm.

Background technique

As an important means of airborne geophysical prospecting, aeromagnetic prospecting has been widely used in the field of geophysics. Traditional aeromagnetic exploration platforms are dominated by manned machines. In the past ten years, with the development of UAV technology, UAVs have been widely used in the field of aeromagnetic exploration. Compared with manned machines, UAVs are cheaper , high efficiency, safety and other significant advantages. However, due to the small size of the aircraft and the short baseline between the probes of the UAV, the interference magnetic field of the aircraft is very significant in the aeromagnetic gradient data, which seriously affects the data quality and final mapping effect of the aeromagnetic survey. Therefore, it is of great significance to effectively remove the influence of aircraft interference magnetic field in aeromagnetic gradient measurement.

At present, foreign magnetic compensation equipment is mainly used in the field of aeromagnetic exploration in China, such as the AADC series magnetic compensation instrument of RMS company and the aeromagnetic compensation equipment of PICO company. The compensation algorithm of the above compensation equipment is designed based on the traditional aeromagnetic compensation algorithm. The characteristic of this algorithm is that first pass the data collected by the optical pump magnetometer and fluxgate magnetometer through a low-pass filter to filter out part of the noise that is not related to the aircraft magnetic interference, and then use the least squares algorithm to achieve the removal The purpose of the aircraft to interfere with the magnetic field.

The existing UAV aeromagnetic survey compensation has the following defects:

(1) At present, the aeromagnetic compensation method of unmanned aerial vehicles mainly continues the magnetic compensation method of manned aircraft, but the flying height of common unmanned aerial vehicles is limited, and it is impossible to achieve a high flight altitude of 3000 meters.

At the same time, common UAVs have poor maneuverability after carrying a certain load, and cannot achieve roll, pitch, yaw and other maneuvers in standard calibration flights. Therefore, calibration flights designed for manned aircraft cannot be applied to common UAVs.

(2) There is a problem of multicollinearity in the traditional aeromagnetic compensation model, and the least squares algorithm cannot solve this problem well.

Contents of the invention

In view of the problems existing in the existing solutions, in order to overcome the shortcomings of the above-mentioned prior art solutions, the present invention proposes a UAV aeromagnetic gradient interference removal method based on the ε-SVR algorithm.

According to one aspect of the present invention, a method for removing aeromagnetic gradient interference based on the ε-SVR algorithm is provided, including: obtaining a compensation model for the interference of the aircraft itself based on the ε-SVR algorithm; obtaining aircraft survey data; and using the The compensation model compensates the survey data to remove the influence of the aircraft's disturbing magnetic field.

As can be seen from the foregoing technical solutions, the present invention has the following beneficial effects:

(1) The UAV aeromagnetic gradient interference removal method adopts a new ε-SVR algorithm, which can effectively realize the low-altitude calibration flight of the UAV, and effectively avoid the deterioration of the compensation results caused by the multicollinearity in the aeromagnetic compensation model.

(2) Binary wavelet processing is performed on the signal data to remove interference irrelevant to aircraft maneuvering, and there is no distortion in the signal phase.

Description of drawings

Fig. 1 is the flow chart of the UAV aeromagnetic gradient interference removal method based on ε-SVR algorithm according to the embodiment of the present invention;

Fig. 2 is the flowchart of obtaining compensation model in Fig. 1;

Fig. 3 is the flowchart of the concrete operation of ε-SVR algorithm in Fig. 2;

Fig. 4 is a graph comparison chart before and after gradient data compensation in exploration flight.

Detailed ways

Certain embodiments of the invention will be described more fully hereinafter with reference to the accompanying drawings, in which some, but not all embodiments are shown. Indeed, various embodiments of the invention may be embodied in many different forms and should not be construed as limited to these set forth embodiments; rather, these embodiments are provided so that this invention will satisfy applicable legal requirements.

