Robust estimation method and system for aerial natural field source dip based on spatial fusion

Abstract

The present application relates to the technical field of geophysical electromagnetic detection, and specifically discloses an airborne natural field source dip robust estimation method and system based on spatial fusion, which comprises the following steps: obtaining a set of original dip estimation values obtained by short-time window frequency domain analysis of three-component magnetic field data; calculating the skin depth of a target frequency, and determining an adaptive spatial search radius positively correlated with the target frequency based on the skin depth of the target frequency; taking an output survey point as the center and the adaptive spatial search radius as the neighborhood, searching and collecting original dip estimation values falling within the neighborhood in the set of original dip estimation values across survey lines to form a neighborhood candidate set; and performing robust spatial fusion processing on the original dip estimation values in the neighborhood candidate set to obtain the final dip result of the output survey point at the current target frequency; the present application realizes low-frequency strong fusion to suppress noise and high-frequency weak fusion to retain local structural details, thereby effectively improving the quality of dip data.

Description

Technical Field

[0001] This invention relates to the field of geophysical electromagnetic detection technology, specifically to a robust estimation method and system for airborne natural field source tilters based on space fusion. Background Technology

[0002] In recent years, unmanned aerial vehicle (UAV) platforms have been widely used for electromagnetic detection of natural field sources due to their advantages such as high mobility, low operating costs, and rapid data acquisition in complex terrain conditions. During UAV aerial observations, the platform's own motion, attitude changes, and radiation from its motor and power systems often result in strong motion noise and structured electromagnetic interference superimposed on the three-component magnetic field time-domain data. This noise often exhibits cross-component correlation and is a non-stationary process that varies with the flight trajectory. Under these conditions, tilt estimates obtained based on short-window frequency domain analysis are prone to discrete points and jumps along the survey line, directly affecting the reliability of the interpretation of underground electrical structures based on tilt data.

[0003] In existing technologies, tilt estimation is typically processed on a single survey line basis: frequency domain spectral quantities are calculated within a fixed or empirically selected time window, and the tilt result at a certain location is obtained by solving the least squares or cross-spectral matrix. To improve stability, engineering practices often employ methods such as extending the time window, smoothing along the line, or performing overall fitting of the entire survey line to suppress random fluctuations. However, due to the significant spatial correlation and cross-survey line consistency characteristics of UAV noise, processing only along a single survey line is insufficient to fully utilize the redundant information provided by adjacent and intersecting survey lines within the same underground induction volume, which may lead to unstable tilt results. Furthermore, when strong interference windows or abnormal attitude segments are mixed into the short-time window estimation values, using simple mean or standard deviation thresholds for removal often fails to simultaneously consider the real and imaginary parts of the tilt and the correlation between different tilt components, easily resulting in insufficient or incorrect removal, thus affecting the reliability of the results.

[0004] Therefore, there is an urgent need for a method to improve the tilter mass of three-component magnetic field data from UAV airborne natural field sources. Summary of the Invention

[0005] The purpose of this invention is to provide a robust estimation method and system for airborne natural field source tilters based on spatial fusion, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A robust estimation method for tiltons of aerospace natural field sources based on spatial fusion, the method comprising:

[0008] A set of original tilt estimate values ​​is obtained by short-time window frequency domain analysis of three-component magnetic field data from UAV airborne natural field sources. The original tilt estimate values ​​include spatial coordinates.

[0009] Calculate the skin depth of the target frequency, and determine the adaptive spatial search radius that is positively correlated with the target frequency based on the skin depth of the target frequency;

[0010] For each output measurement point, with the output measurement point as the center and the adaptive spatial search radius as the neighborhood, the original tilt estimation values ​​are searched across the measurement line in the set of original tilt estimation values ​​and the original tilt estimation values ​​falling into the neighborhood are collected to form a neighborhood candidate set.

