A low-altitude unmanned aerial vehicle magnetic detection method and system for magnetic target positioning

Abstract

This invention relates to a low-altitude unmanned aerial vehicle (UAV) magnetic detection method and system for magnetic target localization, belonging to the field of magnetic detection technology. The method includes: determining the three-dimensional measurement point trajectories of each quantum magnetic sensor based on the UAV's positioning data, altitude data, and attitude data at various sampling times during flight; compensating for the magnetic field data of multiple spatial measurement points synchronously collected by each quantum magnetic sensor at each sampling time based on the three-dimensional measurement point trajectories; calculating the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio based on the compensated magnetic field data; the magnetic gradient tensor norm characterizes the energy intensity of the magnetic gradient; the magnetic gradient vertical coupling ratio characterizes the coupling relationship between the horizontal and vertical magnetic gradients; and performing magnetic target localization inversion based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio to obtain the position and burial depth of the magnetic target.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of magnetic detection technology, specifically to a low-altitude UAV magnetic detection method and system for locating magnetic targets. Background Technology

[0002] Currently, the detection of magnetic targets in buried subways mainly employs methods such as handheld magnetometers, vehicle-mounted magnetic measurement systems, conventional airborne magnetic measurement systems, and low-altitude unmanned aerial vehicle (UAV) magnetic detection systems equipped with classic magnetic sensors.

[0003] However, while existing technologies can detect low-altitude magnetic fields, they are limited by factors such as the background noise of classical magnetic sensors, platform magnetic interference, the influence of flight, and insufficient dimensionality of the acquired magnetic field data. Therefore, it is still difficult to stably and accurately complete the rapid positioning and depth inversion of buried magnetic targets. Summary of the Invention

[0004] The purpose of the embodiments in this specification is to provide a low-altitude UAV magnetic detection method and system for magnetic target localization, so as to overcome the problem that existing methods are difficult to stably and accurately complete the rapid localization and burial depth inversion of buried magnetic targets.

[0005] To address the aforementioned technical issues, this specification provides, on the one hand, a low-altitude UAV magnetic detection method for magnetic target localization, applied to the ground end of a magnetic detection and positioning system; the magnetic detection and positioning system also includes an air end; the air end includes a UAV and a quantum magnetic detection module installed on the UAV, the quantum magnetic detection module and the magnetic source of the UAV body satisfy a preset distance constraint, and the quantum magnetic detection module includes multiple quantum magnetic sensors distributed in a regular tetrahedral pattern; The method includes: Based on the UAV's positioning data, altitude data, and attitude data at each sampling moment during flight, the three-dimensional measurement point trajectory of each quantum magnetic sensor is determined; Based on the three-dimensional measurement point trajectory, the magnetic field data of multiple spatial measurement points synchronously collected by each quantum magnetic sensor at each sampling time are compensated. Based on the compensated magnetic field data, the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio are calculated; the magnetic gradient tensor norm is used to characterize the energy intensity of the magnetic gradient; the magnetic gradient vertical coupling ratio is used to characterize the coupling relationship between the horizontal and vertical magnetic gradients. Based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio, the location and burial depth of the magnetic target are obtained through magnetic target localization inversion.

[0006] Furthermore, determining the three-dimensional measurement point trajectory of each quantum magnetic sensor based on the UAV's positioning data, altitude data, and attitude data at each sampling moment during flight includes: Based on the positioning data, determine the geographic coordinates of the UAV at each sampling time; Based on the attitude data, the mounting arm length of each quantum magnetic sensor is transformed from the body coordinate system to the geographic coordinate system to obtain the offset of each quantum magnetic sensor relative to the UAV in the geographic coordinate system; the mounting arm length is used to characterize the fixed spatial offset of each quantum magnetic sensor relative to the UAV. By fusing the geographic coordinates of the UAV at each sampling time with the offset of each quantum magnetic sensor, the geographic coordinates of each quantum magnetic sensor at each sampling time are obtained. The geographic coordinates of each quantum magnetic sensor at multiple sampling times are fused in chronological order to obtain the three-dimensional measurement point trajectory of each quantum magnetic sensor.

[0007] Furthermore, the compensation for the magnetic field data of multiple spatial measurement points synchronously collected by each quantum magnetic sensor at each sampling time based on the three-dimensional measurement point trajectory includes: Based on the three-dimensional measurement point trajectories of each quantum magnetic sensor, the actual spatial coordinates of each quantum magnetic sensor in the geographic coordinate system at each sampling time are extracted; the actual spatial coordinates include the horizontal position and the height above the ground. Based on the comparison between the actual ground clearance of each quantum magnetic sensor and the preset reference height, the magnetic field data is subjected to height normalization compensation. Based on the comparison between the actual horizontal position of each quantum magnetic sensor and the horizontal position of the UAV, as well as the attitude data of the UAV, horizontal position registration compensation is performed on the magnetic field data.

[0008] Further, the calculation of the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio based on the compensated magnetic field data includes: Based on the compensated magnetic field data at each sampling time and the relative spatial positions of each spatial measuring point at that sampling time in the regular tetrahedron, multiple tensor components of the magnetic gradient tensor at the geometric center of the regular tetrahedron at that sampling time are calculated; each tensor component is used to characterize the rate of change of a magnetic field component along a specific spatial direction. Based on each tensor component of the magnetic gradient tensor at each sampling moment, calculate the magnetic gradient tensor norm and the magnetic gradient vertical coupling ratio corresponding to that sampling moment.

[0009] Furthermore, the method also includes: For each sampling time, based on the spatial geometric positions of N spatial measuring points in a regular tetrahedron, a magnetic field prediction model is constructed to predict the magnetic field data at the Nth spatial measuring point from the magnetic field data of any N-1 spatial measuring points; the magnetic field prediction model is used to characterize the spatial distribution relationship of the magnetic field under the constraints of the physical laws of the magnetic field passive region. Each spatial measuring point is taken as the predicted point. Based on the compensated magnetic field data of the remaining N-1 measuring points, the theoretical magnetic field data at the predicted point is calculated by the magnetic field prediction model. The weighting coefficient of the compensated magnetic field data at the predicted point is determined based on the vector difference between the compensated magnetic field data and the theoretical magnetic field data; the weighting coefficient is negatively correlated with the vector difference. Further, the step of calculating multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron at each sampling moment based on the compensated magnetic field data at each sampling moment and the relative spatial positions of each spatial measurement point in the tetrahedron at that sampling moment includes: Based on the compensated magnetic field data and corresponding weighting coefficients at each sampling time, as well as the relative spatial positions of each spatial measurement point in the tetrahedron at that sampling time, calculate multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron at that sampling time.

[0010] Furthermore, the magnetic target localization inversion based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio to obtain the position and burial depth of the magnetic target includes: Based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio, a magnetic anomaly distribution map of the target region is generated using a magnetic source physical prior model; the magnetic source physical prior model is used to characterize the attenuation relationship of the theoretical tensor features generated by the equivalent magnetic source with spatial distance. Based on the magnetic anomaly distribution map, one or more candidate anomaly regions in the target area are identified. For each candidate region of anomalies, magnetic target localization inversion is performed to obtain the location and burial depth of the magnetic target in that candidate region of anomalies.

[0011] Furthermore, the step of generating a magnetic anomaly distribution map of the target region using a priori physical model of the magnetic source, based on the magnetic gradient tensor, the magnetic gradient tensor norm, and the magnetic gradient vertical coupling ratio, includes: The magnetic gradient tensor's multiple tensor components, the magnetic gradient tensor norm, and the magnetic gradient vertical coupling ratio are used as tensor features of the magnetic anomaly distribution map to be generated. An objective function is constructed for the two-dimensional mesh numerical model to be solved in the target region. The objective function includes: a fitting error term between the tensor eigenvectors to be solved at each two-dimensional mesh point and the measured tensor eigenvectors at each sampling time; a spatial smoothing constraint term between two-dimensional mesh points; and a priori constraint term based on the magnetic source physical prior model. The measured tensor eigenvector at each sampling time is used to characterize the tensor eigenvector at the geometric center of the tetrahedron at that sampling time. The spatial smoothing constraint term is used to characterize the smoothness of the change of tensor eigenvectors between adjacent two-dimensional mesh points. The priori constraint term is used to characterize the degree of deviation between the tensor eigenvectors to be solved in each two-dimensional mesh and the theoretical tensor eigenvectors given by the magnetic source physical prior model. Find a two-dimensional grid numerical model that minimizes the objective function; Based on the obtained two-dimensional grid numerical model, a magnetic anomaly distribution map of the target area is generated.

[0012] Further, the step of performing magnetic target localization inversion for each anomaly candidate region to obtain the position and burial depth of the magnetic target in that anomaly candidate region includes: For each anomaly candidate region, multiple magnetic location inversion models are used for inversion. The multiple magnetic location inversion models include at least a magnetic dipole model, a cylinder model, and a finite-length magnetic anomaly model. The magnetic dipole model is used for characterization. The cylinder model is used for characterization. The finite-length magnetic anomaly model is used for characterization. Based on the inversion accuracy and model complexity of each magnetic positioning inversion model, a target magnetic positioning inversion model is selected from the plurality of magnetic positioning inversion models. Based on the inversion results of the target magnetic positioning inversion model, the location and burial depth of the magnetic target in the anomaly candidate region are determined.

[0013] Furthermore, the method of performing magnetic target localization inversion based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio to obtain the position and burial depth of the magnetic target also includes: Based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio, the location inversion of magnetic targets is performed to obtain the position, burial depth, and uncertainty of the magnetic targets. Based on the uncertainty, calculate the expected information gain of each candidate retest route of the UAV; Based on the expected information gain, a target retest route is selected from multiple candidate retest routes, and the altitude, direction, and survey line density of the target retest route are determined. The position and burial depth of the magnetic target are corrected based on the retest data obtained by the UAV during its flight along the target retest route.

[0014] Furthermore, embodiments of this specification provide a magnetic detection and positioning system; the magnetic detection and positioning system includes an airborne end and a ground end; the airborne end includes a drone and a quantum magnetic detection module installed on the drone; the quantum magnetic detection module and the magnetic source of the drone body satisfy a preset distance constraint, and the quantum magnetic detection module includes multiple quantum magnetic sensors distributed in a regular tetrahedral pattern; the ground end is used to perform the above-mentioned low-altitude drone magnetic detection method for positioning magnetic targets.

[0015] As can be seen from the technical solutions provided in the embodiments of this specification above, the embodiments of this specification can effectively reduce the dynamic interference of the UAV's own motors, batteries, and other components on magnetic field measurements and improve data accuracy by limiting the quantum magnetic detection module and the magnetic source of the UAV to meet the preset distance constraints. Using multiple quantum magnetic sensors distributed in a tetrahedral pattern can avoid the background noise of classical magnetic sensors, thereby simultaneously acquiring multi-point spatial magnetic field information, laying the foundation for magnetic gradient tensor calculation. Furthermore, by using positioning, altitude, and attitude data to determine the three-dimensional measurement point trajectories of each quantum magnetic sensor, and compensating for the magnetic field data accordingly, measurement distortion caused by attitude changes and altitude fluctuations during UAV flight can be eliminated. Finally, based on the magnetic gradient tensor, tensor norm, and vertical coupling ratio, inversion can be performed, comprehensively utilizing the spatial variation characteristics of the magnetic field to stably and accurately determine the position and burial depth of the magnetic target. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the accompanying drawings used in the description of the embodiments or prior art will be briefly introduced below.

[0017] Figure 1 This is a schematic diagram of the structural composition of a magnetic detection and positioning system provided in the embodiments of this specification; Figure 2 This is a flowchart of a low-altitude UAV magnetic detection method for magnetic target localization provided in the embodiments of this specification.

[0018] Explanation of reference numerals in the attached diagram: 1. Unmanned Aerial Vehicle (UAV); 2. Quantum Magnetic Probe Module; 3. Positioning Module; 4. Altitude Measurement Module; 5. Camera Module; 6. Differential Positioning Module; 7. Data Transmission Module; 8. Image Transmission Module; 9. UAV Remote Control Module; 10. Data Processing Module. Detailed Implementation

[0019] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0020] It should be noted that the terms "first," "second," etc., used in this specification, claims, and the foregoing drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0021] This specification provides a magnetic detection and positioning system, as illustrated in the embodiments below. Figure 1 As shown.

[0022] In some embodiments, the magnetic detection and positioning system includes an airborne end and a ground end; the airborne end includes a UAV 1 and a quantum magnetic detection module 2 installed on the UAV 1; the quantum magnetic detection module 2 and the magnetic source of the UAV 1 satisfy a preset distance constraint; the quantum magnetic detection module 2 includes a plurality of quantum magnetic sensors distributed in a regular tetrahedral pattern.

