An intelligent remote sensing identification system for mineral resource exploration

Abstract

The application discloses a kind of intelligent remote sensing identification systems for mineral resources exploration, belong to remote sensing detection technical field.System includes acquisition module, emits linear frequency modulation pulse sequence to exploration area and collects echo, obtains full polarized complex scattering matrix data set;Tomographic focusing module, tomographic focusing processing is carried out to scattering matrix, according to the difference of penetration depth under different polarization base separates surface scattering component and deep body scattering component, obtains the depth positioning parameter set of mineralization anomaly body;Parameter extraction module extracts cross-polarization phase difference parameter, body scattering depolarization rate parameter and complex dielectric constant real part parameter, generates ore-bearing indication characteristic vector;Imaging and identification module, according to variable scale detection step, point-by-point focusing imaging is generated to generate multi-polarization synthetic aperture radar image, and the spatial position calibration of ore-bearing indication characteristic vector is carried out, obtains target mineralization body spatial distribution identification result.

Description

Technical Field

[0001] This invention relates to the field of remote sensing technology, specifically to an intelligent remote sensing identification system for mineral resource exploration. Background Technology

[0002] In mineral resource exploration, when using remote sensing to identify mineralization anomalies, traditional multi-polarization synthetic aperture radar (MPAR) systems primarily acquire surface or near-surface geological information by analyzing the intensity and polarization characteristics of the scattered energy from the ground surface. This method treats all received radar echoes as contributions from the same scattering layer, failing to effectively separate the volume scattering signal penetrating underground from the surface scattering signal. This results in extracted parameters related to mineral lattice structure and dielectric properties being mixed with surface interference, making it difficult to reflect the true state of underground mineralization targets. Furthermore, in the imaging processing stage, the detection step size used for azimuth focusing is usually set to a fixed value, failing to consider the differences in scattering response scales of mineralization anomalies of different sizes and burial depths. A fixed step size can lead to oversampling omissions of small-scale anomalies or undersampling and aliasing of large-scale anomalies, resulting in blurred mineralization boundaries and insufficient spatial positioning accuracy in the final image. Existing technologies for interpreting deep mineralization information lack a mechanism to separate surface and deep scattering contributions based on differences in radar wave penetration depth under different polarization bases. They also lack techniques for dynamically adapting the extracted electromagnetic scattering feature vectors to the detection step size. Therefore, how to separate the deep volume scattering component from the fully polarized complex scattering matrix and obtain reliable depth positioning parameters, and how to adaptively adjust the detection step size based on the mineralization indicator feature vector to achieve precise focused imaging of mineralized targets, have become pressing problems to be solved. Summary of the Invention

[0003] This paper presents an intelligent remote sensing identification system for mineral resource exploration. The system separates the deep volume scattering components and determines the depth positioning parameters through tomographic focusing processing. Based on the ore-bearing indicator feature vector, it obtains the variable-scale detection step size by searching the neighborhood in the strategy library. Combined with point-by-point focusing imaging, it realizes the spatial distribution identification of mineralization anomalies.

[0004] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides an intelligent remote sensing identification system for mineral resource exploration, the system comprising an acquisition module, a tomographic focusing module, a parameter extraction module, a step size determination module, and an imaging and identification module. The acquisition module transmits a linear frequency modulated pulse sequence to the exploration area and acquires echo signals reflected from ground features, obtaining a set of fully polarized complex scattering matrix data. The tomographic focusing module performs tomographic focusing processing on the set of fully polarized complex scattering matrix data, separating the surface scattering component and the deep volume scattering component based on the differences in radar wave penetration depth under different polarization bases, obtaining a set of depth positioning parameters for mineralized anomalies. The parameter extraction module extracts cross-polarization phase difference parameters, volume scattering depolarization parameters, and real part parameters of the complex permittivity related to the mineral lattice structure from the set of depth positioning parameters, generating a mineral-bearing indicator feature vector. The step size determination module uses the mineral-bearing indicator feature vector as an index to perform a neighborhood search in a pre-constructed remote sensing detection strategy library for mining areas, determining a variable-scale detection step size that matches the current size of the mineralized anomaly. The imaging and recognition module performs point-by-point focusing imaging on the fully polarized complex scattering matrix data set according to the variable-scale detection step size, generating a multi-polarization synthetic aperture radar (MPAR) image. The MPAR image is then used to spatially locate the ore-bearing indicator feature vector, obtaining the spatial distribution recognition result of the target mineralization body. This system utilizes tomographic focusing to separate scattering contributions at different depths and adaptively adjusts the imaging process based on the variable-scale detection step size, enabling high-resolution focused images and accurate three-dimensional spatial coordinates to be obtained for mineralization anomalies of varying depths and scales.

[0005] As a technical solution of the present invention, the acquisition module transmits the linear frequency modulated pulse sequence in groups according to different polarization bases, including horizontal-horizontal polarization bases, horizontal-vertical polarization bases, vertical-horizontal polarization bases, and vertical-vertical polarization bases; it synchronously records the transmission timestamp of each pulse under each polarization base, and performs range compression on the echo signal according to the transmission timestamps to obtain the fully polarized complex scattering matrix data set. Through fully polarized multi-base transmission and synchronous time-series recording, the polarization scattering information of the ground target and the phase information of each scattering center in the depth direction are completely preserved, providing high-fidelity basic data for subsequent tomographic focusing and parameter inversion.