[0011] Robust spatial fusion processing is performed on the original tilt estimate values ​​in the neighborhood candidate set to obtain the final tilt result of the output measurement point at the current target frequency.

[0012] As a further aspect of the present invention, determining the adaptive spatial search radius positively correlated with the target frequency specifically includes:

[0013] Calculate the skin depth based on the background resistivity and the target frequency;

[0014] The search radius is made proportional to the skin depth, and a minimum and maximum radius are set as constraints.

[0015] As a further embodiment of the present invention, the cross-survey line search includes: incorporating the original tilt estimates that satisfy the distance condition on adjacent survey lines and intersecting survey lines outside the survey line where the current output survey point is located into the neighborhood candidate set.

[0016] As a further embodiment of the present invention, the robust spatial fusion process includes at least: an outlier removal step based on multivariate statistics and a spatial weighted average step based on distance.

[0017] As a further embodiment of the present invention, the outlier removal step based on multivariate statistics includes:

[0018] The complex information of the tilts of each candidate point in the neighborhood candidate set is constructed into a multivariate feature vector, which includes at least two tilt components T. zx With T zy The real and imaginary parts;

[0019] Mahalanobis distance is calculated based on the multivariate distribution of the candidate set as an outlier index;

[0020] Candidate points whose outlier exceeds a preset threshold will be removed or their fusion weight will be reduced.

[0021] As a further embodiment of the present invention, the distance-based spatial weighted averaging step includes:

[0022] The distance weight is determined based on the distance between the candidate point and the output measurement point, and the search radius.

[0023] The final tilt result is obtained by weighting the complex values ​​of the tilt points using the distance weights.

[0024] As a further embodiment of the present invention, the distance weight is determined using a cosine taper window function.

[0025] As a further embodiment of the present invention, when the original tilt estimate also includes a quality index, the step of determining the distance weight further includes: determining the quality weight according to the quality index, taking the product of the distance weight and the quality weight as a comprehensive weight, and using the comprehensive weight to perform a weighted average.

[0026] This invention also provides a robust estimation system for tiltons of airborne natural field sources based on spatial fusion, the system comprising:

[0027] The data acquisition module is used to acquire the set of original tilter estimates obtained by short-time window frequency domain analysis of the three-component magnetic field data of the UAV airborne natural field source, wherein the original tilter estimates include spatial coordinates;

[0028] The radius determination module is used to calculate the skin depth of the target frequency and determine the adaptive spatial search radius that is positively correlated with the target frequency based on the skin depth of the target frequency.

[0029] The neighborhood determination module is used to, for each output measurement point, take the output measurement point as the center and the adaptive spatial search radius as the neighborhood, search across the measurement line in the set of original tilt estimates and collect the original tilt estimates that fall within the neighborhood to form a neighborhood candidate set.

[0030] The fusion module is used to perform robust spatial fusion processing on the original tilt estimate values ​​in the neighborhood candidate set to obtain the final tilt result of the output measurement point at the current target frequency.

[0031] Compared with the prior art, the beneficial effects of the present invention are: by establishing the relationship between skin depth and search radius, low-frequency strong fusion and high-frequency weak fusion are achieved, thereby improving stability while preserving spatial details as much as possible; by introducing robust statistical ideas, outlier identification and weighting are performed on the estimated values ​​of candidate tilters in the neighborhood, reducing the influence of strong interference windows on the final tilter results. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention.

[0033] Figure 1 This is a flowchart illustrating a robust estimation method for airborne natural field source tilters based on spatial fusion, provided in an embodiment of the present invention.

[0034] Figure 2 This is a schematic diagram of adjacent survey lines and intersecting survey lines provided in an embodiment of the present invention. Detailed Implementation

[0035] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0036] like Figure 1 As shown in the embodiment of the present invention, a robust estimation method for tiltons of airborne natural field sources based on spatial fusion is provided, the method comprising:

[0037] A set of original tilt estimate values ​​is obtained by short-time window frequency domain analysis of three-component magnetic field data from UAV airborne natural field sources. The original tilt estimate values ​​include spatial coordinates.