[0023] The airborne end carries the detection payload and flies at low altitude along a preset route, while the ground end receives data transmitted back from the airborne end and performs data processing and inversion positioning. The airborne end includes a UAV 1 and a quantum magnetic detection module 2 installed on the UAV 1. The UAV 1 carries the quantum magnetic detection module 2 and moves along the flight measurement line. The UAV 1 platform can be, but is not limited to, a quadcopter, a hexacopter, or a vertical takeoff and landing compound wing. The magnetic source of the quantum magnetic detection module 2 and the UAV 1 body meets a preset distance constraint. This preset distance constraint is used to characterize the predetermined interval between the quantum magnetic detection module 2 and the motor, battery, and metal structural components of the UAV 1, so as to reduce the influence of the body's magnetic interference on the magnetic field measurement. The quantum magnetic detection module 2 includes multiple quantum magnetic sensors distributed in a tetrahedral pattern. The tetrahedral distribution means that the four quantum magnetic sensors are located at the four vertices of a tetrahedron, such that the spatial distance between any two sensors is equal. The quantum magnetic sensors can be magnetometers based on diamond nitrogen-vacancy color centers or atomic magnetometers based on the principle of spinless exchange relaxation, used to synchronously collect the magnetic field components in three orthogonal directions at the measurement point in space. By using preset distance constraints and a regular tetrahedral array, the problems of large magnetic interference and insufficient single-point measurement information in existing technologies were solved, achieving low machine interference and synchronous acquisition of multiple spatial points, providing the original data foundation for subsequent high-precision positioning.

[0024] In some embodiments, the quantum magnetic detection module 2 includes N quantum magnetic sensors, which are distributed in a regular tetrahedral pattern, with each quantum magnetic sensor located at a vertex of the regular tetrahedron, and N is 4.

[0025] Four quantum magnetic sensors are labeled as sensor 1, sensor 2, sensor 3, and sensor 4. A regular tetrahedron is used to represent that the straight-line distance between any two vertices is equal, and the distance from each vertex to the geometric center of the tetrahedron is equal. Each quantum magnetic sensor synchronously outputs three orthogonal magnetic field components at its respective vertex, denoted as the north component, east component, and vertical component. The four quantum magnetic sensors form a symmetrical three-dimensional array in space, enabling the acquisition of three-dimensional magnetic field vectors at four different spatial locations at the same sampling time. The quantum magnetic detection module 2 also includes a synchronization signal acquisition board, which provides a unified trigger signal and a unified timestamp to the four quantum magnetic sensors, ensuring that the four sensors complete data acquisition within the same sampling period. Based on the strict regular tetrahedral distribution, the problems of unclear sensor spatial relationships and large tensor solution errors in existing technologies are solved. By simplifying the difference calculation formula of the magnetic gradient tensor through symmetrical geometric relationships, the sensitivity of sensor position errors to tensor solution is reduced, thereby improving the accuracy and stability of the full tensor solution.

[0026] In some embodiments, the quantum magnetic detection module 2 includes N quantum magnetic sensors, which are arranged in an approximately tetrahedral spatial array, where N is an integer greater than or equal to 4.

[0027] An approximate regular tetrahedron is used to characterize that the edge lengths of the spatial tetrahedron formed by the four sensors are not exactly equal, but the deviation is controlled within a preset engineering tolerance range; for example, the relative deviation between any two edge lengths does not exceed five percent. The quantum magnetic detection module 2 also includes a sensor coordinate calibration unit, used to actually measure the precise spatial coordinates of each quantum magnetic sensor in the UAV 1 body coordinate system after installation and record them as calibration coordinates. In subsequent data processing, the magnetic gradient tensor is calculated based on the calibration coordinates of each quantum magnetic sensor obtained from the actual calibration, without assuming that the distances between the sensors are strictly equal. The approximate regular tetrahedron distribution solves the problem in existing technologies where a strictly regular tetrahedron array cannot be achieved due to processing or installation errors. By calibrating the actual coordinates and solving the tensor accordingly, the requirements for machining accuracy and installation accuracy are reduced, making the solution easier to implement in engineering and mass-produce, while maintaining high tensor calculation accuracy.

[0028] In some embodiments, the quantum magnetic detection module 2 includes N quantum magnetic sensors, which are distributed in three-dimensional space using other geometric configurations that are not tetrahedral; the other geometric configurations include, but are not limited to, cube arrays, triangular pyramid arrays or cross-shaped three-dimensional arrays, where N is an integer greater than or equal to 4.

[0029] Other geometric configurations include cubic arrays, triangular pyramid arrays, or cross-shaped solid arrays. A cubic array represents eight quantum magnetic sensors located at the eight vertices of a cube. A triangular pyramid array represents four quantum magnetic sensors located at the four vertices of a triangular pyramid, with a triangular base and vertices above the base. A cross-shaped solid array represents multiple quantum magnetic sensors located on three orthogonal axes in space, forming a three-dimensional cross distribution. Regardless of the geometric configuration, the array must be able to simultaneously acquire triaxial magnetic field data from at least four spatial measurement points in the target area, and based on this data, all independent components of the magnetic gradient tensor or equivalent tensor invariants can be solved. When using more than four sensors, weighted least squares or robust estimation methods are used to improve the noise resistance of tensor solutions by utilizing redundant data. Non-tetrahedral configurations address the special space constraints of different UAV platforms, providing a variety of flexible array layout schemes, enabling the system to adapt to UAV platforms of different sizes and payload capacities, thus expanding the applicability and scenario adaptability of the technology.

[0030] In some embodiments, the airborne terminal further includes a positioning module 3, an altitude measurement module 4, and a camera module 5; the positioning module 3 is disposed on the top of the UAV 1 platform and is used to obtain the latitude and longitude position of the UAV 1; the altitude measurement module 4 is disposed on the lower part of the UAV 1 platform and is used to measure the flight altitude of the UAV 1 relative to the ground; the camera module 5 is disposed on the part of the UAV 1 platform facing the ground and is used to collect images of the target area below the flight.

[0031] The positioning module 3 is installed on the top of the UAV 1 platform in an unobstructed position to acquire latitude and longitude position data of the UAV 1 at various sampling times during flight. Positioning module 3 represents a device that receives global navigation satellite signals and works with ground differential base stations to achieve centimeter-level positioning; its components include a satellite signal receiving antenna and a real-time dynamic differential calculation chip. The altimeter module 4 is installed on the lower part of the UAV 1 platform, with its detection direction pointing vertically downwards towards the ground, to acquire the real-time flight altitude of the UAV 1 relative to the ground surface directly below. Altimeter module 4 represents a device that measures vertical distance by emitting laser pulses and receiving ground echoes; it may include a laser rangefinder, which includes a laser emitter, a receiving optical system, and a timing circuit, or alternatively, a radar altimeter or an ultrasonic altimeter. The camera module 5 is installed on the ground-facing part of the UAV 1 platform, with its optical axis aligned with the vertical direction of the UAV 1, to acquire visible light or infrared images of the target area below it during flight. Camera module 5 represents a sensor component that converts optical images into digital image signals. It comprises an optical lens, an image sensor, and an image compression coding unit. The dual-light camera simultaneously includes a visible light channel and a thermal imaging channel. The integration of positioning module 3, altimeter module 4, and camera module 5 solves the problem of spatiotemporal reference separation for multi-source data. By using a unified timestamp, it achieves synchronous recording of magnetic field data, spatial location, altitude above ground, and surface imagery, providing an altitude reference for subsequent attitude-altitude joint compensation and image-based auxiliary data for target confirmation.

[0032] In some embodiments, the ground terminal includes a differential positioning module 6, a data transmission receiving module 7, an image transmission receiving module 8, a UAV remote control module 9, and a data processing module 10. The differential positioning base station forms a differential positioning link with the UAV 1's onboard positioning module 3, used to provide positioning correction information to the UAV 1. The data transmission receiving module is communicatively connected to the data processing module 10, used to receive magnetic field data, positioning data, altitude measurement data, and attitude data transmitted back by the UAV 1. The image transmission receiving module 8 is communicatively connected to the data processing module 10, used to receive surface images of the target area transmitted back by the UAV 1. The UAV remote control module 9 is communicatively connected to the UAV 1, used to control the flight of the UAV 1 or input flight mission parameters. The data processing module 10 is used to synchronize, compensate, map, invert, and display the received data.

[0033] The differential positioning module 6 may include a differential positioning base station, deployed in an open, unobstructed location near the area to be measured. It establishes a differential positioning link with the onboard positioning module 3 of the UAV 1, and is used to send positioning correction signals to the UAV 1 to eliminate common errors in satellite positioning. The differential positioning base station represents a fixed ground reference station with known precise coordinates, and its components include a satellite signal receiver, a data processing unit, and a wireless data transmitter. The data transmission module 7 communicates with the data processing module 10 via a serial interface or a universal serial bus, and is used to receive magnetic field data, positioning data, altitude data, and attitude data transmitted back by the UAV 1. The data transmission module 7 represents a wireless data receiver operating in the industrial, scientific, and medical frequency bands, and its components include a receiving antenna, a demodulator, and a data verification unit. The image transmission module 8 communicates with the data processing module 10 via a video capture card or a network interface, and is used to receive surface images of the target area transmitted back by the UAV 1. The image transmission module 8 represents a wireless receiving device for receiving analog or digital video signals, and its components include a receiving antenna, a video decoder, and an image buffer unit. The UAV remote control module 9 communicates with UAV 1 via a radio control link, used for manually controlling the flight attitude of UAV 1 or uploading preset flight mission parameters to UAV 1. The UAV remote control module 9 represents a handheld remote controller, comprising a control joystick, a status display screen, a mode switch, and an RF transmission module. The data processing module 10 runs on a ground computer, used for time synchronization alignment, attitude-altitude-positioning joint compensation, magnetic gradient tensor calculation, physical constraint 2D mapping, multi-model inversion positioning, and result display output of received multi-source data. The data processing module 10 represents an integrated software system, comprising a data receiving interface, a synchronization timing control unit, a compensation algorithm library, a mapping and inversion calculation engine, and a graphical user interface. The collaborative work of the various modules on the ground solves the problem of rapid data processing and target positioning on-site, achieving real-time closed-loop processing from raw data reception to target position and depth output, significantly improving the efficiency and reliability of field exploration operations.

[0034] In some embodiments, the data processing module 10 may include a microcontroller unit (MCU) and a central processing unit (CPU). Of course, the data processing module 10 may also include other devices capable of control functions, such as desktop computers, laptops, and other computer equipment. The data processing module 10 can determine the three-dimensional measurement point trajectory of each quantum magnetic sensor based on the UAV's positioning data, altitude data, and attitude data at each sampling moment during flight; based on the three-dimensional measurement point trajectory, it compensates for the magnetic field data of multiple spatial measurement points synchronously collected by each quantum magnetic sensor at each sampling moment; based on the compensated magnetic field data, it calculates the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio; the magnetic gradient tensor norm is used to characterize the energy intensity of the magnetic gradient; the magnetic gradient vertical coupling ratio is used to characterize the coupling relationship between the horizontal and vertical magnetic gradients; based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio, it performs magnetic target localization inversion to obtain the position and burial depth of the magnetic target.

[0035] In some embodiments, the aforementioned quantum magnetic sensor is preferably a quantum magnetometer based on diamond nitrogen-vacancy color centers. In alternative embodiments, other high-sensitivity quantum magnetic sensors may also be used, as long as they can meet the sensitivity and synchronous sampling requirements for low-altitude ferromagnetic target detection, such as atomic magnetometers based on the spin-free exchange relaxation principle, optically pumped quantum magnetometers, and other quantum magnetic sensors capable of multi-point vector magnetic field component measurement.

[0036] In some embodiments, the quantum magnetic detection module is preferably installed under the UAV fuselage and separated from the fuselage by a mounting rod. In alternative embodiments, other installation methods that can reduce magnetic interference from the UAV and complete flight detection can also be used, such as pod-type installation, flexible suspension installation, rigid extension rod extension installation, belly center installation or offset installation, and other installation methods that can reduce magnetic interference from the UAV fuselage.

[0037] In some embodiments, the height measurement module is preferably a laser altimeter. In alternative embodiments, radar altimeters, ultrasonic altimeters, relative height measurement schemes based on vision or terrain matching, and other devices capable of providing UAV height information relative to the ground may also be used.

[0038] In some embodiments, the camera module described above is preferably a dual-light camera. In alternative embodiments, a single-light camera, a visible light and infrared combined camera, a multispectral camera, a thermal imaging camera, and other image acquisition devices that can be used for auxiliary interpretation of abnormal areas may also be used.

[0039] In some embodiments, the positioning module is preferably a positioning method that combines an airborne satellite positioning module with a ground differential positioning base station. In alternative embodiments, RTK positioning, PPK positioning, inertial navigation and satellite integrated navigation, and other positioning schemes that can provide high-precision spatial positioning information may also be used.