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be described in further detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

In this field, the calibration flight refers to the process of modeling and matching the interference magnetic field of the aircraft before the aeromagnetic survey flight. The specific operation is to first fly the flight platform to a certain height, complete the flight in four orthogonal headings, and use the magnetic compensation algorithm to model the data obtained during the flight, so as to fit the interference magnetic field of the aircraft.

The embodiment of the present invention provides a UAV aeromagnetic gradient interference removal method based on the ε-SVR algorithm. Based on the ε-SVR algorithm, the compensation model for the interference of the aircraft itself is obtained during the calibration flight, and the compensation model is applied to the exploration. The flight obtains aeromagnetic survey data and compensated aeromagnetic gradient data.

A specific embodiment of the present invention provides a method for removing the aeromagnetic gradient interference of an unmanned aerial vehicle based on the ε-SVR algorithm, including the following steps, as shown in Figure 1:

S101. Obtain a compensation model for the interference of the UAV itself based on the ε-SVR algorithm during the UAV calibration flight.

The feature of using this method to obtain the compensation model is that it uses the method of machine learning, regards the calibration flight as a learning process, uses the ε-SVR algorithm to obtain the regression hyperplane, and uses the ε-SVR algorithm based on the characteristics of structural risk minimization, which can effectively To solve the problem of multicollinearity in the aeromagnetic compensation model, the calibration flight of the UAV in the present invention can be a low-altitude flight, and there is no need for the UAV to perform large maneuvering flight actions such as roll, pitch, and yaw.

S102, acquiring exploration data during the unmanned aerial vehicle exploration flight;

The exploration data includes: aeromagnetic survey gradient data, UAV interference data and interference data irrelevant to aircraft maneuvering.

S103, performing binary wavelet processing on the obtained exploration data;

The binary wavelet processing of the exploration data can remove the interference irrelevant to the maneuvering of the UAV, such as high-frequency electrical noise, etc.; and it will not cause distortion of the phase of the data signal. After binary wavelet processing, the exploration data only include aeromagnetic gradient data and UAV interference data.

S104, using the compensation model obtained in S101 to compensate the exploration data.

The compensation model is used to compensate the exploration data processed by the binary wavelet, remove the interference data of the UAV itself, and obtain accurate aeromagnetic gradient data.

In this embodiment, step S101 specifically includes, as shown in FIG. 2:

S201: Obtain calibration data during the calibration flight of the UAV;

The survey data includes calibration survey gradient data, UAV interference data and interference data irrelevant to aircraft maneuvering.

S202: Perform binary wavelet processing on the obtained calibration data;

Binary wavelet processing of the calibration data can remove interference that is not related to the maneuvering of the UAV, such as high-frequency electrical noise, and will not cause distortion of the phase of the data signal. After binary wavelet processing, the exploration data only include aeromagnetic gradient data and UAV interference data.

S203: Perform down-sampling processing on the calibration data processed by binary wavelet;

Since the ε-SVR algorithm has a good learning generalization ability for small-scale samples, processing the ε-SVR algorithm on the calibration data after downsampling can significantly speed up the processing efficiency of the algorithm.

S204: Perform ε-SVR algorithm processing on the down-sampled calibration data to obtain a compensation model.

As shown in Figure 3, the specific operation of this step is as follows:

S301: Bring the downsampled calibration data into the original optimization problem;

The expression of the original optimization problem is shown in formula (1).

where w is the fitting coefficient of the hyperplane, C is the penalty parameter, ξ _i and is the relaxation parameter, b is the intercept of the hyperplane, ε is the data fluctuation range that the hyperplane can tolerate, G _di is the influence value of the disturbance magnetic field of the aircraft at the i-th sample point on the total magnetic field gradient data, which can be determined by the optical pump magnetic force Measured by the instrument, st represents the constraint condition.

It is worth noting that the penalty parameter removes the aeromagnetic abnormal signal as an abnormal point from the regression problem, which effectively solves the problem that the UAV cannot fly high and complete a large-angle maneuvering flight.