[0038] Calculate the skin depth of the target frequency, and determine the adaptive spatial search radius that is positively correlated with the target frequency based on the skin depth of the target frequency;

[0039] For each output measurement point, with the output measurement point as the center and the adaptive spatial search radius as the neighborhood, the original tilt estimation values ​​are searched across the measurement line in the set of original tilt estimation values ​​and the original tilt estimation values ​​falling into the neighborhood are collected to form a neighborhood candidate set.

[0040] Robust spatial fusion processing is performed on the original tilt estimate values ​​in the neighborhood candidate set to obtain the final tilt result of the output measurement point at the current target frequency.

[0041] In this embodiment, a data preprocessing step is also included: acquiring the three-component magnetic field time-domain data and synchronous positioning information collected by the UAV. The data is preprocessed to meet the basic conditions for subsequent short-window frequency domain estimation. The preprocessing includes at least one or a combination of the following: aligning the magnetic field data with the positioning and attitude data in time, and resampling to a uniform sampling rate if necessary; rotating the machine's magnetic field to a uniform geographic coordinate system, such as the NED coordinate system; marking or removing abnormal segments such as saturation, loss of lock, and abrupt attitude changes, and optionally performing detrending and impulse suppression.

[0042] The above preprocessing is used to reduce the impact of non-geophysical factors on tilt estimation, but it does not require the complete elimination of motion noise. Subsequent steps further improve the tilt quality through spatially robust fusion.

[0043] The preprocessed three-component magnetic field data is divided into short time windows, with a window length of several seconds, such as 1 second. The window shift can be 25%–50% of the window length, i.e., an overlap ratio of 50%–75%, to balance frequency domain stability and spatial continuity. Within each time window, frequency domain calculations are performed on the target frequency to obtain the original estimate of the tilton corresponding to that window. .

[0044] The At least include: GPS coordinates of the window's center ,frequency and the complex value of the tilting element Further save window-level quality metrics, such as the error, variance, or quality index of the estimate, for subsequent weighting.

[0045] This step generates a large set of tilt estimates distributed along the flight survey line trajectory. , as input data for spatial fusion.

[0046] In a preferred embodiment of the present invention, determining the adaptive spatial search radius positively correlated with the target frequency specifically includes:

[0047] Calculate the skin depth based on the background resistivity and the target frequency;

[0048] The search radius is made proportional to the skin depth, and a minimum and maximum radius are set as constraints.

[0049] In this embodiment, in order to improve stability at low frequencies and preserve spatial details as much as possible at high frequencies, for each target frequency... Calculate skin depth And determine the search radius accordingly. .

[0050] Skin depth can be approximated using engineering methods:

[0051]

[0052] in The background resistivity can be taken as a representative value of the measurement area or an empirical average value of the entire area, for example, 1000 ohm-meters. Further let... and Proportionately, and It is limited to between the minimum and maximum radii, thus forming an adaptive spatial fusion scale with "large low-frequency radius and small high-frequency radius".

[0053] The minimum radius should be no less than 1-2 times the survey line spacing to ensure that information from adjacent survey lines can be incorporated; the maximum radius is used to avoid excessive smoothing, for example, taking 3-5 times the survey line spacing, or not exceeding the desired minimum structural scale.

[0054] like Figure 2 As shown, in a preferred embodiment of the present invention, the cross-survey line search includes: incorporating the original tilt estimates that satisfy the distance condition on adjacent survey lines and intersecting survey lines outside the survey line where the current output survey point is located into the neighborhood candidate set.

[0055] In this embodiment, to ensure that tilter results at different frequencies can be output at the same spatial location, a set of output measurement points is defined within the measurement area. ,in, This indicates the output measurement point number. This indicates the j-th output measurement point. Output measurement points can be laid out along each measurement line at fixed intervals, such as 50 to 300 meters, or they can be laid out using a regular grid.