[0040] In some embodiments, the data transmission method described above is preferably a separate wireless data transmission and image transmission method. In alternative embodiments, the following methods may also be used: integrated data and image transmission method, airborne local storage and post-flight export processing method, and a combination of real-time transmission and local backup method.

[0041] In some embodiments, the flight control methods described above are preferably manual detection mode and path planning detection mode. In alternative embodiments, semi-automatic assisted flight mode, preset grid scanning mode, adaptive target tracking and retesting mode, multi-line stitching scanning mode, and low-altitude magnetic detection mode capable of completing the target area may also be used.

[0042] Corresponding to the aforementioned magnetic detection and positioning system, this specification provides an embodiment of a low-altitude UAV magnetic detection method for locating magnetic targets, referring to... Figure 2 As shown. In practice, this may include the following steps: S201: Based on the UAV's positioning data, altitude data, and attitude data at each sampling moment during flight, determine the three-dimensional measurement point trajectory of each quantum magnetic sensor.

[0043] In some embodiments, step S201 may specifically include: determining the geographic coordinates of the UAV at each sampling time based on positioning data; converting the mounting arm length of each quantum magnetic sensor from the body coordinate system to the geographic coordinate system based on attitude data to obtain the offset of each quantum magnetic sensor relative to the UAV in the geographic coordinate system; the mounting arm length is used to characterize the fixed spatial offset of each quantum magnetic sensor relative to the UAV; fusing the geographic coordinates of the UAV at each sampling time with the offset of each quantum magnetic sensor to obtain the geographic coordinates of each quantum magnetic sensor at each sampling time; fusing the geographic coordinates of each quantum magnetic sensor at multiple sampling times in chronological order to obtain the three-dimensional measurement point trajectory of each quantum magnetic sensor.

[0044] This step converts the UAV's own motion and attitude information into a continuous motion path for each quantum magnetic sensor in real geographic space. Positioning data, provided by the onboard satellite positioning module in conjunction with ground-based differential base stations, characterizes the longitude, latitude, and ellipsoidal altitude of the UAV's center point at each sampling time. Altitude data characterizes the vertical distance of the UAV relative to the ground directly below. Attitude data, represented as quaternions, characterizes the rotation of the UAV's body coordinate system relative to the geographic coordinate system. Quantum magnetic sensors, mounted on a support rod beneath the UAV, represent magnetic field measurement elements; each sensor has a fixed mounting arm length vector relative to the UAV's center. The fused three-dimensional measurement point trajectory characterizes the spatial coordinate sequence of each quantum magnetic sensor during flight, including eastward coordinates, northward coordinates, and altitude.

[0045] In some embodiments, the above-mentioned fusion of the geographic coordinates of the UAV at each sampling time with the offset of each quantum magnetic sensor to obtain the geographic coordinates of each quantum magnetic sensor at each sampling time may specifically include: directly adding the geographic coordinates of the UAV at each sampling time to the geographic coordinate system offset of each quantum magnetic sensor at the corresponding time to obtain the geographic coordinates of each quantum magnetic sensor at each sampling time.

[0046] Based on the positioning data, the geographic coordinates of the UAV's center point at each sampling time can be extracted. These geographic coordinates include eastward coordinates, northward coordinates, and ellipsoidal altitude. Based on the attitude data, the mounting arm length vectors of each quantum magnetic sensor are transformed from the body coordinate system to the geographic coordinate system, obtaining the geographic coordinate system offset of each quantum magnetic sensor relative to the UAV's center point. The mounting arm length characterizes the fixed spatial offset of the quantum magnetic sensor in the UAV's body coordinate system, including projection components on the UAV's forward, rightward, and downward axes. The geographic coordinates of the UAV at each sampling time are directly added to the corresponding geographic coordinate system offsets of each quantum magnetic sensor to obtain the geographic coordinates of each quantum magnetic sensor at each sampling time. Finally, the geographic coordinates of each quantum magnetic sensor at all sampling times are sequentially connected according to the chronological order of the sampling times to form the three-dimensional measurement point trajectory of that quantum magnetic sensor.

[0047] In some embodiments, the above-mentioned fusion of the geographic coordinates of the UAV at each sampling time with the offset of each quantum magnetic sensor to obtain the geographic coordinates of each quantum magnetic sensor at each sampling time may further include: converting the mounting arm length of each quantum magnetic sensor to the geographic coordinate system based on the attitude data at each sampling time to obtain the offset; adding the geographic coordinates of the UAV to the offset, and performing a moving average filter on the summation results of adjacent sampling times to obtain the geographic coordinates of each quantum magnetic sensor at each sampling time.

[0048] Based on the attitude data at each sampling time, the mounting arm length vector of each quantum magnetic sensor is transformed to the geographic coordinate system, obtaining the geographic coordinate system offset of each quantum magnetic sensor. The geographic coordinates of the UAV are added to the corresponding offset to obtain the preliminary geographic coordinates of each quantum magnetic sensor. For the preliminary geographic coordinate sequence of each quantum magnetic sensor, a moving average filter is applied along the time axis, i.e., a weighted average of the coordinate values ​​at three adjacent sampling times is performed, with the current time having a higher weight than the times before and after it, resulting in the filtered geographic coordinates. The filtered geographic coordinates at each sampling time are used as the final geographic coordinates of the quantum magnetic sensor at each sampling time, and then connected in chronological order to form a three-dimensional measurement point trajectory. The moving average filter is used to reduce random noise in the positioning data and high-frequency jitter in attitude measurement, making the trajectory smoother.

[0049] In some embodiments, the above-mentioned fusion of the geographic coordinates of the UAV at each sampling time with the offsets of each quantum magnetic sensor to obtain the geographic coordinates of each quantum magnetic sensor at each sampling time may further include: constructing a state equation with the geographic coordinates of the UAV, attitude data, and the mounting arm length of each quantum magnetic sensor as input, wherein the state equation is used to characterize the recursive relationship of the spatial position of each quantum magnetic sensor between adjacent sampling times; using the positioning data and altimetry data at each sampling time as observation values, establishing an observation equation between the spatial position of each quantum magnetic sensor and the observation values; predicting the predicted position and its prediction uncertainty of each quantum magnetic sensor at the current sampling time based on the optimal estimate of the spatial position of each quantum magnetic sensor at the previous sampling time through the state equation; calculating the update gain according to the principle of minimizing the estimation error covariance based on the predicted position, prediction uncertainty, and the observation values ​​at the current sampling time; using the update gain, weightedly fusing the predicted position with the observation values ​​at the current sampling time to obtain the optimal estimate of the spatial position of each quantum magnetic sensor at the current sampling time; and using the optimal estimate obtained recursively at each sampling time as the geographic coordinates of each quantum magnetic sensor at each sampling time.

[0050] A state equation can be constructed using the UAV's geographic coordinates, attitude data, and the arm lengths of each quantum magnetic sensor as input. This state equation characterizes the recursive relationship of the spatial positions of each quantum magnetic sensor between adjacent sampling times, specifically by establishing a state transition matrix for position and velocity based on a uniform motion model. Using the positioning and altitude data at each sampling time as observations, an observation equation is established between the spatial positions of each quantum magnetic sensor and the observed values. This observation equation maps the state variables to the measurement space and considers measurement noise. Based on the optimal estimate of the spatial position of each quantum magnetic sensor and its estimated covariance at the previous sampling time, the predicted position and predicted covariance of each quantum magnetic sensor at the current sampling time are predicted using the state equation. Based on the predicted position, predicted covariance, and the observed value and observation noise covariance at the current sampling time, an updated gain matrix is ​​calculated according to the principle of minimizing the estimation error covariance. Using this updated gain, the predicted position and the observed value at the current sampling time are weighted and fused to obtain the optimal estimate of the spatial position of each quantum magnetic sensor at the current sampling time and its updated covariance. The optimal estimates obtained recursively at each sampling time are used as the geographic coordinates of each quantum magnetic sensor at each sampling time, and then connected in chronological order to form a three-dimensional measurement point trajectory. The state recursion utilizes the system's dynamic model, and the observation update utilizes real-time measurement data. It can adaptively balance the relationship between model predictions and measured data, significantly improving the robustness and continuity of trajectory estimation under abnormal conditions such as brief loss of positioning signal or sudden attitude changes.

[0051] S202: Based on the three-dimensional measurement point trajectory, the magnetic field data of multiple spatial measurement points synchronously collected by each quantum magnetic sensor at each sampling time are compensated.

[0052] In some embodiments, step S202 may specifically include: extracting the actual spatial coordinates of each quantum magnetic sensor in the geographic coordinate system at each sampling time based on the three-dimensional measurement point trajectory of each quantum magnetic sensor; the actual spatial coordinates include horizontal position and altitude above the ground; performing height normalization compensation on the magnetic field data based on the comparison result between the actual altitude above the ground of each quantum magnetic sensor and the preset reference altitude; and performing horizontal position registration compensation on the magnetic field data based on the comparison result between the actual horizontal position of each quantum magnetic sensor and the horizontal position of the UAV, as well as the attitude data of the UAV.

[0053] The three-dimensional measurement point trajectory is used to characterize the true spatial coordinates of each quantum magnetic sensor at each sampling moment, including horizontal position and altitude above the ground. Magnetic field data is used to characterize the magnetic induction intensity components in three orthogonal directions measured by each quantum magnetic sensor based on quantum effects (such as energy level splitting of diamond nitrogen-vacancy color centers or atomic spin precession). The compensation process includes two sub-steps: altitude normalization compensation and horizontal position registration compensation. Altitude normalization compensation corrects magnetic field data measured at different altitudes to the same reference altitude, eliminating the attenuation or enhancement effect of magnetic anomaly amplitude caused by altitude changes. Horizontal position registration compensation normalizes the magnetic field data measured due to sensor horizontal offset caused by attitude tilt to a position directly below the UAV's center point, eliminating the abnormal pattern offset introduced by horizontal offset. By jointly utilizing the altitude and horizontal information in the three-dimensional measurement point trajectory, a unified spatial reduction of magnetic field data is achieved, providing more consistent input data for subsequent tensor calculations.

[0054] In some embodiments, the above-mentioned height normalization compensation of the magnetic field data based on the comparison results between the actual ground height of each quantum magnetic sensor and the preset reference height may specifically include: calculating the ratio of the actual ground height of each quantum magnetic sensor to the preset reference height, and multiplying the magnetic field data collected by the quantum magnetic sensor by the cube of the ratio according to the inverse cubic relationship of magnetic field dipole attenuation to obtain the magnetic field data corrected to the preset reference height.

[0055] The actual ground-level altitude of each quantum magnetic sensor at each sampling moment is extracted from the three-dimensional measurement point trajectory of each quantum magnetic sensor. This actual ground-level altitude characterizes the vertical distance from the sensor to the ground surface directly below. A fixed nominal value, such as two meters, is preset as the target altitude for all data correction. The ratio of the actual ground-level altitude of each quantum magnetic sensor to the preset reference altitude is calculated. A ratio greater than one indicates that the actual flight altitude is higher than the reference altitude, and a ratio less than one indicates that it is lower than the reference altitude. According to the physical attenuation law of the magnetic dipole field, the magnetic anomaly amplitude is inversely proportional to the cube of the distance. The original magnetic field data collected by the quantum magnetic sensor is multiplied by the cube of the ratio of the actual ground-level altitude to the preset reference altitude to obtain the equivalent magnetic field data corrected to the preset reference altitude. This scalar multiplication correction driven by physical laws has low computational cost and maintains the inherent characteristics of magnetic field attenuation. It is suitable for detection scenarios with flat terrain and small altitude fluctuations, and can quickly eliminate the influence of altitude differences on the anomaly amplitude.

[0056] In some embodiments, the above-mentioned height normalization compensation of the magnetic field data based on the comparison results between the actual ground clearance of each quantum magnetic sensor and the preset reference height may further include: constructing a height deviation sequence along the flight trajectory based on the difference between the actual ground clearance of each quantum magnetic sensor and the preset reference height; performing empirical mode decomposition on the height deviation sequence to separate low-frequency components characterizing terrain undulations and high-frequency components characterizing attitude fluctuations; compensating the magnetic field data using a preset geomagnetic field vertical gradient model for the low-frequency components; compensating the magnetic field data using a magnetic dipole decay model for the high-frequency components; and superimposing the low-frequency compensation results with the high-frequency compensation results to obtain the height-normalized compensated magnetic field data.