Let A _i = (cosX _i , cosY _i , cosZ _i , H _ei cosX _i cosY _i ,H _ei cos ² Y _i ,H _ei cosX _i cosZ _i ,H _ei cosY _i cosZ _i ,H _ei cos ² Z _i ,H _ei cosX _i (cosX _i )',H _ei cosX _i ( cosY _i )',H _ei cosX _i (cosZ _i )',H _ei cosY _i (cosX _i )',H _ei cosY _i (cosY _i )',H _ei cosY _i (cosZ _i )',H _ei cosZ _i ( cosX _i )',H _ei cosZ _i (cosY _i )',H _ei cosZ _i (cosZ _i )') is the characteristic matrix formed by the measured value of the fluxgate at the i-th sample point, for the i-th sample point, cosX _i , cosY _i , cosZ _i are direction cosines, (cosX _i )′, (cosY _i )′, (cosZ _i )′ are the derivatives of direction cosines with respect to time, where:

cosX _i = T _i /H _ei

cosY _i =L _i /H _ei

cosZ _i =V _i /H _ei

T _i , L _i and V _i are the measured values of the fluxgate at the ith sample point,

The kernel function can map the inner product operation to a high-dimensional space, and realize the calculation of the hyperplane in the high-dimensional space,

In the problem of aeromagnetic gradient compensation, the kernel function can choose the exponential form, and its mapping method is shown in the following formula (2).

In this implementation, a reasonable selection of the mapping method Φ(A _i ) of the kernel function can obtain an excellent compensation effect. To a certain extent, it makes up for the possible high-order interference and partial nonlinear problems in the compensation model. The nonlinear mapping of the characteristic matrix A _i is shown in formula (3).

S302: converting the original optimization problem into a dual optimization problem;

Using the Lagrange multiplier method, the original optimization problem can be written as shown in formula (4).

By solving the partial derivative of the above formula and making it zero, it can be obtained as shown in formula (5).

in As an intermediate variable, substituting formula (5) into formula (4), the dual optimization problem of the original optimization problem can be obtained as shown in formula (6).

where α _i and is the dual optimization problem to be solved.

S303: Solving the dual problem to obtain a compensation model.

By solving the dual problem, the regression hyperplane of the original optimization problem is obtained, and the compensation model is obtained as shown in formula (7).

Among them, A is the characteristic matrix of the sample points to be compensated, and f(A) is the fitted interference magnetic field.

Fig. 4 is the comparison result of the gradient data before and after compensation in the exploration flight. Fig. 4(a) and Fig. 4(b) are the direct mapping of the existing vertical gradient and horizontal gradient data. It can be seen that there are obvious ripples and Stripes, Fig. 4(c) and Fig. 4(d) are the mapping results processed by the method in the example of the present invention. It can be seen that the ripples and strips have been removed, and the aeromagnetic anomaly is clearly visible. Although the embodiment of the present invention performs binary wavelet processing and down-sampling processing on the calibration data and binary wavelet processing on the exploration data, these steps are not necessary, and those skilled in the art can omit these steps according to specific situations.

It should be noted that the removal method of the aeromagnetic gradient interference based on the ε-SVR algorithm in the present invention is not only used in the case of the drone in the embodiment, but also can be used in the case of other aircraft for exploration.

It should also be noted that the text may provide examples of parameters that include specific values, but these parameters need not be exactly equal to the corresponding values, but may approximate the corresponding values within acceptable error tolerances or design constraints.

It should be noted that, in the accompanying drawings or in the text of the specification, implementations that are not shown or described are forms known to those of ordinary skill in the art, and are not described in detail. In addition, the above definitions of each element and method are not limited to the various specific structures, shapes or methods mentioned in the embodiments, and those of ordinary skill in the art can easily modify or replace them, for example:

(1) Unless specifically described or steps that must occur sequentially, the order of the above steps is not limited to that listed above, and can be changed or rearranged according to the desired design.

(2) ε-SVR algorithm processing can use different types of kernel functions, such as linear kernel functions to replace the kernel functions described in this paper.