[0056] The optimal setting for output measurement points should balance sampling capability with the lowest target frequency, ensuring that the sampling capability corresponds to the lowest frequency. A sufficient number can be collected within the neighborhood. This satisfies the needs of robust statistics and fusion computing.

[0057] For each output measurement point and each frequency ,by Centered on, with The search radius is derived from the original set of estimates for the short-window tilt. The search yields a neighborhood candidate set. .in, Each candidate All include the coordinates of their window center. Corresponding frequency and the complex value of the oscillator It may also include quality metrics.

[0058] The key to the neighborhood search is that the candidate set is not limited to the current survey line; any candidate point on an adjacent or intersecting survey line that falls within the neighborhood is included in the candidate set. This naturally creates multi-survey line redundancy constraints under low-frequency conditions to suppress motion noise. Figure 2 As shown, Figure 2In the diagram, L10 is adjacent to L09 and L11, and intersects with L02 and L03. Purple dots represent output measurement points, and black circles represent the search radius. Under high-frequency conditions, due to automatic neighborhood shrinkage, data near the output measurement points are primarily fused to maintain spatial detail resolution as much as possible. If the number of candidate points for a certain output measurement point at a certain frequency is insufficient, the point can be marked as low confidence or not output; or the search radius can be appropriately expanded without exceeding the maximum radius; or the number of effective samples can be increased by merging frequency bands.

[0059] As a preferred embodiment of the present invention, the robust spatial fusion process includes at least: an outlier removal step based on multivariate statistics and a spatial weighted average step based on distance.

[0060] The outlier removal step based on multivariate statistics includes:

[0061] The complex information of the tilts of each candidate point in the neighborhood candidate set is constructed into a multivariate feature vector, which includes at least two tilt components. and The real and imaginary parts;

[0062] Mahalanobis distance is calculated based on the multivariate distribution of the candidate set as an outlier index;

[0063] Candidate points whose outlier exceeds a preset threshold will be removed or their fusion weight will be reduced.

[0064] In this embodiment, the obtained set of neighborhood candidate tilt estimates is... Because these candidate points may contain tilt estimates corresponding to strong interference, abrupt attitude changes, or non-stationary signal segments, they often exhibit abrupt changes in the real or imaginary part of the tilt, or... and Simultaneous anomalies can significantly shift the subsequent averaging results. Therefore, this invention preferably performs outlier removal before spatial weighting to ensure that the candidate points participating in the fusion have relatively consistent statistical characteristics.

[0065] Specifically, the complex information of each candidate tilt point at frequency f is constructed into a multivariate feature vector: ;

[0066] Based on the overall distribution of the candidate set, the central location and dispersion are calculated, and the correlation between the four components mentioned above is specifically considered to construct a robust outlier index. Outlier can be expressed using multivariate distance measures, such as Mahalanobis distance-like indices: ;

[0067] in and Let represent the center and covariance of the candidate set, respectively. Robust estimation is preferred to reduce the contamination of the statistics by outliers themselves. Exceeding the preset threshold When this happens, the corresponding candidate tilt estimates are discarded, or weighted according to outlier degree; preferably, a threshold is used. Choose a number of 4 to 6, for example, 5, to strike a balance between outlier removal and retention of valid samples. This step yields a cleansing candidate set. This is used for subsequent spatial weighted fusion.

[0068] In a preferred embodiment of the present invention, the distance-based spatial weighted averaging step includes:

[0069] The distance weight is determined based on the distance between the candidate point and the output measurement point, and the search radius.

[0070] The final tilt result is obtained by weighting the complex values ​​of the tilt points using the distance weights.

[0071] The distance weights are determined using a cosine taper window function.