[0006] As a technical solution of the present invention, the tomographic focusing module arranges the data of each pixel in the fully polarized complex scattering matrix dataset in time-delay order to construct a complex scattering intensity profile along the radar line of sight. The complex scattering intensity profile is then subjected to de-clipping processing on a depth-by-depth-unit basis to separate the surface scattering component corresponding to the surface roughness of the ground object and the deep-body scattering component corresponding to the structure of the underground target body. The depth unit position corresponding to the energy peak value in the deep-body scattering component is extracted as the depth positioning parameter of the mineralization anomaly. Utilizing the difference in penetration depth under different polarization bases for de-clipping separation can effectively suppress surface scattering interference caused by surface vegetation and roughness, highlighting the deep-body scattering signal that reflects the mineral lattice structure and joint development, thus achieving a reliable estimation of the burial depth of the mineralization anomaly.

[0007] As a technical solution of the present invention, the parameter extraction module performs phase calculation on the co-polarized components and cross-polarized components under the linear polarization basis in the fully polarized complex scattering matrix data set, calculates the phase difference between the cross-polarized components and the co-polarized components, and uses it as the cross-polarization phase difference parameter; it performs polarization decomposition on the fully polarized complex scattering matrix data set to obtain the power proportions of odd-order scattering, even-order scattering, and volume scattering, and uses the ratio of the power proportion of volume scattering to the total scattering power as the volume scattering depolarization rate parameter; it extracts the backscattering coefficient amplitude value of the deep volume scattering component at the system center frequency, and inverts this amplitude value according to the pre-calibrated incident angle-dielectric constant mapping relationship, using the inversion result as the real part parameter of the complex dielectric constant. Thus, the generated ore-bearing indicator feature vector integrates the cross-polarization phase difference closely related to the mineral lattice structure, the volume scattering proportion reflecting the depolarization effect, and the real part of the complex dielectric constant characterizing the mineral's dielectric properties, making the ore-bearing feature description have strong physical interpretability and anti-interference ability.

[0008] As a technical solution of the present invention, the step size determination module acquires multiple historical ore-bearing indicator feature vectors and corresponding multiple historical variable-scale detection step sizes as retrieval samples; maps the multiple historical ore-bearing indicator feature vectors to a multi-dimensional space with electromagnetic scattering parameters as the axis, and labels the points in the multi-dimensional space with the corresponding multiple historical variable-scale detection step sizes to construct the remote sensing detection strategy library for the mining area; in the multi-dimensional space, a dynamic search neighborhood is defined with the spatial point corresponding to the current ore-bearing indicator feature vector as the center, and the median of all historical variable-scale detection step sizes in the dynamic search neighborhood is calculated as the variable-scale detection step size. By constructing the detection strategy library through historical detection samples and using median neighborhood search, the step size can adapt to the distribution characteristics of the current ore-bearing indicator feature vector, maintaining azimuth resolution while avoiding focusing mismatch or computational redundancy caused by a fixed step size.

[0009] Furthermore, the step size determination module extracts all historical ore-bearing indicator feature vectors within the dynamic search neighborhood and calculates the Euclidean distance between each historical ore-bearing indicator feature vector and the current ore-bearing indicator feature vector to obtain a distance set. The distance values ​​in the distance set are sorted from smallest to largest, and the distance located at a preset quantile value is selected as the radius of the dynamic search neighborhood. Adaptively determining the neighborhood radius based on the distance quantile ensures that the neighborhood contains a sufficient number of effective historical samples under different feature space densities, improving the robustness of the variable-scale detection step size determination.

[0010] As a technical solution of the present invention, the imaging and recognition module uses the variable-scale detection step size as the number of pulse accumulation points for the azimuth matched filter, and performs point-by-point focusing imaging on the signal at each azimuth position in the fully polarized complex scattering matrix data set to generate a focused single-view complex image; it extracts the total polarization power value of each pixel in the single-view complex image, and selects pixels with a total polarization power value greater than a preset power threshold as candidate mineralized pixels. By using an adaptive step size for pulse accumulation and focusing, pixels with sufficient signal-to-noise ratio can still be formed in deep regions with low echo intensity, improving the detection probability of weak-signal mineralized anomalies.

[0011] Furthermore, spatial connectivity analysis is performed on the candidate mineralized pixels to merge spatially adjacent candidate mineralized pixels into a single mineralization anomaly patch. The geometric center coordinates and projected area of ​​each mineralization anomaly patch are calculated. The geometric center coordinates are then matched with the depth positioning parameters to obtain the three-dimensional spatial coordinates of the mineralization anomaly. By merging connectivity components and depth matching, discrete candidate pixels in the two-dimensional image are integrated into a complete mineralization anomaly with three-dimensional spatial attributes, significantly reducing fragmentation and false alarm targets.