[0057] The actual ground-level altitude of each quantum magnetic sensor at each sampling moment is extracted from the three-dimensional measurement point trajectory of each quantum magnetic sensor, and the altitude deviation value is obtained by subtracting it from the preset reference altitude. The altitude deviation values ​​at all sampling moments are arranged in chronological order to construct an altitude deviation sequence along the flight trajectory. Empirical mode decomposition (EMD) is performed on this altitude deviation sequence, decomposing it into several intrinsic mode function components and a residual component. Through spectral analysis or zero-crossing rate identification, components with frequencies below the preset cutoff frequency are classified as low-frequency components representing terrain undulations, and components with frequencies above the preset cutoff frequency are classified as high-frequency components representing attitude fluctuations or airflow disturbances. For altitude changes caused by low-frequency components, a pre-established geomagnetic field vertical gradient model is used for compensation. The geomagnetic field vertical gradient model is used to characterize the coefficient of linear variation of the geomagnetic field background value with altitude in the region, and is obtained through ground measurements or regional geomagnetic maps. One specific compensation method is to multiply the altitude change corresponding to the low-frequency component by the vertical gradient coefficient to obtain the background field change, and then subtract this change from the raw magnetic field data collected by the quantum magnetic sensors. To compensate for altitude variations caused by high-frequency components, a magnetic dipole attenuation model is employed. This model scales the quantum magnetic anomaly field by the cube of the ratio of the actual altitude to the reference altitude, where the quantum magnetic anomaly field is obtained by subtracting the background magnetic field from the original quantum magnetic field data. The low-frequency compensation results are then superimposed with the high-frequency compensation results to obtain the complete quantum magnetic field data after altitude normalization compensation. By separating terrain undulations and flight disturbances, and employing background field gradient compensation and anomaly field attenuation compensation respectively, the altitude correction accuracy is improved in complex terrain and turbulent environments.

[0058] In some embodiments, the above-mentioned horizontal position registration compensation of the magnetic field data based on the comparison results between the actual horizontal position of each quantum magnetic sensor and the horizontal position of the UAV, as well as the attitude data of the UAV, may specifically include: calculating the horizontal offset vector between the actual horizontal position of each quantum magnetic sensor and the horizontal position of the center point of the UAV, and weighting the magnetic field data collected by each quantum magnetic sensor inverse distance according to the horizontal offset vector to the horizontal position of the center point of the UAV, so as to complete the horizontal position registration compensation.

[0059] The actual horizontal position of each quantum magnetic sensor at each sampling moment is extracted from the three-dimensional measurement point trajectory of each quantum magnetic sensor. This actual horizontal position includes east and north coordinates. The horizontal position of the UAV center point at the same sampling moment is also extracted. The horizontal offset vector between the actual horizontal position of each quantum magnetic sensor and the horizontal position of the UAV center point is calculated. This horizontal offset vector includes the offset direction and offset distance. Based on this horizontal offset vector, the magnetic field data collected by the corresponding quantum magnetic sensor is normalized to the horizontal position of the UAV center point using an inverse distance weighting method. Inverse distance weighting is used to characterize that the larger the offset distance of a sensor, the smaller the weight of its data in the normalization process; the smaller the offset distance of a sensor, the larger the weight of its data. Specifically, for the magnetic field value to be determined at the UAV center point, a weighted average of the measured magnetic field values ​​of the surrounding quantum magnetic sensors is calculated. The weight coefficient of each sensor is the square of the inverse of its offset distance, and then normalized.

[0060] In some embodiments, the above-mentioned horizontal position registration compensation of the magnetic field data based on the comparison results between the actual horizontal positions of each quantum magnetic sensor and the horizontal position of the UAV, and the attitude data of the UAV, may further include: constructing a spatial coordinate system transformation matrix based on the horizontal offset vectors of each quantum magnetic sensor and the attitude data of the UAV, so as to transform the coordinates of the measurement points of each quantum magnetic sensor in the body coordinate system to a horizontal reference system with the horizontal projection center of the UAV as the origin; constructing an objective function with the transformed measured magnetic field data of each quantum magnetic sensor as a constraint, the objective function including: a weighted residual sum of squares between the measured magnetic field values ​​of each sensor and the magnetic field vector of the center point of the horizontal reference system to be determined, and a magnetic field smoothing constraint term based on the assumption of continuous spatial distribution of the magnetic field; solving for the optimal estimate of the magnetic field vector that minimizes the objective function value; and using the optimal estimate as the magnetic field data after horizontal position registration compensation.

[0061] The actual horizontal position and horizontal offset vector of each quantum magnetic sensor at each sampling moment are extracted from the three-dimensional measurement point trajectory of each quantum magnetic sensor. Attitude data of the UAV at the same sampling moment are obtained. Based on the horizontal offset vector and attitude data of each quantum magnetic sensor, a spatial coordinate system transformation matrix is ​​constructed. This transformation matrix is ​​used to transform the coordinates of the measurement points of each quantum magnetic sensor in the body coordinate system to a horizontal reference system with the UAV's horizontal projection center as the origin, eliminating changes in the sensor's horizontal position caused by roll and pitch motions. An objective function is constructed using the transformed measured magnetic field data of each quantum magnetic sensor as constraints. This objective function contains two terms: the first term is the weighted sum of squared residuals between the measured magnetic field values ​​of each sensor and the magnetic field vector at the center point of the horizontal reference system to be determined, where the weighting coefficients are negatively correlated with the sensor offset distance; the second term is a magnetic field smoothing constraint term based on the assumption of continuous spatial distribution of the magnetic field, used to penalize drastic jumps in the magnetic field vector at the center point between adjacent sampling moments. The objective function is solved using gradient descent or Newton's method, and its value is minimized through iterative updates. The solved magnetic field vector is used as the optimal estimate at the horizontal position of the UAV's center point, and is output as the magnetic field data after horizontal position registration and compensation. By introducing smoothing constraints and performing overall optimization over all sampling times, abnormal fluctuations in the compensation results caused by single-point measurement noise or attitude changes can be suppressed, significantly improving the spatial consistency and noise robustness of the compensated magnetic field data.

[0062] S203: Based on the compensated magnetic field data, calculate the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio; the magnetic gradient tensor norm is used to characterize the energy intensity of the magnetic gradient; the magnetic gradient vertical coupling ratio is used to characterize the coupling relationship between the horizontal magnetic gradient and the vertical magnetic gradient.

[0063] In some embodiments, step S203 may specifically include: calculating multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron at each sampling time based on the compensated magnetic field data at each sampling time and the relative spatial positions of each spatial measurement point at that sampling time in the tetrahedron; each tensor component is used to characterize the rate of change of a magnetic field component along a specific spatial direction; and calculating the magnetic gradient tensor norm and magnetic gradient vertical coupling ratio corresponding to that sampling time based on each tensor component of the magnetic gradient tensor at each sampling time.

[0064] The compensated magnetic field data is used to characterize the triaxial magnetic induction intensity values, which have undergone high normalization and horizontal position registration. Each sampling time corresponds to the triaxial magnetic field values ​​of four spatial measurement points. The magnetic gradient tensor is used to characterize the first-order spatial rate of change of the magnetic field vector in three-dimensional space, and is composed of multiple tensor components. The magnetic gradient tensor norm is used to characterize the energy intensity of the magnetic gradient tensor, i.e., the square root of the sum of the squares of all tensor components, reflecting the total drastic degree of spatial change of the magnetic field, and is related to the target magnetic moment and distance. The magnetic gradient vertical coupling ratio is used to characterize the coupling relationship between the horizontal and vertical magnetic gradients, specifically defined as the ratio of the horizontal magnetic gradient amplitude to the vertical magnetic gradient amplitude, which can reflect the relative relationship between the target's horizontal offset and burial depth. By calculating the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio based on quantum magnetic sensor data, the problem of converting discrete measurement point magnetic field values ​​into a continuous tensor field is solved. The above tensor features compress the original four measurement point data into a tensor feature sequence with clear physical meaning, providing more robust input parameters for subsequent inversion.

[0065] In some embodiments, the above-mentioned calculation of multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron at each sampling time based on the compensated magnetic field data at each sampling time and the relative spatial positions of each spatial measuring point at that sampling time in the tetrahedron can specifically include: establishing a system of linear equations between each component of the magnetic gradient tensor and the magnetic field difference between the measuring points based on the compensated magnetic field data of multiple spatial measuring points and their spatial coordinates in the tetrahedron array; and solving the system of linear equations using the least squares method to obtain multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron.

[0066] Based on the triaxial magnetic field data of the four spatial measurement points after compensation at each sampling time, and the spatial coordinates of each measurement point relative to the geometric center of the tetrahedral array, a system of linear equations can be established to represent the magnetic field differences between the components of the magnetic gradient tensor and the measurement points. This system of linear equations characterizes the linear relationship between the tensor components to be determined at the geometric center of the array and the measured magnetic field values ​​at the four measurement points, with each measurement point contributing three equations. The least squares method is used to solve this system of linear equations, i.e., to find the tensor component values ​​that minimize the sum of squared residuals of the system. The solved values ​​are used as multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron, including the northward component along the northward guide, the northward component along the eastward guide, the northward component along the vertical guide, the eastward component along the northward guide, the eastward component along the eastward guide, the eastward component along the vertical guide, the vertical component along the northward guide, the vertical component along the eastward guide, and the vertical component along the vertical guide, with a total of six independent components.

[0067] In some embodiments, the calculation of multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron at each sampling time based on the compensated magnetic field data at each sampling time and the relative spatial positions of each spatial measuring point at that sampling time in the tetrahedron may further include: establishing an initial linear equation set for each component of the magnetic gradient tensor based on the compensated magnetic field data and spatial coordinates of multiple spatial measuring points; introducing physical constraints of zero magnetic field divergence and zero curl, and adding the physical constraints as additional equations to the initial linear equation set to construct a constrained overdetermined equation set; solving the overdetermined equation set using the Lagrange multiplier method, so that the tensor components obtained by the solution simultaneously satisfy the magnetic field measurement difference relation and the constraints of Maxwell's equations in the passive region, thereby obtaining multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron.

[0068] Based on the triaxial magnetic field data of multiple spatial measurement points after compensation at each sampling time and their spatial coordinates in a tetrahedral array, an initial linear equation set for each component of the magnetic gradient tensor is established. Two physical constraints are introduced for the passive region of the magnetic field: the divergence of the magnetic field is zero, and the curl of the magnetic field is zero. The zero divergence constraint is manifested in the fact that the sum of the three diagonal tensor components is equal to zero, that is, the sum of the northward component along the northward guide, the eastward component along the eastward guide, and the vertical component along the vertical guide is zero. The zero curl constraint is manifested in the symmetry of the off-diagonal components, that is, the northward component along the eastward guide is equal to the eastward component along the northward guide, the northward component along the vertical guide is equal to the vertical component along the northward guide, and the eastward component along the vertical guide is equal to the vertical component along the eastward guide. These physical constraints are added as additional equations to the initial linear equation set to construct a constrained overdetermined equation set, the number of which exceeds the number of tensor components to be determined. The overdetermined equations are solved using the Lagrange multiplier method. Lagrange multipliers are introduced to couple the constraints with the objective function. By solving the augmented linear system, tensor component values ​​that simultaneously satisfy the magnetic field measurement difference relation and the constraints of Maxwell's equations in the passive region are obtained. By forcing the divergence and curl to be zero, the solved tensor components possess physical consistency, improving the reliability of the tensor parameters and the stability of subsequent inversion.

[0069] In some embodiments, the above-mentioned calculation of the magnetic gradient tensor norm and magnetic gradient vertical coupling ratio corresponding to each sampling moment based on each tensor component of the magnetic gradient tensor at each sampling moment may specifically include: squaring each of the multiple tensor components of the magnetic gradient tensor at each sampling moment, summing them, and then taking the square root of the sum to obtain the magnetic gradient tensor norm at that sampling moment; summing the square of the rate of change of the northward component of the magnetic field along the vertical direction and the square of the rate of change of the eastward component of the magnetic field along the vertical direction, taking the square root, and then dividing by the sum of the absolute value of the rate of change of the vertical component of the magnetic field along the vertical direction and a very small positive number to obtain the magnetic gradient vertical coupling ratio at that sampling moment.

[0070] For each sampling time, obtain all independent tensor components of the calculated magnetic gradient tensor. There are six independent tensor components: the northward component along the northward guide, the eastward component along the eastward guide, the vertical component along the vertical guide, the northward component along the eastward guide, the northward component along the vertical guide, and the eastward component along the vertical guide. Square each of these six tensor components and sum them to obtain the sum of the six squared values. Then take the square root of this sum and use the result as the magnetic gradient tensor norm for that sampling time. This norm comprehensively reflects the overall intensity of the magnetic field change in the three directions. Extract the three tensor components: the northward component along the vertical guide, the eastward component along the vertical guide, and the vertical component along the vertical guide. Add the square of the northward component along the vertical guide and the square root of the eastward component along the vertical guide to obtain the horizontal magnetic gradient magnitude. Take the absolute value of the vertical component along the vertical guide and add a very small positive number to prevent the denominator from being zero; a very small positive number is, for example, 10 to the power of -9. Divide the magnitude of the horizontal magnetic gradient by the absolute value of the vertical component with a very small positive number along the vertical derivative, and the resulting ratio is taken as the magnetic gradient vertical coupling ratio at that sampling moment.