[0072] In this embodiment, after obtaining the purification candidate set Then, the present invention measures each output measurement point In frequency Spatial weighted fusion is then performed to obtain enhanced tilt results. This fusion process follows the principle of "greater weight for nearby components, smaller weight for distant components, and smooth attenuation at the boundary" to avoid spatial artifacts caused by hard truncation and to maintain a smooth transition of results when using multi-line information over a wider range at low frequencies.

[0073] Specifically, the distance between each candidate point and the output measurement point is calculated. And based on the search radius Construct distance weights .

[0074] The distance weighting preferably adopts a tapered window form: a larger weight, such as 1, is assigned to the central region near the output measurement point; as the distance gradually approaches... At this time, the weights are smoothly decayed to 0 according to a cosine taper, i.e., a Tukey window form, thus ensuring the gradual change in neighborhood fusion. Through this design, at low frequencies... It is relatively large and can automatically incorporate redundant information from adjacent and intersecting survey lines to suppress noise; at high frequencies, due to It is relatively small, and only local neighborhood points are merged to preserve local structural details as much as possible.

[0075] In a preferred embodiment of the present invention, when the original tilt estimate also includes a quality index, the step of determining the distance weight further includes: determining the quality weight based on the quality index, using the product of the distance weight and the quality weight as a comprehensive weight, and using the comprehensive weight to perform a weighted average.

[0076] In this embodiment, when the short-window tilt estimate is accompanied by a window-level quality metric, such as coherence, fitting residual, or estimation variance, a quality weight can be defined. This ensures that high-quality candidate points account for a higher proportion in the fusion process, while low-quality candidate points account for a lower proportion.

[0077] The overall weight can be written as:

[0078] ;

[0079] Finally, respectively and The weighted average of the two components yields the enhanced tilt result at the output measurement point:

[0080] ;

[0081] in Indicates the candidate point corresponding to or .

[0082] To provide credibility labels for the output results, uncertainty indices can be given based on the weighted dispersion of candidate points after fusion. For example, statistics can be given for the real and imaginary parts separately, so as to distinguish between stable and unstable regions when plotting and interpreting tilt arrows.

[0083] The above steps form a complete processing flow of "short-window inclination estimation - physical scale constraint - cross-survey line spatial fusion": First, time synchronization, coordinate unification and outlier segment processing are completed for the three-component magnetic field data to ensure the stability of short-window frequency domain estimation; then, the set of inclination estimation values ​​distributed along the track is obtained within a short time window, and the search radius that varies with frequency is determined based on the skin depth, so as to expand the neighborhood at low frequencies to enhance stability and shrink the neighborhood at high frequencies to preserve spatial details; then, using the output measurement point as a unified carrier, candidate points in the neighborhood are collected across the survey line, and robust outlier removal and tapered window weighted averaging are combined to reduce data jumps caused by motion noise, thereby improving the quality of inclination results.

[0084] This invention also provides a robust estimation system for airborne natural field source tilters based on spatial fusion, the system comprising:

[0085] The data acquisition module is used to acquire the set of original tilter estimates obtained by short-time window frequency domain analysis of the three-component magnetic field data of the UAV airborne natural field source, wherein the original tilter estimates include spatial coordinates;

[0086] The radius determination module is used to calculate the skin depth of the target frequency and determine the adaptive spatial search radius that is positively correlated with the target frequency based on the skin depth of the target frequency.

[0087] The neighborhood determination module is used to, for each output measurement point, take the output measurement point as the center and the adaptive spatial search radius as the neighborhood, search across the measurement line in the set of original tilt estimates and collect the original tilt estimates that fall within the neighborhood to form a neighborhood candidate set.

[0088] The fusion module is used to perform robust spatial fusion processing on the original tilt estimate values ​​in the neighborhood candidate set to obtain the final tilt result of the output measurement point at the current target frequency.