[0012] Furthermore, the cross-polarization phase difference parameter, the volume scattering depolarization parameter, and the real part parameter of the complex permittivity of all pixels within the mineralization anomaly patch are averaged to obtain the average feature vector of the patch, which serves as the spatial location calibration result of the mineralization indicator feature vector. Feature averaging on a patch-by-pattern basis can suppress parameter estimation noise for individual pixels, making the calibration result more representative of the overall properties of the mineralization anomaly.

[0013] Furthermore, the three-dimensional spatial coordinates are combined with the average feature vector to generate a composite data record containing spatial location information and mineralization attributes. This composite data record is then spatially overlaid with a pre-stored geological map of the mining area to output the spatial distribution identification result of the target mineralization body. This method integrates the three-dimensional coordinates and physical parameters of the mineralization anomaly retrieved from remote sensing with the geological map, providing an intuitive and information-rich decision-making basis for subsequent exploration engineering deployment.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: The tomographic focusing module processes the fully polarized complex scattering matrix dataset. Based on the differences in radar wave penetration depth under horizontal-horizontal, horizontal-vertical, vertical-horizontal, and vertical-vertical polarization bases, the complex scattering intensity profile of each pixel along the radar line of sight is deskewed unit by depth. This separates the surface scattering component corresponding to the surface roughness of the ground object and the deep volume scattering component corresponding to the structure of the underground target. The depth unit position corresponding to the energy peak in the deep volume scattering component is extracted as the depth positioning parameter of the mineralization anomaly. This method physically distinguishes the sources of surface scattering and volume scattering, eliminating the interference of surface scattering on the extraction of deep mineralization information. It allows the acquired depth positioning parameters to be directly associated with the actual occurrence depth of the mineralization anomaly, providing a clean deep volume scattering data source for subsequent feature parameter inversion. The parameter extraction module extracts the cross-polarization phase difference parameter, the volume scattering depolarization rate parameter, and the real part parameter of the complex permittivity from the depth positioning parameter set to generate a mineralized indicator feature vector. The step size determination module uses the ore-bearing indicator feature vector as an index to perform a neighborhood search in a pre-built remote sensing detection strategy library for mining areas. This strategy library is established by mapping multiple historical ore-bearing indicator feature vectors to a multi-dimensional space with electromagnetic scattering parameters as the axis and labeling the historical variable-scale detection step sizes. During the search, a dynamic search neighborhood is defined centered on the spatial point of the current ore-bearing indicator feature vector, and the median of all historical variable-scale detection step sizes within the neighborhood is calculated as the currently applicable variable-scale detection step size. This mechanism allows the detection step size to automatically adapt to the spatial extension scale of the mineralization anomaly based on its electromagnetic scattering characteristics, avoiding the azimuth undersampling ambiguity caused by a fixed step size for small-scale mineralization targets and the redundant calculations caused by a fixed step size for large-scale mineralization targets. This maintains the boundary sharpness and internal texture details of various mineralization patches in multi-polarization synthetic aperture radar images. The imaging and recognition module performs point-by-point focusing imaging based on the variable-scale detection step size and completes three-dimensional spatial coordinate matching by combining depth positioning parameters. The final output of the target mineralization spatial distribution recognition result is improved in terms of both geometric integrity and spatial positioning accuracy. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a schematic diagram of the structure of an intelligent remote sensing identification system for mineral resource exploration; Figure 2 This is a flowchart of the fully polarized pulse grouping and range compression processing. Figure 3 This is a flowchart for extracting ore-bearing indicator feature vectors. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] See Figure 1 This invention provides an intelligent remote sensing identification system for mineral resource exploration, comprising an acquisition module, a tomographic focusing module, a parameter extraction module, a step size determination module, and an imaging and identification module. The acquisition module transmits a linear frequency modulated pulse sequence to the exploration area and acquires echo signals reflected from ground features, obtaining a set of fully polarized complex scattering matrix data. The tomographic focusing module performs tomographic focusing processing on this set of fully polarized complex scattering matrix data, separating the surface scattering component and the deep volume scattering component based on the differences in radar wave penetration depth under different polarization bases, obtaining a set of depth positioning parameters for mineralized anomalies. The parameter extraction module extracts cross-polarization phase difference parameters, volume scattering depolarization parameters, and the real part of the complex permittivity parameter related to the mineral lattice structure from the depth positioning parameter set, generating a mineral-bearing indicator feature vector. The step size determination module uses the mineral-bearing indicator feature vector as an index to perform a neighborhood search in a pre-constructed remote sensing detection strategy library for mining areas, determining a variable-scale detection step size that matches the current size of the mineralized anomaly. The imaging and recognition module performs point-by-point focusing imaging on the fully polarized complex scattering matrix data set based on the variable scale detection step size, generates a multi-polarization synthetic aperture radar image, and uses the multi-polarization synthetic aperture radar image to calibrate the spatial location of the ore-bearing indicator feature vector, thereby obtaining the spatial distribution recognition result of the target mineralized body.