[0071] In some embodiments, the calculation of the magnetic gradient tensor norm and magnetic gradient vertical coupling ratio corresponding to each sampling moment based on each tensor component of the magnetic gradient tensor at each sampling moment may further include: performing time-series analysis on each tensor component of the magnetic gradient tensor at each sampling moment to identify and remove outlier tensor components caused by transient perturbations; calculating the initial norm and initial vertical coupling ratio using the retained tensor components; establishing a cross-validation relationship between each tensor component based on the physical constraints that the divergence and curl of the magnetic gradient tensor are zero, and calculating the consistency residual between tensor components based on the cross-validation relationship; determining the confidence weights of the initial norm and initial vertical coupling ratio based on the consistency residuals; and determining the magnetic gradient tensor norm and magnetic gradient vertical coupling ratio at the sampling moment based on the initial norm, initial vertical coupling ratio, and corresponding confidence weights.

[0072] Six independent components of the magnetic gradient tensor at each sampling time are obtained, and a time series for each component is constructed along the time axis. A sliding window analysis is performed on each time series, calculating the mean and standard deviation of the components within the window. Sampling points deviating from the mean by more than three times the standard deviation are marked as outlier tensor components, and these outliers are removed from the component set at that sampling time. Using the retained tensor components, the initial norm is calculated using the square root of the square, and the initial vertical coupling ratio is calculated by dividing the horizontal amplitude by the vertical amplitude. Simultaneously, based on the physical constraints of zero divergence and zero curl of the magnetic gradient tensor, a cross-validation relationship is established between the tensor components. Specifically, the zero divergence constraint requires the sum of the three diagonal components to be close to zero, and the diagonal consistency residual is calculated accordingly. The zero curl constraint requires the difference between each pair of off-diagonal components to be close to zero, and three sets of symmetry residuals are calculated accordingly. The sum of the squares of these four residuals is taken as the consistency residual between the tensor components at that sampling time. Based on the magnitude of the consistency residuals, confidence weights for the initial norm and initial vertical coupling ratio are determined: high confidence weights are set when the consistency residuals are less than a low threshold, and low confidence weights are set when the consistency residuals are greater than a high threshold. The initial norm and initial vertical coupling ratio are multiplied by their respective confidence weights, and then divided by the sum of the weights to obtain the weighted adjusted norm and vertical coupling ratio, which serve as the final magnetic gradient tensor norm and magnetic gradient vertical coupling ratio at that sampling time. By dynamically adjusting the confidence levels of the feature quantities through temporal outlier removal and physical consistency checks, the smoothness and robustness of the feature sequence are improved.

[0073] S204: Based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio, the location and burial depth of the magnetic target are obtained by performing magnetic target localization inversion.

[0074] In some embodiments, step S204 may specifically include: generating a magnetic anomaly distribution map of the target region using a magnetic source physical prior model based on the magnetic gradient tensor, the magnetic gradient tensor norm, and the magnetic gradient vertical coupling ratio; the magnetic source physical prior model is used to characterize the attenuation relationship of the theoretical tensor features generated by the equivalent magnetic source with spatial distance; determining one or more anomaly candidate regions of the target region based on the magnetic anomaly distribution map; and performing magnetic target localization inversion for each anomaly candidate region to obtain the position and burial depth of the magnetic target in the anomaly candidate region.

[0075] The magnetic gradient tensor is used to characterize the spatial rate of change matrix of the magnetic field vector. The magnetic gradient tensor norm is used to characterize the overall magnitude of the tensor. The magnetic gradient vertical coupling ratio is used to characterize the ratio of the horizontal gradient to the vertical gradient. The inversion process involves establishing a forward model and iteratively optimizing it to make the magnetic field characteristics modeled by the model approximate the measured characteristics, thereby estimating the target parameters.

[0076] In some embodiments, the above-mentioned generation of a magnetic anomaly distribution map of the target region using a priori physical model of the magnetic source based on the magnetic gradient tensor, the magnetic gradient tensor norm, and the magnetic gradient vertical coupling ratio may further include: using multiple tensor components of the magnetic gradient tensor, the magnetic gradient tensor norm, and the magnetic gradient vertical coupling ratio as tensor features of the magnetic anomaly distribution map to be generated; constructing an objective function for a two-dimensional grid numerical model to be solved in the target region, wherein the objective function includes: a fitting error term between the tensor feature vector to be solved at each two-dimensional grid point and the measured tensor feature vector at each sampling time, and a spatial smoothing constraint term between two-dimensional grid points. The method includes prior constraints based on a priori physical model of the magnetic source; the measured tensor eigenvector at each sampling time is used to characterize the tensor eigenvector at the geometric center of the tetrahedron at that sampling time; the spatial smoothness constraint is used to characterize the smoothness of the change of tensor eigenvector between adjacent two-dimensional grid points; the prior constraints are used to characterize the degree of deviation between the tensor eigenvector to be solved for each two-dimensional grid and the theoretical tensor eigenvector given by the priori physical model of the magnetic source; the method solves for a two-dimensional grid numerical model that minimizes the objective function; and the method generates a magnetic anomaly distribution map of the target region based on the solved two-dimensional grid numerical model.

[0077] Multiple tensor components of the magnetic gradient tensor, the magnetic gradient tensor norm, and the magnetic gradient vertical coupling ratio can be used together as tensor features of the target graph. These feature values ​​at each sampling time or at each grid point constitute a multidimensional feature vector. A two-dimensional grid is constructed for the target region, with each grid point corresponding to a multidimensional feature vector to be solved, the dimension of which is equal to the number of selected tensor features. An objective function is established, comprising three parts: a fitting error term between the multidimensional feature vector to be solved at each two-dimensional grid point and the measured multidimensional feature vector at each sampling time; a spatial smoothing constraint term between two-dimensional grid points; and a priori constraint term based on the prior physical model of the magnetic source. The fitting error term penalizes the deviation between the grid interpolation result and the measured data, using a weighted sum of squares, independently weighted for each feature dimension, or using Mahalanobis distance. The spatial smoothing constraint term penalizes drastic changes in the multidimensional feature vector between adjacent grid points, and can be the sum of squares of the Laplacian operator of each feature component or the square of the vector gradient magnitude. The prior constraint term penalizes the deviation between the unsolved mesh multidimensional eigenvector and the theoretical multidimensional eigenvector given by the prior physical model of the magnetic source. The prior physical model of the magnetic source characterizes the attenuation of theoretical tensor eigenvalues ​​(including corresponding tensor components, tensor norm, and vertical coupling ratio) generated by an equivalent point dipole or magnetized sphere with spatial distance; its parameters (location, burial depth, magnetic moment) are obtained through preliminary analysis. Solving the two-dimensional mesh numerical model that minimizes this objective function can be done using the conjugate gradient method or the direct method. The multidimensional eigenvectors of each mesh point in the solved two-dimensional mesh numerical model are plotted as a two-dimensional image according to the mesh coordinates. This can be done by plotting each feature separately or by merging multiple features into a comprehensive anomaly map, serving as the magnetic anomaly distribution map of the target area. By introducing physical prior constraints, the generated anomaly map better matches the spatial distribution characteristics of the actual magnetic source, improving the accuracy of anomaly localization.

[0078] In some embodiments, the above-mentioned determination of one or more candidate anomalies in the target area based on the magnetic anomaly distribution map may specifically include: performing a global extreme value search on the magnetic anomaly distribution map to extract all local maxima points; taking each local maxima point as the center, searching outward layer by layer until the magnetic anomaly amplitude drops below the maximum value multiplied by a preset proportional threshold, and taking a continuous closed region that meets the descent condition as a candidate anomaly region.

[0079] A global extremum search is performed on the magnetic anomaly distribution map, scanning the magnetic anomaly amplitudes of all grid points to identify all local maxima. A local maximum is defined as a grid point whose amplitude is greater than the amplitudes of all its eight neighboring grid points. Each local maximum is used as a seed point, and the search expands outward layer by layer from that point. During each layer expansion, it is checked whether the amplitudes of adjacent grid points on the current boundary are still higher than the maximum multiplied by a preset proportional threshold, which is set to 50%. When the search reaches a certain layer, if the amplitudes of all adjacent grid points are lower than this product, the expansion stops. A continuous closed region formed by all the searched grid points is marked as an anomaly candidate region. For cases where multiple maximum points are too close together, causing region overlap, the region corresponding to the maximum with the larger amplitude is retained, and the overlapping parts are merged. The region boundary is controlled by the proportional threshold to adapt to the scale changes of targets with different intensities, without requiring prior target size information.

[0080] In some embodiments, the above-mentioned determination of one or more candidate anomalous regions of a target region based on a magnetic anomaly distribution map may further include: performing a multi-scale morphological top-hat transformation on the magnetic anomaly distribution map to extract bright feature regions at different scales; calculating the ratio of the mean of the magnetic gradient tensor norm to the background variance within each bright feature region as the concentration degree, and calculating the spatial gradient modulus of the magnetic gradient vertical coupling ratio within that region as the mutation rate; identifying bright feature regions with a concentration degree exceeding a first preset threshold and a mutation rate exceeding a second preset threshold as candidate anomalous regions; and retaining the scale division corresponding to the largest product of concentration degree and mutation rate for the same spatial region extracted repeatedly at different scales.

[0081] A set of circular or square structural elements of different sizes can be selected, with the diameters of the structural elements increasing sequentially from smallest to largest, taking values ​​of half, one, and two times the expected burial depth, respectively. For each structural element of a certain size, an erosion operation is first performed on the magnetic anomaly distribution map using that structural element, followed by a dilation operation on the erosion result to obtain the opening operation result at that size. Subtracting the opening operation result from the original magnetic anomaly distribution map yields the top-hat transformation map at that size, where the connected regions with positive grayscale values ​​are the bright feature regions extracted at that scale. This process is repeated for all structural elements of different sizes to extract the bright feature regions at each scale.

[0082] For each bright feature region, the ratio of the mean of its internal magnetic gradient tensor norm to the background variance of the entire magnetic anomaly distribution map is calculated. This ratio is used as a concentration index; higher concentration indicates a more focused anomaly. Simultaneously, the average spatial gradient modulus of the vertical coupling ratio of the magnetic gradient within this region is calculated as a mutation rate index; higher mutation rates indicate a stronger coupling between the target's horizontal offset and burial depth. Concentration and mutation rate thresholds are set. Bright feature regions with a concentration exceeding a first preset threshold and a mutation rate exceeding a second preset threshold are identified as anomaly candidate regions.

[0083] For bright feature regions repeatedly extracted at different scales and corresponding to the same spatial region, spatial overlap is used for determination. The product of the concentration and mutation rate of the region at each scale is compared, and the scale division corresponding to the largest product is retained, while repeated regions at other scales are discarded. Multi-scale morphological operations are used to automatically adapt the spatial distribution range of the target, and tensor norm and vertical coupling ratio are combined for secondary screening to improve the detection rate of weak anomalies and small targets and reduce the risk of false anomalies.

[0084] In some embodiments, the above-mentioned magnetic target localization inversion for each anomaly candidate region to obtain the position and burial depth of the magnetic target in the anomaly candidate region may further include: for each anomaly candidate region, performing inversion using multiple magnetic localization inversion models; the multiple magnetic localization inversion models include at least a magnetic dipole model, a cylinder model, and a finite-length magnetic anomaly model; the magnetic dipole model is used to characterize the spatial distribution characteristics of the magnetic gradient tensor generated by approximately spherical or equiaxed magnetic targets; the cylinder model is used to characterize the spatial distribution characteristics of the magnetic gradient tensor generated by long tubular or columnar magnetic targets; the finite-length magnetic anomaly model is used to characterize the spatial distribution characteristics of the magnetic gradient tensor generated by rod-shaped or strip-shaped magnetic targets with finite extension length; based on the inversion accuracy and model complexity of each magnetic localization inversion model, a target magnetic localization inversion model is selected from the multiple magnetic localization inversion models; based on the inversion results of the target magnetic localization inversion model, the position and burial depth of the magnetic target in the anomaly candidate region are determined.

[0085] The magnetic dipole model is used to characterize the spatial distribution of the magnetic gradient tensor generated by approximately spherical or equiaxed magnetic targets. The parameters to be estimated are the target center location, burial depth, and magnetic moment vector. The cylinder model is used to characterize the spatial distribution of the magnetic gradient tensor generated by long tubular or cylindrical magnetic targets. The parameters to be estimated include the cylinder center location, burial depth, length, radius, and magnetization direction. The finite-length magnetic anomaly model is used to characterize the spatial distribution of the magnetic gradient tensor generated by rod-shaped or strip-shaped magnetic targets with finite extension length. The parameters to be estimated fall between those of the magnetic dipole model and the cylinder model.