[0089] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A robust estimation method for tilters of aerospace natural field sources based on spatial fusion, characterized in that, The method includes: A set of original tilt estimate values ​​is obtained by short-time window frequency domain analysis of three-component magnetic field data from UAV airborne natural field sources. The original tilt estimate values ​​include spatial coordinates. Calculate the skin depth of the target frequency, and determine the adaptive spatial search radius that is positively correlated with the target frequency based on the skin depth of the target frequency; For each output measurement point, with the output measurement point as the center and the adaptive spatial search radius as the neighborhood, the original tilt estimation values ​​are searched across the measurement line in the set of original tilt estimation values ​​and the original tilt estimation values ​​falling into the neighborhood are collected to form a neighborhood candidate set. Robust spatial fusion processing is performed on the original tilt estimate values ​​in the neighborhood candidate set to obtain the final tilt result of the output measurement point at the current target frequency.

2. The robust estimation method for tiltors of aerospace natural field sources based on spatial fusion according to claim 1, characterized in that, The determination of the adaptive spatial search radius positively correlated with the target frequency specifically includes: Calculate the skin depth based on the background resistivity and the target frequency; The search radius is made proportional to the skin depth, and a minimum and maximum radius are set as constraints.

3. The robust estimation method for tilters of aerospace natural field sources based on spatial fusion according to claim 1, characterized in that, The cross-survey line search includes: incorporating the original tilt estimates that meet the distance conditions on adjacent and intersecting survey lines outside the survey line where the current output survey point is located into the neighborhood candidate set.

4. The robust estimation method for tilters of aerospace natural field sources based on spatial fusion according to claim 1, characterized in that, The robust spatial fusion process includes at least: an outlier removal step based on multivariate statistics and a spatial weighted average step based on distance.

5. The robust estimation method for airborne natural field source tilters based on spatial fusion according to claim 4, characterized in that, The outlier removal step based on multivariate statistics includes: The complex information of the tilts of each candidate point in the neighborhood candidate set is constructed into a multivariate feature vector, which includes at least two tilt components T. zx With T zy The real and imaginary parts; Mahalanobis distance is calculated based on the multivariate distribution of the candidate set as an outlier index; Candidate points whose outlier exceeds a preset threshold will be removed or their fusion weight will be reduced.

6. The robust estimation method for airborne natural field source tilters based on spatial fusion according to claim 4, characterized in that, The distance-based spatial weighted averaging step includes: The distance weight is determined based on the distance between the candidate point and the output measurement point, and the search radius. The final tilt result is obtained by weighting the complex values ​​of the tilt points using the distance weights.

7. The robust estimation method for tiltors of aerospace natural field sources based on spatial fusion according to claim 6, characterized in that, The distance weights are determined using a cosine taper window function.

8. The robust estimation method for tiltors of aerospace natural field sources based on spatial fusion according to claim 6, characterized in that, When the original tilt estimate also includes a quality index, the step of determining the distance weight further includes: determining the quality weight based on the quality index, using the product of the distance weight and the quality weight as a comprehensive weight, and using the comprehensive weight to perform a weighted average.

9. A robust estimation system for tilters of aerospace natural field sources based on spatial fusion, used to implement the robust estimation method for tilters of aerospace natural field sources based on spatial fusion as described in any one of claims 1-8, characterized in that, The system includes: The data acquisition module is used to acquire the set of original tilter estimates obtained by short-time window frequency domain analysis of the three-component magnetic field data of the UAV airborne natural field source, wherein the original tilter estimates include spatial coordinates; The radius determination module is used to calculate the skin depth of the target frequency and determine the adaptive spatial search radius that is positively correlated with the target frequency based on the skin depth of the target frequency. The neighborhood determination module is used to, for each output measurement point, take the output measurement point as the center and the adaptive spatial search radius as the neighborhood, search across the measurement line in the set of original tilt estimates and collect the original tilt estimates that fall within the neighborhood to form a neighborhood candidate set. The fusion module is used to perform robust spatial fusion processing on the original tilt estimate values ​​in the neighborhood candidate set to obtain the final tilt result of the output measurement point at the current target frequency.

Images