[0019] Example 1: In specific implementation, please refer to Figure 2The acquisition module is equipped with a multi-channel waveform generator and a polarization-switching antenna feed. The linear frequency modulated pulse sequence transmitted by the acquisition module to the exploration area is grouped according to polarization bases. The grouping method is as follows: one group of pulses is configured to transmit and receive horizontally polarized waves, forming a horizontal-horizontal polarization base; another group is configured to transmit and receive horizontally polarized waves, forming a horizontal-vertical polarization base; a third group is configured to transmit and receive vertically polarized waves, forming a vertical-horizontal polarization base; and a fourth group is configured to transmit and receive vertically polarized waves, forming a vertical-vertical polarization base. During transmission, at the leading edge of each pulse, the clock synchronization unit in the acquisition module records the transmission timestamp and binds and stores the timestamp with the corresponding polarization base identifier and pulse number.

[0020] After receiving the echo signals reflected from ground objects, the acquisition module performs range compression processing on the echo signals under each polarization base. The specific method of range compression processing is as follows: using the transmission timestamp corresponding to the echo of that polarization base, the zero point of the echo signal is determined; and the raw echo data within a synthetic aperture processing interval is extracted from the continuous received data stream, denoted as... ,in, An identifier representing the emission polarization mode. The value is the horizontal polarization identifier. Or vertical polarization identifier , An identifier representing the receiving polarization mode. The value is or , This is a time delay variable relative to zero time. The acquisition module uses the modulation frequency of the transmitted linear frequency modulated pulse sequence. Construct a matching reference signal, the expression of which is: , here The imaginary unit satisfies , Pi This represents the time coordinate of the distance to the compressed output. Frequency modulation. The value of depends on the bandwidth of the transmitted linear frequency modulated pulse. and pulse width ,satisfy After aligning according to the transmission timestamp, range compression is performed on the original echo data. This is achieved by performing a convolution operation with the matched reference signal. The expression for the convolution operation is: in, Indicates time The distance obtained at that point is compressed into a complex value, and the integral variable is... The value of is within the entire echo sampling interval, and the integral result gives a complex number that contains both amplitude and phase information. For each polarization basis combination, the above range compression operation is performed to obtain... , , , Four complex functions together constitute the fully polarimetric complex scattering matrix data set. Each range-azimuth resolved cell in the fully polarimetric complex scattering matrix data set corresponds to one... The complex matrix, whose elements are the four complex numbers mentioned above, fully records the complex scattering amplitude and relative phase information of the resolution unit.

[0021] Example 2: In practice, after acquiring the fully polarized complex scattering matrix data set, the tomographic focusing module performs a serialization and arrangement operation along the radar line-of-sight for the data of each pixel in the data set. For a pixel with a fixed azimuth and fixed range, the fully polarized complex scattering matrix data set stores the complex values ​​of four polarization channels: horizontal-horizontal polarization basis, horizontal-vertical polarization basis, vertical-horizontal polarization basis, and vertical-vertical polarization basis. The tomographic focusing module arranges the complex values ​​of each polarization channel according to their corresponding time delay order, which is determined by the difference between the transmission timestamp and the echo reception timestamp recorded by the acquisition module. After the arrangement is completed, each pixel forms a one-dimensional complex sequence along the radar line-of-sight under one polarization channel. The one-dimensional complex sequences of the four polarization channels together constitute the complex scattering intensity profile of that pixel, and the position of each element in the complex scattering intensity profile corresponds to a depth cell.

[0022] After obtaining the complex scattering intensity profile for each pixel, the tomographic focusing module performs de-chirping processing on the complex scattering intensity profile at depth units. The de-chirping processing is performed separately for each depth unit in the complex scattering intensity profile. The de-chirping process is implemented as follows: the tomographic focusing module pre-stores a reference signal with the same modulation frequency as the emitted linear frequency modulated pulse sequence, and performs a conjugate multiplication operation between the complex value corresponding to each depth unit in the complex scattering intensity profile and this reference signal. The construction of the reference signal depends on the modulation frequency of the emitted linear frequency modulated pulse sequence. Frequency modulation The value is determined by the pulse bandwidth configured at the time of transmission. and pulse width Decision, satisfaction Let the th complex scattering intensity profile contain... The complex value corresponding to each depth unit is denoted as The first tomographic focusing module generates The reference signal value corresponding to each depth unit is denoted as , The expression depends on the depth cell number. Harmonic frequency The calculation method for deskewing per depth element is expressed as follows: in, The first one after deslant removal The complex value of a depth unit. This is the depth cell number. The range of values ​​is from arrive integers, This represents the total number of depth cells in the complex scattering intensity profile. The first scattering intensity profile represents the complex scattering intensity profile. The original complex values ​​of each depth unit. Representing the Reference signal value corresponding to each depth unit The complex conjugate of . Where the reference signal Depend on Give, The imaginary unit satisfies , Pi Representing the The signal propagation delay time corresponding to each depth unit Depend on The product of the distance and the sampling interval is used to determine the distance.

[0023] After declipping processing at each depth unit, the differences in the distribution of scattering responses corresponding to surface roughness and subsurface target structure in the depth direction of the complex scattering intensity profile are distinguished. The tomographic focusing module segments the declipping complex scattering intensity profile according to a pre-set depth threshold, classifying the portion with a depth unit number less than or equal to the depth threshold as the surface scattering component, and the portion with a depth unit number greater than the depth threshold as the deep volume scattering component. The depth threshold is set based on the nominal penetration depth of the emitted linear frequency modulated pulse sequence in typical surface media of the exploration area. The nominal penetration depth is pre-obtained through measurements of the dielectric constant of surface samples in the exploration area and calculations of the radar wave attenuation coefficient.