[0086] Nonlinear optimization is performed on each model, using, but not limited to, the Levenberg-Marquardt algorithm or particle swarm optimization, to minimize the sum of squared residuals between the tensor features derived from the model's forward modeling and the measured features. The sum of squared residuals after optimization for each model, along with the number of parameters to be estimated in that model, is recorded; this number represents the model complexity. Based on the inversion accuracy and model complexity of each model, the Akaike information criterion value or the Bayesian information criterion value is calculated for each model. The Akaike information criterion value is calculated as the logarithm of the sum of squared residuals plus twice the number of parameters; the Bayesian information criterion value is calculated as the logarithm of the sum of squared residuals plus the number of parameters multiplied by the logarithm of the number of sampling points. The magnetic positioning inversion model with the smallest information criterion value is selected as the target magnetic positioning inversion model for the anomaly candidate region. The planar coordinate parameters and vertical depth parameters of the magnetic target are read from the inversion results of this target model, serving as the location and burial depth of the magnetic target in the anomaly candidate region. By running multiple physical models in parallel and automatically selecting the optimal model based on the information criterion, the target shape does not need to be known in advance, improving the adaptability of the inversion and the reliability of the results.

[0087] In some embodiments, the selection of a target magnetic positioning inversion model based on the inversion accuracy and model complexity of each magnetic positioning inversion model may further include: randomly dividing the measured data at each sampling time into a training set and a validation set; using the training set to estimate the parameters of each magnetic positioning inversion model, and then calculating the prediction residuals and their covariance matrix on the validation set; calculating the model complexity penalty term based on the number of parameters to be estimated for each model; determining the spatial range parameter of the residuals based on the spatial autocorrelation analysis of the prediction residuals of each model; using the weighted sum of the root mean square error of the prediction residuals, the model complexity penalty term, and the spatial range parameter as a comprehensive evaluation index; and selecting the magnetic positioning inversion model with the smallest comprehensive evaluation index as the target magnetic positioning inversion model.

[0088] The measured data at all sampling times were randomly divided into a training set and a validation set. The training set was used for model parameter estimation, and the validation set was used to evaluate prediction performance. The ratio of the training set to the validation set was 7:3. Parameter estimation was performed on each magnetic positioning inversion model using the training set. After optimizing the model parameters, these parameters were substituted into the model and used to calculate the theoretical prediction values ​​at the sampling locations in the validation set. The difference between these theoretical predictions and the measured values ​​in the validation set was calculated to obtain the prediction residuals. The root mean square error (RMSE) of the prediction residuals on the validation set was calculated as the prediction accuracy index of the model. Simultaneously, a model complexity penalty term was calculated based on the number of parameters to be estimated in each model. For example, the penalty form in the Bayesian information criterion was used, which is the number of parameters multiplied by the natural logarithm of the number of sampling points. Further spatial autocorrelation analysis was performed on the prediction residuals on the validation set: using the residual values ​​as variables, the semi-variogram of the residuals at different spatial distances was calculated; a semi-variogram model (such as a spherical model or an exponential model) was fitted, and the range parameters were extracted from it. The range parameter characterizes the maximum distance at which the residuals maintain spatial correlation. A larger range indicates a large-scale systematic bias in the residuals, while a smaller range indicates stronger randomness in the residuals. The root mean square error of the predicted residuals, the model complexity penalty term, and the spatial range parameter are multiplied by preset weighting coefficients (e.g., root mean square error weight 0.5, complexity penalty weight 0.3, range parameter weight 0.2), and summed to obtain a comprehensive evaluation index. The comprehensive evaluation index is calculated for all candidate models, and the model with the smallest index value is selected as the target magnetic positioning inversion model. Cross-validation is used to test the model's extrapolation performance, and spatial autocorrelation analysis is used to determine whether there are structural biases in the residuals that are not explained by the model, thus selecting a model with strong generalization ability and good spatial randomness in the residuals.

[0089] In some embodiments, the above method may further include: for each sampling time, constructing a magnetic field prediction model based on the spatial geometric positions of N spatial measurement points in a regular tetrahedron, using magnetic field data from any N-1 spatial measurement points to predict the magnetic field data at the Nth spatial measurement point; where N is an integer greater than or equal to 4; the magnetic field prediction model is used to characterize the spatial distribution relationship of the magnetic field under the constraints of the physical laws of the magnetic field passive region; taking each spatial measurement point as the predicted point, calculating the theoretical magnetic field data at the predicted point using the magnetic field prediction model based on the compensated magnetic field data from the remaining N-1 measurement points; determining the weighting coefficient of the compensated magnetic field data at the predicted point based on the vector difference between the compensated magnetic field data and the theoretical magnetic field data; the weighting coefficient is negatively correlated with the vector difference. Based on this, in step S203 above, the calculation of multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron at each sampling time can be further included: the calculation of multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron at each sampling time based on the compensated magnetic field data at each sampling time and the corresponding weight coefficients, as well as the relative spatial positions of each spatial measurement point at the sampling time in the tetrahedron.

[0090] Based on the spatial geometric positions of N spatial measuring points in a tetrahedral array, a magnetic field prediction model is constructed. This model characterizes the analytical relationship between the magnetic field vector at any spatial location and the magnetic field vectors of surrounding measuring points, under the constraints of the physical laws of the magnetic field source-free region (including zero divergence and zero curl). One specific form of this prediction model is as follows: using the triaxial magnetic field data of any N minus one spatial measuring point, the theoretical triaxial magnetic field data at the Nth spatial measuring point can be uniquely predicted through interpolation based on the Laplace equation or the equivalent source method. Taking each spatial measuring point as the predicted point, the compensated magnetic field data of the remaining N minus one measuring points are substituted into the magnetic field prediction model to calculate the theoretical triaxial magnetic field data at that predicted point. Subtracting the theoretical magnetic field data from the actual compensated magnetic field data at the predicted point yields a three-dimensional vector difference. This vector difference reflects the degree of inconsistency between the measured data and the predicted data based on physical laws at that point; a higher degree of inconsistency indicates a greater influence of local interference or sensor malfunction on the data at that point. The weighting coefficients of the compensated magnetic field data at the predicted point are determined based on the vector difference. These weighting coefficients are negatively correlated with the magnitude of the vector difference; that is, the larger the vector difference, the smaller the weighting coefficients. By utilizing the physical laws of magnetic fields for self-consistency testing, the reliability of the data at each measurement point is quantitatively evaluated, providing a basis for subsequent weighted solutions and thus improving the robustness and anti-interference capability of tensor calculations.

[0091] In some embodiments, determining the weighting coefficient of the compensated magnetic field data at the predicted point based on the vector difference between the compensated magnetic field data and the theoretical magnetic field data at the predicted point may specifically include: calculating the magnitude of the vector difference; comparing the magnitude with a preset first magnitude threshold and a second magnitude threshold, wherein the first magnitude threshold is less than the second magnitude threshold; when the magnitude is less than or equal to the first magnitude threshold, setting the weighting coefficient to a first value; when the magnitude is greater than the first magnitude threshold and less than the second magnitude threshold, setting the weighting coefficient to a second value, wherein the second value is less than the first value; and when the magnitude is greater than or equal to the second magnitude threshold, setting the weighting coefficient to a third value, wherein the third value is less than the second value.

[0092] The magnitude of the vector difference between the measured magnetic field data and the theoretical magnetic field data at each predicted point can be calculated. This magnitude characterizes the total deviation. Two preset magnitude thresholds are defined as a first magnitude threshold and a second magnitude threshold, where the first magnitude threshold is less than the second magnitude threshold. The calculated magnitude is compared with these two thresholds. When the magnitude is less than or equal to the first magnitude threshold, the data quality at that measurement point is considered excellent, and the weighting coefficient is set to a first value, which is one. When the magnitude is greater than the first magnitude threshold and less than the second magnitude threshold, the data quality at that measurement point is considered moderate, and the weighting coefficient is set to a second value, which is less than the first value, for example, 0.5. When the magnitude is greater than or equal to the second magnitude threshold, the data quality at that measurement point is considered poor, and the weighting coefficient is set to a third value, which is less than the second value, for example, 0.1. By assigning values ​​in segments, the continuously changing deviation is discretized into three levels, each level corresponding to a fixed weighting coefficient.

[0093] In some embodiments, determining the weighting coefficient of the compensated magnetic field data at the predicted point based on the vector difference between the compensated magnetic field data and the theoretical magnetic field data at the predicted point may further include: calculating the projection component of the vector difference onto a constrained subspace where the magnetic gradient tensor divergence is zero and the residual component perpendicular to the constrained subspace; calculating an anomaly index based on the ratio of the energy of the vertical residual component to the total energy of the vector difference; setting the weighting coefficient to zero when the anomaly index exceeds a preset upper limit threshold; constructing a Gaussian kernel function based on the magnitude of the vector difference when the anomaly index does not exceed the upper limit threshold; using the value of the Gaussian kernel function as the weighting coefficient, wherein the value of the Gaussian kernel function is a negative exponential decay form with the magnitude of the vector difference as the independent variable.

[0094] The vector difference between the measured magnetic field data and the theoretical magnetic field data at each predicted point can be calculated and projected onto two orthogonal subspaces. The first subspace is a constraint subspace where the magnetic gradient tensor divergence is zero, consisting of all magnetic field vector differences that satisfy the zero divergence condition. The second subspace is the residual subspace perpendicular to the constraint subspace. The ratio of the energy of the projected components of the vector difference onto the residual subspace (i.e., the sum of squares of these components) to the total energy of the vector difference (i.e., the square of the magnitude of the vector difference) is calculated and defined as the anomaly index. The closer the anomaly index is to one, the more likely the difference between the measured and theoretical data originates from components that do not satisfy the zero divergence constraint, which may be caused by non-magnetic source interference or sensor malfunction. A preset upper threshold, such as 0.8, is used. When the anomaly index exceeds this upper threshold, the data at that measurement point is determined to be mainly caused by interference, and its weight coefficient is set to zero, indicating that the data at that measurement point is completely removed. When the anomaly index does not exceed the upper limit threshold, a Gaussian kernel function is constructed based on the magnitude of the vector difference. The value of the Gaussian kernel function is defined as a negative exponential decay form with the magnitude of the vector difference as the independent variable. Specifically, the exponential decay factor is equal to the square of the negative magnitude of the vector difference divided by the square of twice the bandwidth parameter. The value of the Gaussian kernel function is used as the weighting coefficient of the measurement point data, so that the weighting coefficient continuously and smoothly decays as the magnitude of the vector difference increases. By using projection analysis to eliminate interference that violates basic physical laws, while retaining deviations that conform to physical constraints and continuously weighting them, this method is suitable for complex electromagnetic environments with non-Gaussian interference or partial sensor failure.

[0095] In some embodiments, the above-mentioned calculation of multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron at each sampling time based on the compensated magnetic field data and corresponding weight coefficients at each sampling time, as well as the relative spatial positions of each spatial measuring point at that sampling time in the tetrahedron, may specifically include: establishing a system of linear equations between each component of the magnetic gradient tensor and the magnetic field difference between the measuring point and the measuring point based on the compensated magnetic field data and spatial coordinates of each spatial measuring point; using the weight coefficient of each measuring point as the weight of the corresponding equation, solving the system of linear equations using the weighted least squares method to minimize the weighted sum of squared residuals, thereby obtaining multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron.

[0096] For each sampling time, compensated triaxial magnetic field data from four spatial measurement points are acquired, along with the corresponding weighting coefficient for each point. The spatial coordinates of each measurement point within the tetrahedral array are obtained, with the array's geometric center as the origin. Based on the definition of the magnetic field gradient tensor, a system of linear equations is established between the components of the magnetic gradient tensor (a total of six independent components) and the magnetic field values ​​at each measurement point. Each measurement point contributes three linear equations, corresponding to the three magnetic field components, with the unknowns being the six tensor components at the array center. The three equations contributed by each measurement point are multiplied by the weighting coefficient for that point, forming a weighted system of linear equations. The weighted least squares method is used to solve this system of equations, i.e., to find a set of tensor component values ​​that minimizes the sum of squares of the weighted residuals of each equation. The obtained values ​​are used as multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron.

[0097] In some embodiments, the calculation of multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron at each sampling time, based on the compensated magnetic field data and corresponding weight coefficients at each sampling time, and the relative spatial positions of each spatial measuring point at that sampling time within the tetrahedron, may further include: constructing an initial linear equation system based on the compensated magnetic field data, weight coefficients, and spatial coordinates of each spatial measuring point; introducing physical constraints of zero magnetic field divergence and zero curl as additional conditions; setting initial iteration values; calculating the theoretical magnetic field value of each measuring point based on the currently solved tensor components in each iteration, and recalculating the residual of each measuring point; updating the weight coefficients of each measuring point using the ratio of the residual of the current iteration to the residual of the previous iteration as an adjustment factor; repeating the iteration until the change in the tensor components is less than a preset convergence threshold, and using the tensor components at convergence as multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron.