(3) The ε-insensitive loss function and penalty parameters can be replaced by different functions and values according to the actual data.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention, and are not intended to limit the present invention. Within the spirit and principles of the present invention, any modifications, equivalent replacements, improvements, etc., shall be included in the protection scope of the present invention.

Claims (7)

1. a kind of minimizing technology of the aeromagnetic interference based on ε-SVR algorithms, which is characterized in that including：

The compensation model interfered for aircraft itself is obtained based on ε-SVR algorithms；

Obtain aircraft survey data；And

Survey data is compensated using the compensation model, to remove the influence that aircraft interferes magnetic field, in aircraft mark It is fixed to include for the compensation model that aircraft itself interferes based on the acquisition of ε-SVR algorithms in-flight：

Nominal data is obtained, which in-flight obtains in calibration；

Down-sampled processing is carried out to the nominal data；And

ε-SVR algorithm process is carried out to down-sampled treated nominal data and obtains compensation model,

Further include between obtaining survey data and being compensated to survey data using the compensation model：For surveying for acquisition It visits data and carries out dyadic wavelet processing；And

Further include in acquisition nominal data and between the down-sampled processing of nominal data progress：The nominal data of acquisition carries out Dyadic wavelet processing.

2. minimizing technology according to claim 1, which is characterized in that carry out ε-to down-sampled treated nominal data SVR algorithm process obtains compensation model：

Treated that nominal data is brought into original optimization problem by down-sampled；

Original optimization problem is converted into primal-dual optimization problem；And

Primal-dual optimization problem is solved, compensation model is obtained.

3. minimizing technology according to claim 2, which is characterized in that the expression formula of the original optimization problem is as follows：

G_di-w^TΦ(A_i)-b≤ε+ξ_i,

Wherein：W is the fitting coefficient of hyperplane, w^TIt is the transposition of the fitting coefficient of hyperplane, C is punishment parameter, ξ_iWithTo relax Henan parameter, b are the intercept of hyperplane, and ε is the data fluctuations range that hyperplane can be tolerated, G_diFor i-th sample point aircraft It interferes magnetic field to the influence numerical value of the total field gradient data of magnetic, is measured by optical pumped magnetometer, s.t. indicates constraints；

Wherein：A_iFor the eigenmatrix that the measured value of i-th of sample point fluxgate is formed, Φ (A_i) it is kernel function mapping mode, institute State that down-sampled treated that nominal data includes A_iAnd G_di。

4. minimizing technology according to claim 3, which is characterized in that the kernel function mapping mode

5. minimizing technology according to claim 3, which is characterized in that i-th of sample point eigenmatrix A_i= (cosX_i, cosY_i, cosZ_i,H_eicosX_icosY_i,H_eicos²Y_i,H_eicosX_icosZ_i,H_eicosY_icosZ_i, H_eicos²Z_i,H_eicosX_i(cosX_i)',H_eicosX_i(cosY_i)',H_eicosX_i(cosZ_i)',H_eicosY_i(cosX_i)', H_eicosY_i(cosY_i)',H_eicosY_i(cosZ_i)',H_eicosZ_i(cosX_i)',H_eicosZ_i(cosY_i)',H_eicosZ_i (cosZ_i) ') wherein, cosX_i, cosY_i, cosZ_iFor direction cosines, (cosX_i) ', (cosY_i) ', (cosZ_i) ' it is direction cosines To the derivative of time,

cosX_i=T_i/H_ei

cosY_i=L_i/H_ei

cosZ_i=V_i/H_ei

T_i、L_iAnd V_iFor the measured value of the fluxgate of i-th of sample point,

6. minimizing technology according to claim 3, which is characterized in that the primal-dual optimization problem is：

Wherein, α_iWithFor the to be solved of primal-dual optimization problem.

7. minimizing technology according to claim 6, which is characterized in that the compensation model is

Wherein A is the eigenmatrix of sample point to be compensated.