[0024] After obtaining the deep volume scattering component, the tomographic focusing module performs energy calculations on the complex values ​​of all depth cells contained in the deep volume scattering component. The energy calculation is achieved by calculating the squared modulus of the complex value corresponding to each depth cell; that is, for the th... Complex values ​​of depth units Energy value The tomographic focusing module searches for the depth cell with the highest energy value among all depth cells covered by the deep volume scattering component, and records the depth index corresponding to the depth cell with the highest energy value as the peak depth cell index. The tomographic focusing module multiplies the peak depth cell index by the actual spatial interval of the depth cells to obtain the depth positioning parameters of the mineralization anomaly. The actual spatial interval of the depth cells is determined by the range sampling rate and the propagation speed of electromagnetic waves in the medium. The above operation is performed on each pixel in the fully polarized complex scattering matrix dataset that identifies the energy peak of the deep volume scattering component. The depth positioning parameters of the mineralization anomaly corresponding to all pixels are summarized to form a depth positioning parameter set.

[0025] Example 3: In specific implementation, please refer to Figure 3 After acquiring the fully polarized complex scattering matrix data set and the depth positioning parameter set, the parameter extraction module performs the calculation of the cross-polarization phase difference parameter. The parameter extraction module extracts the co-polarization component and cross-polarization component under the linear polarization basis from the fully polarized complex scattering matrix data set. The co-polarization component refers to the complex value corresponding to the horizontal-horizontal polarization basis. The complex values ​​corresponding to the vertical-vertical polarization basis Cross-polarization components refer to the complex values ​​corresponding to the horizontal-vertical polarization basis. Complex values ​​corresponding to vertical-horizontal polarization bases The parameter extraction module performs phase calculations on the co-polarization and cross-polarization components. Phase calculation is achieved by extracting the argument of the complex values. For horizontal-horizontal polarization basis complex values... Its phase value Depend on Give, This indicates the operation of taking the argument of a complex number, where the argument can take values ​​ranging from 1 to 2. For horizontally-vertically polarized basis complex values Its phase value Depend on The parameter extraction module calculates the difference between the phase values ​​of the cross-polarization component and the phase values ​​of the same polarization component to obtain the horizontal cross-polarization phase difference. Similarly, the vertical cross-polarization phase difference The parameter extraction module extracts the horizontal cross-polarization phase difference. Phase difference with vertical cross-polarization The average value is then used as the cross-polarization phase difference parameter.

[0026] The parameter extraction module extracts the volume scattering depolarization parameter. It performs polarization decomposition on the fully polarized complex scattering matrix dataset using an eigenvalue-based polarization target decomposition method. The complex scattering matrix is ​​converted into a polarization coherence matrix, which is: The Hermitian matrix is ​​composed of elements that are second-order statistics of complex values ​​under different combinations of polarization bases. The parameter extraction module performs eigenvalue decomposition on the polarization coherence matrix to obtain three eigenvalues. , and The three eigenvalues ​​satisfy The parameter extraction module calculates the probability distribution of the scattering mechanism based on three eigenvalues, and decomposes the total polarization power into odd-order scattering power using the probability distribution of the scattering mechanism. Even-order scattering power Volume scattering power The volume scattering depolarization parameter is the volume scattering power. With total scattering power The ratio of total scattered power Depend on Calculations show that This indicates operations involving complex modulus values. The expression for the volume scattering depolarization parameter is: .

[0027] The parameter extraction module extracts the real part of the complex permittivity. It obtains the deep volume scattering component from the tomography focusing module and extracts the deep volume scattering component at the system center frequency. The amplitude of the backscattering coefficient at that location. (System center frequency) The center frequency of the transmitted linear frequency modulated pulse sequence is determined during the system design phase. The backscattering coefficient amplitude is extracted by obtaining the de-skewed complex values ​​of the deep volume scattering components from the parameter extraction module. Calculate the center frequency of the system. The modulus squared value at the corresponding depth cell is then used for antenna gain correction and range attenuation compensation based on the radar equations. Antenna gain correction uses a pre-calibrated antenna pattern gain value, and range attenuation compensation follows the electromagnetic wave propagation attenuation model in a medium. The parameter extraction module inverts the corrected backscattering coefficient amplitude value based on a pre-calibrated incident angle-dielectric constant mapping relationship. This mapping relationship is established by pre-measuring standard samples with known dielectric constants at multiple incident angles. The mapping relationship is stored in a two-dimensional lookup table, with the incident angle on the horizontal axis and the backscattering coefficient amplitude value on the vertical axis. Each cell in the lookup table stores the corresponding real part of the complex dielectric constant. The parameter extraction module performs interpolation retrieval in the two-dimensional lookup table based on the radar incident angle and the corrected backscattering coefficient amplitude value corresponding to the current pixel, using the retrieval result as the real part parameter of the complex dielectric constant.