[0098] An initial system of linear equations can be established, with the six independent components of the magnetic gradient tensor as unknowns. Simultaneously, physical constraints of zero magnetic field divergence and zero curl are introduced, and these constraints are added to the system of equations as supplementary equations, forming an overdetermined system. Initial values ​​are set for iteration, for example, by substituting initial weight coefficients into weighted least squares to obtain an initial solution. The iteration loop then begins: in each iteration, based on the tensor components obtained from the current solution, the theoretical magnetic field value at each measurement point is calculated using forward modeling; the difference between the measured magnetic field value and the theoretical value at that measurement point is taken as the residual for that point; the ratio of the residual magnitude of this iteration to the residual magnitude of the previous iteration is used as an adjustment factor; the weight coefficients of each measurement point are updated using this adjustment factor. One specific update method is to multiply the weight coefficient of the previous iteration by the reciprocal of the adjustment factor or to use a function inversely proportional to the residual magnitude, so that the weight of measurement points with increasing residuals decreases, and the weight of measurement points with decreasing residuals increases. Weighted least squares are then re-executed using the updated weight coefficients to obtain new tensor components. The iterations are repeated until the change between the tensor components obtained from two adjacent iterations is less than a preset convergence threshold, for example, the relative change of each component is less than one-thousandth. The tensor components at convergence are then used as multiple tensor components of the magnetic gradient tensor at the geometric center of a regular tetrahedron. By iteratively reweighting, the weights of measurement points that deviate significantly from the physical model are adaptively reduced, significantly improving the tensor solution accuracy in the presence of multiple interference sources or system biases.

[0099] In some embodiments, step S204 may further include: performing magnetic target localization inversion based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio to obtain the position, burial depth, and uncertainty of the magnetic target; if the uncertainty does not meet the preset certainty condition, calculating the expected information gain of each candidate retest route of the UAV based on the uncertainty; selecting the target retest route from multiple candidate retest routes based on the expected information gain, and determining the altitude, direction, and survey line density of the target retest route; generating a new position, new burial depth, and new uncertainty of the magnetic target based on the retest data during the flight of the UAV along the target retest route; iteratively executing the steps of calculating the expected information gain, selecting the target retest route, and generating the new position, new burial depth, and new uncertainty until the new uncertainty meets the preset certainty condition.

[0100] Inversion is performed based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio to obtain the position, burial depth, and uncertainties of these parameters of the magnetic target. Uncertainty characterizes the confidence interval or variance of the inverted parameter estimates and can be determined by the diagonal elements of the parameter covariance matrix during the inversion process. A certainty condition can be preset, such as a position uncertainty of less than 0.5 meters and a burial depth uncertainty of less than 0.3 meters. If the current uncertainty does not meet this condition, the expected information gain that the UAV can obtain by performing retests on each candidate retest route is calculated based on the uncertainty. Candidate retest routes include different flight altitudes (e.g., low altitude 5 meters, medium altitude 10 meters, high altitude 15 meters), different crossing directions (e.g., perpendicular to the initial survey line, parallel to the initial survey line, and at a 45-degree angle to the initial survey line), and different survey line densities (e.g., sparse, standard, dense). Based on the expected information gain, a target retest route is selected from multiple candidate retest routes, and the flight altitude, crossing direction, and survey line density of this route are determined. The UAV is controlled to fly again along the target retest route to acquire retest data. Based on the retest data, the data compensation, tensor calculation, and inversion steps are repeated to generate a new position, new burial depth, and new uncertainty for the magnetic target. The steps of calculating the expected information gain, selecting the target retest route, retesting, and inversion updating are iteratively executed until the new uncertainty meets the preset certainty condition. The parameter estimation error is adaptively reduced through closed-loop iteration to ensure that the final output target position and burial depth meet the preset accuracy requirements.

[0101] In some embodiments, the above-mentioned calculation of the expected information gain of each candidate retest route of the UAV based on uncertainty may specifically include: obtaining the covariance matrix of the target position parameters and burial depth parameters obtained from the initial inversion, and calculating the determinant of the covariance matrix; for each candidate retest route, predicting the correction effect of the retest data on the covariance matrix based on the flight altitude, crossing direction and survey line density corresponding to the route, and calculating the determinant of the corrected predicted covariance matrix; taking the logarithm of the ratio of the determinant of the initial covariance matrix to the determinant of the predicted covariance matrix as the expected information gain of the candidate retest route.

[0102] The covariance matrices of the target position parameters (eastward and northward coordinates) and burial depth parameters can be extracted from the initial inversion results. These covariance matrices are 3x3 symmetric positive definite matrices, with diagonal elements representing the eastward, northward, and burial depth variances, and off-diagonal elements representing the covariances between the parameters. The determinant of this covariance matrix is ​​calculated; a smaller determinant value indicates lower overall parameter uncertainty. For each candidate remeasurement route, a correction model for the parameter covariance matrix based on the flight altitude, crossing direction, and survey line density is established. This correction model can be based on Bayesian update theory: the amount of observational information in the remeasurement data is determined by the geometric layout of the remeasurement route; for example, lower flight altitude, denser survey lines, or a better crossing direction provide more information, thus significantly reducing the covariance matrix. The covariance matrix of the parameters after remeasurement is predicted, and the determinant of the predicted covariance matrix is ​​calculated. Dividing the determinant of the initial covariance matrix by the determinant of the predicted covariance matrix, and then taking the natural logarithm, yields the expected information gain for the candidate retested route. A larger expected information gain indicates that the route can more effectively reduce the overall parameter uncertainty. By using the logarithmic form of the ratio of the covariance matrix determinants, the comprehensive reduction of multidimensional parameter uncertainty is transformed into a single scalar index, providing a unified mathematical basis for route optimization.

[0103] In some embodiments, the above-mentioned calculation of the expected information gain of each candidate retest route of the UAV based on uncertainty may further include: obtaining the covariance matrix of the target position parameters and burial depth parameters obtained from the first inversion, and extracting the variance terms on the diagonal of the covariance matrix; for each candidate retest route, predicting the reduction ratio of the position variance and burial depth variance of the route respectively; and weighting and summing the reduction ratio of the position variance and the reduction ratio of the burial depth variance according to the preset weights of the target positioning task for position accuracy and burial depth accuracy to obtain the comprehensive expected information gain of the candidate retest route.

[0104] The covariance matrix of the target position parameters and burial depth parameters can be extracted from the initial inversion results. Three variance terms on the diagonal of this covariance matrix are then extracted: eastward position variance, northward position variance, and burial depth variance. For each candidate remeasurement route, the reduction ratio of these three variance terms is predicted. The reduction ratio is calculated by analyzing the relationship between the observation geometry of the remeasurement route and the principal axis direction of the uncertainty ellipse of the current parameters: the reduction effect is best when the survey line direction is orthogonal to the major axis direction of the uncertainty ellipse; the higher the survey line density and the lower the flight altitude, the greater the reduction ratio. Preset weights are set for the target positioning task regarding position accuracy and burial depth accuracy, for example, a position weight of 0.7 and a burial depth weight of 0.3. The average of the eastward and northward position variance reduction ratios is taken as the overall position reduction ratio. This overall position reduction ratio is then multiplied by the position weight, and the result is added to the burial depth variance reduction ratio multiplied by the burial depth weight, yielding a weighted sum, which is used as the overall expected information gain for the candidate remeasurement route. When different exploration missions have different requirements for location and burial depth accuracy, a single information gain index may deviate from the actual needs. By introducing an adjustable weighting factor, the route selection can be flexibly matched with mission priorities.

[0105] In some embodiments, the above-mentioned selection of a target retest route from multiple candidate retest routes based on the expected information gain, and determination of the altitude, direction and survey line density of the target retest route, may specifically include: selecting the candidate retest route with the largest expected information gain as the target retest route, and using the preset altitude, preset direction and preset survey line density corresponding to the target retest route as the altitude, direction and survey line density of the actual retest route.

[0106] The expected information gain values ​​of all candidate routes are compared, and the candidate retest route with the highest expected information gain is selected as the target retest route. Each candidate retest route has its corresponding flight altitude, crossing direction, and survey line density preset during generation. The preset altitude, preset direction, and preset survey line density corresponding to the target retest route are directly used as the route parameters for the actual retest.

[0107] In some embodiments, the above-mentioned selection of a target retest route from multiple candidate retest routes based on the expected information gain, and determination of the altitude, direction, and survey line density of the target retest route, may further include: acquiring the minimum safe flight altitude, maximum endurance, and obstacle distribution information of the UAV in the target area; constructing a route optimization model with the expectation of maximizing the expected information gain as the optimization objective and with the constraint that the flight altitude of the retest route is not lower than the minimum safe flight altitude and avoids the airspace where obstacles are located; solving the optimization model to obtain a Pareto front solution set; selecting the solution with the highest expected information gain from the Pareto front solution set, and using the flight altitude, crossing direction, and survey line spacing corresponding to this solution as the altitude, direction, and survey line density of the target retest route.

[0108] The minimum safe flight altitude of the UAV platform (determined by the fuselage structure and rotor downwash, e.g., 1.5 meters), the maximum endurance of the UAV (determined by battery capacity and payload, e.g., 5 kilometers), and obstacle distribution information in the target area can be obtained in advance (constructed in real-time through prior mapping or airborne LiDAR, including the location and height of trees, buildings, high-voltage power line towers, etc.). The optimization objective is to maximize the expected information gain, while introducing constraints: the flight altitude of the retest route must not be lower than the minimum safe flight altitude; the flight path of the retest route must not enter the airspace where obstacles are located (i.e., maintaining a preset safety margin from obstacles, such as two meters horizontally and one meter vertically); the total length of the retest route must not exceed the remaining endurance of the UAV. A route optimization model is constructed, with decision variables including the route's altitude, direction, survey line density, and the start and end points of the survey lines. A multi-objective optimization algorithm (e.g., non-dominated sorting genetic algorithm or particle swarm optimization) is used to solve the model, obtaining a Pareto front solution set that satisfies all constraints. Each solution in this set represents a candidate route, and the information gain cannot be further increased without violating the constraints. The solution with the highest expected information gain is selected from the Pareto front solution set. The corresponding flight altitude, crossing direction, and survey line spacing of this solution are used as the altitude, direction, and survey line density of the target retest route. By embedding safety and energy constraints into the optimization model, the selected route is ensured to be both efficient and safe, making it suitable for retesting tasks in complex terrain and areas with dense obstacles.

[0109] In some embodiments, magnetic targets include, but are not limited to, ferromagnetic targets. Magnetic targets are used to characterize objects that can produce detectable magnetic anomaly responses under the influence of the Earth's magnetic field or an applied magnetic field. Ferromagnetic targets are used to characterize metallic objects with high magnetic permeability, such as buried steel components, iron pipes, unexploded ordnance, cast iron manhole covers, etc. In addition to ferromagnetic targets, magnetic targets may also include subferromagnetic targets (such as geological bodies containing magnetite) and good conductors with induced magnetic fields (such as the eddy current effect of large aluminum or copper cavities under alternating magnetic fields).

[0110] In some embodiments, magnetic field data includes, but is not limited to, triaxial magnetic field data. Magnetic field data is used to characterize the magnetic flux density information output by the quantum magnetic sensor at the measurement point. Triaxial magnetic field data is used to characterize the projected components of the magnetic field vector in three orthogonal directions (such as north, east, and vertical in a geographic coordinate system). Besides triaxial magnetic field data, magnetic field data can also employ scalar total field data (such as the total magnetic field strength value output by a proton magnetometer), or biaxial magnetic field data (such as ignoring the vertical component in specific application scenarios), or derived characteristic quantities such as magnetic field gradient magnitude, magnetic inclination, and magnetic declination extracted from the triaxial components.

[0111] In some alternative embodiments, the data processing flow adopts a sequence of first performing track correction and then magnetic field compensation. Specifically, after simultaneously collecting magnetic field data, positioning data, altimetry data, and attitude data during UAV flight, the flight track is first spatially corrected using the positioning and attitude data to eliminate the influence of track deviation on the coordinates of the measuring points. After track correction, the magnetic field data is then normalized for altitude and registered horizontally based on the corrected sensor spatial coordinates. Subsequently, tensor calculation, two-dimensional mapping, anomaly extraction, and inversion positioning are performed, ultimately outputting the target position and burial depth. This process, which prioritizes track correction, is suitable for detection scenarios with high requirements for spatial reference consistency and large flight track deviations.