[0028] The parameter extraction module combines the cross-polarization phase difference parameter, the bulk scattering depolarization parameter, and the real part of the complex permittivity parameter into a three-dimensional vector, denoted as [parameter name missing]. 3D vector The expression is: in, Represents the ore-bearing indicator feature vector. Represents the cross-polarization phase difference parameter. Due to horizontal cross-polarization phase difference Phase difference with vertical cross-polarization The arithmetic mean is obtained. The range of values ​​is . Representative volume scattering depolarization parameter, It is the ratio of volume scattered power to total scattered power. The range of values ​​is a closed interval. . Represents the real part of the complex permittivity. It is obtained by inverting the amplitude value of the backscattering coefficient. The value of depends on the dielectric properties of the subsurface medium in the exploration area. The above operation is performed on each pixel in the fully polarized complex scattering matrix dataset to generate a mineral-bearing indicator feature vector for each pixel.

[0029] Example 4: In practical implementation, the step size determination module pre-acquires multiple historical mineralization indicator feature vectors and the corresponding historical variable-scale detection step size as retrieval samples. Each historical mineralization indicator feature vector is a three-dimensional vector, with three components: a historical cross-polarization phase difference parameter, a historical volume scattering depolarization parameter, and a historical complex permittivity real part parameter. The historical cross-polarization phase difference parameter is composed of the cross-polarization phase difference value calculated from the fully polarized complex scattering matrix dataset in historical exploration tasks. The historical volume scattering depolarization parameter is composed of the proportion of volume scattering power obtained from polarization decomposition in historical exploration tasks. The historical complex permittivity real part parameter is composed of the real part value of the complex permittivity obtained from the inversion of deep volume scattering components in historical exploration tasks. The historical variable-scale detection step size corresponding to each historical mineralization indicator feature vector is the number of pulse accumulation points of the azimuth matched filter actually used in the historical exploration task. The value of this historical variable-scale detection step size is configured after manually judging the size of the mineralization anomaly in the historical exploration task.

[0030] The step size determination module maps all historical ore-bearing indicator feature vectors to a multi-dimensional space, where the coordinate axes are composed of electromagnetic scattering parameters. The first coordinate axis of the multi-dimensional space is the cross-polarization phase difference axis, the second coordinate axis is the volume scattering depolarization axis, and the third coordinate axis is the real part axis of the complex permittivity. For the... The module uses historical mineralization indicator feature vectors as inputs. It then uses the historical cross-polarization phase difference parameter from these vectors as the coordinate value on the first coordinate axis, the historical volume scattering depolarization parameter as the coordinate value on the second coordinate axis, and the historical real part of the complex permittivity parameter as the coordinate value on the third coordinate axis, thus determining a historical spatial point in multidimensional space. The module also uses the historical variable-scale detection step size corresponding to the historical mineralization indicator feature vector as the label value for this historical spatial point, storing both the coordinates and the label value. The set of all historical spatial points and their label values ​​constitutes the remote sensing detection strategy library for the mining area.

[0031] During the real-time processing phase, the step size determination module obtains the current ore-bearing indicator feature vector from the parameter extraction module. This current ore-bearing indicator feature vector is a three-dimensional vector, with its three components being the current cross-polarization phase difference parameter, the current volume scattering depolarization parameter, and the current real part of the complex permittivity parameter, respectively. In multi-dimensional space, the step size determination module delineates a dynamic search neighborhood centered on the current spatial point corresponding to the current ore-bearing indicator feature vector. The process of delineating this dynamic search neighborhood involves the step size determination module extracting historical spatial points corresponding to all historical ore-bearing indicator feature vectors from the remote sensing detection strategy library for the mining area, calculating the Euclidean distance between each historical spatial point and the current spatial point, and obtaining a distance set composed of multiple Euclidean distance values.

[0032] The Euclidean distance is calculated as follows: Let the current ore-bearing indicator feature vector be... ,in, Represents the current cross-polarization phase difference parameter. Represents the current volume scattering depolarization parameter. Represents the real part of the current complex permittivity; the first The historical ore-bearing indicator feature vectors are ,in, Representing the The historical cross-polarization phase difference parameter of the historical ore-bearing indicator eigenvector. Representing the The historical volume scattering depolarization parameter of the historical mineralization indicator eigenvector. Representing the The real part of the historical complex permittivity parameter of the historical ore-bearing indicator eigenvector. Current spatial point and the... Euclidean distance between historical points Calculate according to the following formula: in, Represents the current spatial point and the first Euclidean distance between historical points in space This represents the sequence number of the historical ore-bearing indicator feature vector. The range of values ​​is from arrive integers, This represents the total number of historical ore-bearing indicator feature vectors in the remote sensing detection strategy library for mining areas. Represents the current cross-polarization phase difference parameter. Representing the The historical cross-polarization phase difference parameter of the historical mineralization indicator feature vector, both are in radians. Represents the current volume scattering depolarization parameter. Representing the The historical volume scattering depolarization parameter of the historical mineralization indicator feature vector is a dimensionless ratio. This represents the real part of the current complex permittivity. Representing the The real part of the historical complex permittivity parameter of the historical ore-bearing indicator eigenvector is a dimensionless real number.

[0033] After obtaining the distance set, the step size determination module sorts all distance values ​​in the set in ascending order to form an ordered distance sequence. The step size determination module then selects the distance located at a preset quantile value within the ordered distance sequence as the radius of the dynamic search neighborhood. The preset quantile value is set to... The setting is based on: when the preset quantile value is taken as... In this approach, the dynamic search neighborhood can encompass three-quarters of the nearby historical spatial points while excluding the remaining quarter, striking a balance between the representativeness of the included samples and the locality of the neighborhood's extent. The step size determination module uses the current spatial point as the center and the distance selected from the ordered distance sequence as the radius to determine a spherical neighborhood in multidimensional space; this spherical neighborhood is the dynamic search neighborhood.

[0034] The step size determination module extracts the labeled values ​​corresponding to all historical spatial points within the dynamic search neighborhood, specifically the historical variable-scale detection step size labeled for each historical spatial point falling within the dynamic search neighborhood, forming a neighborhood step size set. The step size determination module sorts all historical variable-scale detection step sizes in the neighborhood step size set according to their numerical values ​​and takes the median of the sorted sequence as the variable-scale detection step size. When the number of historical variable-scale detection step sizes in the neighborhood step size set is odd, the median is the value in the middle position after sorting; when the number of historical variable-scale detection step sizes in the neighborhood step size set is even, the median is the arithmetic mean of the two middle values ​​after sorting. The step size determination module outputs the calculated median as the variable-scale detection step size that matches the current ore-bearing indicator feature vector.

[0035] Example 5: In practice, the imaging and recognition module obtains the variable-scale detection step size from the step size determination module and uses this variable-scale detection step size as the number of pulse accumulation points for the azimuth matched filter. The imaging and recognition module performs point-by-point focusing imaging on the signal at each azimuth position in the fully polarized complex scattering matrix dataset. For a signal sequence at a fixed distance, the azimuth matched filter operates as follows: the imaging and recognition module uses the number of pulse accumulation points specified by the variable-scale detection step size. The coherent accumulation length, used as the coherent accumulation length, is used to perform a weighted summation of the azimuth sampling points within that length. The weighting coefficients of the azimuth matched filter are determined by the azimuth reference function, which is constructed based on the radar platform's motion parameters and the Doppler frequency history. Let the first... The directional position is at the 1st The complex value of each distance cell is After azimuth matched filter processing, the focused complex value Depend on Coefficients of the azimuth reference function Discrete convolution is performed to obtain, where, These are the coefficient indices of the azimuth reference function. The value range is determined by the number of pulse accumulation points. The decision is made. The imaging and recognition module performs the above point-by-point focusing imaging operation on the complex values ​​of the four polarization bases in the fully polarized complex scattering matrix data set, generating a focused single-view complex image. Each pixel in the single-view complex image contains the focused complex values ​​of the four polarization channels.

[0036] The imaging and recognition module extracts the total polarization power value of each pixel in the single-view complex image. The total polarization power value is obtained by summing the squares of the moduli of the complex focusing values ​​of the four polarization channels of that pixel. Let the pixel be... The horizontal-horizontal polarization basis focusing complex value is The complex value of the horizontal-vertical polarization base focusing is The complex value of vertical-horizontal polarization base focusing is The complex value of vertical-vertical polarization base focusing is Then the pixel Total polarization power value Calculate according to the following formula: in, Representing pixels The total polarization power, expressed in watts; For a single-view complex image, the distance pixel index is... The range of values ​​is from arrive integers, The total number of distance pixels in a single-view complex image; This represents the orientation pixel number of a single-view complex image. The range of values ​​is from arrive integers, This represents the total number of orientation pixels in a single-view complex image. Representing pixels The horizontal-horizontal polarization basis complex value after point-by-point focusing imaging Representing pixels The horizontal-vertical polarization basis complex value after point-by-point focusing imaging Representing pixels The complex values ​​of the vertical-horizontal polarization basis after point-by-point focusing imaging. Representing pixels The vertical-vertical polarization basis complex value after point-by-point focusing imaging. This represents the operation of taking the square modulo of a complex number.

[0037] The imaging and recognition module compares the total polarization power value of each pixel with a preset power threshold, which is determined based on the statistical distribution of the total polarization power values ​​of all pixels in the single-view complex image. The preset power threshold is set by the imaging and recognition module calculating the mean of the total polarization power values ​​of all pixels in the single-view complex image. and standard deviation Set the preset power threshold to The setting is based on the premise that the probability of a positive deviation exceeding three standard deviation under a Gaussian distribution meets the confidence level requirements for anomaly detection. For pixels with a total polarization power value greater than a preset power threshold, the imaging and recognition module marks the pixel as a candidate mineralized pixel.