[0112] In some alternative embodiments, anomaly identification employs a template-matching-based intelligent identification algorithm. A pre-constructed magnetic gradient tensor feature template library containing various typical ferromagnetic targets (such as spheres, cylinders, and finite-length rods) is built, with each template corresponding to theoretical tensor anomaly patterns under different burial depths, magnetic moment directions, and signal-to-noise ratios. After obtaining a two-dimensional magnetic anomaly distribution map of the target area, the anomaly map is cross-correlated with each template in the template library using a sliding window, and the similarity coefficient at each grid position is calculated. Regions with similarity exceeding a preset threshold are selected as suspected anomaly candidate regions, and the geometric shape of the target is preliminarily determined based on the template type corresponding to the highest similarity. This method utilizes prior template information to improve the target detection rate in cases of weak anomalies and complex backgrounds.

[0113] In some alternative embodiments, the inversion model employs a combination of a multi-source stacking model and a grid search approach. For anomaly candidate regions, it is assumed that multiple independent magnetic sources exist underground (e.g., two dipoles or one dipole plus a cylinder), with the location, depth, and magnetic moment vector of each source being parameters to be estimated. A grid search method is used to discretize and sample within a preset parameter space: the target region is divided into a fine grid, with each grid point representing a candidate location; at each candidate location, the theoretical tensor characteristics are calculated using the forward formula for dipoles or cylinders, and the sum of squared residuals is calculated by comparing these with the measured characteristics. After traversing all grid points, the grid point with the smallest residual is selected as the initial location of the target, and a local refinement search is performed centered on this point, ultimately outputting the location, depth, and magnetic moment parameters of each magnetic source. This approach is suitable for complex scenarios where multiple targets are close to each other or where the background field is non-uniform.

[0114] In some alternative embodiments, the output includes not only the target's location and depth but also recommended re-survey route parameters based on maximizing information gain. After the initial inversion, the system calculates the covariance matrix of the target's location and depth parameters and extracts the uncertainty ellipse in each direction. Based on the current uncertainty distribution, a set of candidate re-survey routes is automatically generated. Each route has a specific flight altitude (low, medium, and high), crossing direction (along the major axis, minor axis, or at an acute angle to the major axis of the uncertainty ellipse), and survey line density (sparse, standard, and dense). The expected information gain of each candidate route is calculated using the D-optimal criterion, i.e., comparing the logarithmic ratio of the determinant of the predicted covariance matrix after re-survey to the current determinant of the covariance matrix. The route parameters (altitude, direction, and density) with the maximum information gain are output as the recommended re-survey scheme to the ground station interface for the operator to directly execute the second precision exploration flight. This output method realizes an integrated intelligent decision-making closed loop from detection to re-survey.

[0115] As can be seen from the technical solutions provided in the embodiments of this specification above, the embodiments of this specification can effectively reduce the dynamic interference of the UAV's own motors, batteries, and other components on magnetic field measurements and improve data accuracy by limiting the quantum magnetic detection module and the magnetic source of the UAV to meet the preset distance constraints. Using multiple quantum magnetic sensors distributed in a tetrahedral pattern can avoid the background noise of classical magnetic sensors, thereby simultaneously acquiring multi-point spatial magnetic field information, laying the foundation for magnetic gradient tensor calculation. Furthermore, by using positioning, altitude, and attitude data to determine the three-dimensional measurement point trajectories of each quantum magnetic sensor, and compensating for the magnetic field data accordingly, measurement distortion caused by attitude changes and altitude fluctuations during UAV flight can be eliminated. Finally, based on the magnetic gradient tensor, tensor norm, and vertical coupling ratio, inversion can be performed, comprehensively utilizing the spatial variation characteristics of the magnetic field to stably and accurately determine the position and burial depth of the magnetic target.

[0116] It should be understood that in the various embodiments of this specification, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this specification.

[0117] It should also be understood that, in the embodiments of this specification, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this specification generally indicates that the preceding and following related objects have an "or" relationship.

[0118] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0119] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0120] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0121] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational tasks to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The task is a function specified in one or more boxes.

[0122] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A low-altitude UAV magnetic detection method for magnetic target localization, characterized in that, The method includes: a ground-based terminal for a magnetic detection and positioning system; the magnetic detection and positioning system also includes an air-based terminal; the air-based terminal includes a drone and a quantum magnetic detection module installed on the drone, the quantum magnetic detection module and the magnetic source of the drone body satisfy a preset distance constraint, and the quantum magnetic detection module includes multiple quantum magnetic sensors distributed in a tetrahedral pattern; the method includes: Based on the UAV's positioning data, altitude data, and attitude data at each sampling moment during flight, the three-dimensional measurement point trajectory of each quantum magnetic sensor is determined; Based on the three-dimensional measurement point trajectory, the magnetic field data of multiple spatial measurement points synchronously collected by each quantum magnetic sensor at each sampling time are compensated. Based on the compensated magnetic field data, the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio are calculated; the magnetic gradient tensor norm is used to characterize the energy intensity of the magnetic gradient; the magnetic gradient vertical coupling ratio is used to characterize the coupling relationship between the horizontal and vertical magnetic gradients. Based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio, the location and burial depth of the magnetic target are obtained through magnetic target localization inversion.

2. The method according to claim 1, characterized in that, The process of determining the three-dimensional measurement point trajectory of each quantum magnetic sensor based on the UAV's positioning data, altitude data, and attitude data at each sampling moment during flight includes: Based on the positioning data, determine the geographic coordinates of the UAV at each sampling time; Based on the attitude data, the mounting arm length of each quantum magnetic sensor is transformed from the body coordinate system to the geographic coordinate system to obtain the offset of each quantum magnetic sensor relative to the UAV in the geographic coordinate system; the mounting arm length is used to characterize the fixed spatial offset of each quantum magnetic sensor relative to the UAV. By fusing the geographic coordinates of the UAV at each sampling time with the offset of each quantum magnetic sensor, the geographic coordinates of each quantum magnetic sensor at each sampling time are obtained. The geographic coordinates of each quantum magnetic sensor at multiple sampling times are fused in chronological order to obtain the three-dimensional measurement point trajectory of each quantum magnetic sensor.

3. The method according to claim 1, characterized in that, The method of compensating for the magnetic field data of multiple spatial measurement points synchronously collected by each quantum magnetic sensor at each sampling time based on the three-dimensional measurement point trajectory includes: Based on the three-dimensional measurement point trajectories of each quantum magnetic sensor, the actual spatial coordinates of each quantum magnetic sensor in the geographic coordinate system at each sampling time are extracted; the actual spatial coordinates include the horizontal position and the height above the ground. Based on the comparison between the actual ground clearance of each quantum magnetic sensor and the preset reference height, the magnetic field data is subjected to height normalization compensation. Based on the comparison between the actual horizontal position of each quantum magnetic sensor and the horizontal position of the UAV, as well as the attitude data of the UAV, horizontal position registration compensation is performed on the magnetic field data.

4. The method according to claim 1, characterized in that, The step of calculating the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio based on the compensated magnetic field data includes: Based on the compensated magnetic field data at each sampling time and the relative spatial positions of each spatial measuring point at that sampling time in the regular tetrahedron, multiple tensor components of the magnetic gradient tensor at the geometric center of the regular tetrahedron at that sampling time are calculated; each tensor component is used to characterize the rate of change of a magnetic field component along a specific spatial direction. Based on each tensor component of the magnetic gradient tensor at each sampling moment, calculate the magnetic gradient tensor norm and the magnetic gradient vertical coupling ratio corresponding to that sampling moment.

5. The method according to claim 4, characterized in that, The method further includes: For each sampling time, based on the spatial geometric positions of N spatial measurement points in a regular tetrahedron, a magnetic field prediction model is constructed to predict the magnetic field data at the Nth spatial measurement point from the magnetic field data of any N-1 spatial measurement points; where N is an integer greater than or equal to 4; the magnetic field prediction model is used to characterize the spatial distribution relationship of the magnetic field under the constraints of the physical laws of the magnetic field passive region. Each spatial measuring point is taken as the predicted point. Based on the compensated magnetic field data of the remaining N-1 measuring points, the theoretical magnetic field data at the predicted point is calculated by the magnetic field prediction model. The weighting coefficient of the compensated magnetic field data at the predicted point is determined based on the vector difference between the compensated magnetic field data and the theoretical magnetic field data; the weighting coefficient is negatively correlated with the vector difference. The calculation of multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron at each sampling moment, based on the compensated magnetic field data at each sampling moment and the relative spatial positions of each spatial measuring point in the tetrahedron at that sampling moment, includes: Based on the compensated magnetic field data and corresponding weighting coefficients at each sampling time, as well as the relative spatial positions of each spatial measurement point in the tetrahedron at that sampling time, calculate multiple tensor components of the magnetic gradient tensor at the geometric center of the tetrahedron at that sampling time.

6. The method according to claim 1, characterized in that, The method of inverting the magnetic target location based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio to obtain the position and burial depth of the magnetic target includes: Based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio, a magnetic anomaly distribution map of the target region is generated using a magnetic source physical prior model; the magnetic source physical prior model is used to characterize the attenuation relationship of the theoretical tensor features generated by the equivalent magnetic source with spatial distance. Based on the magnetic anomaly distribution map, one or more candidate anomaly regions in the target area are identified. For each candidate region of anomalies, magnetic target localization inversion is performed to obtain the location and burial depth of the magnetic target in that candidate region of anomalies.

7. The method according to claim 6, characterized in that, The method for generating a magnetic anomaly distribution map of the target region using a priori physical model of the magnetic source, based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio, includes: The magnetic gradient tensor's multiple tensor components, the magnetic gradient tensor norm, and the magnetic gradient vertical coupling ratio are used as tensor features of the magnetic anomaly distribution map to be generated. An objective function is constructed for the two-dimensional mesh numerical model to be solved in the target region. The objective function includes: a fitting error term between the tensor eigenvectors to be solved at each two-dimensional mesh point and the measured tensor eigenvectors at each sampling time; a spatial smoothing constraint term between two-dimensional mesh points; and a priori constraint term based on the magnetic source physical prior model. The measured tensor eigenvector at each sampling time is used to characterize the tensor eigenvector at the geometric center of the tetrahedron at that sampling time. The spatial smoothing constraint term is used to characterize the smoothness of the change of tensor eigenvectors between adjacent two-dimensional mesh points. The priori constraint term is used to characterize the degree of deviation between the tensor eigenvectors to be solved in each two-dimensional mesh and the theoretical tensor eigenvectors given by the magnetic source physical prior model. Find a two-dimensional grid numerical model that minimizes the objective function; Based on the obtained two-dimensional grid numerical model, a magnetic anomaly distribution map of the target area is generated.

8. The method according to claim 6, characterized in that, The step of performing magnetic target localization inversion for each anomaly candidate region to obtain the position and burial depth of the magnetic target in that anomaly candidate region includes: For each anomaly candidate region, multiple magnetic positioning inversion models are used for inversion. These multiple magnetic positioning inversion models include at least a magnetic dipole model, a cylinder model, and a finite-length magnetic anomaly model. The magnetic dipole model is used to characterize the spatial distribution of the magnetic gradient tensor generated by approximately spherical or equiaxed magnetic targets. The cylinder model is used to characterize the spatial distribution of the magnetic gradient tensor generated by long tubular or columnar magnetic targets. The finite-length magnetic anomaly model is used to characterize the spatial distribution of the magnetic gradient tensor generated by rod-shaped or strip-shaped magnetic targets with finite extension length. Based on the inversion accuracy and model complexity of each magnetic positioning inversion model, a target magnetic positioning inversion model is selected from the plurality of magnetic positioning inversion models. Based on the inversion results of the target magnetic positioning inversion model, the location and burial depth of the magnetic target in the anomaly candidate region are determined.

9. The method according to claim 1, characterized in that, The method of performing magnetic target localization inversion based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio to obtain the position and burial depth of the magnetic target also includes: Based on the magnetic gradient tensor, magnetic gradient tensor norm, and magnetic gradient vertical coupling ratio, the location inversion of magnetic targets is performed to obtain the position, burial depth, and uncertainty of the magnetic targets. If the uncertainty does not meet the preset certainty condition, calculate the expected information gain of each candidate retest route of the UAV based on the uncertainty. Based on the expected information gain, a target retest route is selected from multiple candidate retest routes, and the altitude, direction, and survey line density of the target retest route are determined. Based on the retest data during the UAV's flight along the target retest route, new positions, new burial depths, and new uncertainties of the magnetic target are generated; The process of iteratively executing the steps of calculating the expected information gain, selecting the target retest route, and generating new location, new burial depth, and new uncertainty continues until the new uncertainty meets the preset certainty condition.

10. A magnetic detection and positioning system, characterized in that, The magnetic detection and positioning system includes an airborne end and a ground end; the airborne end includes a drone and a quantum magnetic detection module installed on the drone; the quantum magnetic detection module and the magnetic source of the drone body meet a preset distance constraint, and the quantum magnetic detection module includes multiple quantum magnetic sensors distributed in a regular tetrahedral pattern; the ground end is used to perform the method as described in any one of claims 1-9.

Images