[0038] The imaging and recognition module performs spatial connectivity analysis on all candidate mineralized pixels. The spatial connectivity analysis uses the four-connectivity neighborhood criterion, meaning a candidate mineralized pixel is considered connected to its direct neighbors in either the range or azimuth direction. The imaging and recognition module traverses all candidate mineralized pixels in the single-view complex image, merging spatially adjacent candidate mineralized pixels into the same mineralization anomaly patch and assigning a unique patch identifier to each anomaly patch. For each mineralization anomaly patch, the imaging and recognition module calculates the geometric center coordinates and projected area. The geometric center coordinates are calculated by taking the arithmetic mean of the range coordinates of all candidate mineralized pixels within the mineralization anomaly patch, and by taking the arithmetic mean of the azimuth coordinates of all candidate mineralized pixels within the mineralization anomaly patch. The projection area is calculated as follows: count the total number of candidate mineralized pixels in the mineralization anomaly patch, multiply the total number by the ground projection area of ​​a single pixel to obtain the projection area of ​​the mineralization anomaly patch. The ground projection area of ​​a single pixel is determined by the range sampling interval, the azimuth sampling interval and the radar incident angle.

[0039] After acquiring the geometric center coordinates of each mineralization anomaly patch, the imaging and recognition module matches these coordinates with a set of depth positioning parameters. The matching method is as follows: for the pixel positions corresponding to the range and azimuth coordinates of the geometric center of the mineralization anomaly patch, the depth positioning parameters of the corresponding mineralization anomaly are retrieved from the depth positioning parameter set. When a corresponding depth positioning parameter exists for that pixel position, the range and azimuth coordinates of the geometric center of the mineralization anomaly patch are combined with the depth positioning parameters to form the three-dimensional spatial coordinates of the mineralization anomaly corresponding to that patch. The three-dimensional spatial coordinates consist of three parts: the depth value along the radar line of sight, the horizontal distance value perpendicular to the radar line of sight, and the azimuth distance value.

[0040] The imaging and recognition module performs spatial location calibration of the ore-bearing indicator feature vector for each mineralization anomaly patch. The module acquires the cross-polarization phase difference parameter, volume scattering depolarization parameter, and real part parameter of the complex permittivity for each candidate mineralized pixel within the anomaly patch. It then arithmetically averages the cross-polarization phase difference parameters of all candidate mineralized pixels to obtain the average cross-polarization phase difference parameter; arithmetically averages the volume scattering depolarization parameters of all candidate mineralized pixels to obtain the average volume scattering depolarization parameter; and arithmetically averages the real part parameter of the complex permittivity for all candidate mineralized pixels to obtain the average real part parameter of the complex permittivity. The average cross-polarization phase difference parameter, average volume scattering depolarization parameter, and average real part parameter of the complex permittivity together constitute the average feature vector of the mineralization anomaly patch. This average feature vector is the spatial location calibration result of the ore-bearing indicator feature vector at that mineralization anomaly patch.

[0041] The imaging and recognition module combines the three-dimensional spatial coordinates and average feature vector corresponding to each mineralization anomaly patch to generate a composite data record. The format of the composite data record is as follows: the first field stores the depth value in the three-dimensional spatial coordinates; the second field stores the horizontal distance value in the three-dimensional spatial coordinates; the third field stores the azimuth distance value in the three-dimensional spatial coordinates; the fourth field stores the average cross-polarization phase difference parameter; the fifth field stores the average volume scattering depolarization parameter; and the sixth field stores the real part of the average complex permittivity parameter. The imaging and recognition module aggregates the composite data records corresponding to all mineralization anomaly patches into a composite data record set.

[0042] The imaging and recognition module acquires a pre-stored geological map of the mining area, which includes vector layers of geological elements such as stratigraphic boundaries, fault lines, known mineral occurrences, and lithological distribution areas. The module then projects each composite data record from the composite data record set onto the coordinate system of the geological map, achieving spatial overlay between the composite data records and the geological map. After spatial overlay, the module marks the planar location of each mineralization anomaly on the geological map with annotation symbols and outputs the corresponding depth value, average cross-polarization phase difference parameter, average volume scattering depolarization parameter, and average complex permittivity real part parameter of the mineralization anomaly as attribute labels next to the annotation symbols, forming the spatial distribution recognition result of the target mineralization body.

[0043] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. An intelligent remote sensing identification system for mineral resource exploration, characterized in that, The system includes: The acquisition module transmits a linear frequency modulated pulse sequence to the exploration area and acquires the echo signals reflected by ground objects to obtain a set of fully polarized complex scattering matrix data. The tomographic focusing module performs tomographic focusing processing on the fully polarized complex scattering matrix data set, and separates the surface scattering component and the deep volume scattering component based on the difference in radar wave penetration depth under different polarization bases, thereby obtaining the depth positioning parameter set of the mineralization anomaly. The parameter extraction module extracts the cross-polarization phase difference parameter, the volume scattering depolarization parameter, and the real part parameter of the complex permittivity related to the mineral lattice structure from the depth positioning parameter set, and generates a mineral-bearing indicator feature vector. The step size determination module uses the mineralized indicator feature vector as an index to perform a neighborhood search in a pre-built remote sensing detection strategy library for mining areas, and determines a variable-scale detection step size that matches the current mineralization anomaly size. The imaging and recognition module performs point-by-point focusing imaging on the fully polarized complex scattering matrix data set according to the variable scale detection step size, generates a multi-polarization synthetic aperture radar image, and uses the multi-polarization synthetic aperture radar image to spatially calibrate the ore-bearing indicator feature vector to obtain the spatial distribution recognition result of the target mineralization body.

2. The intelligent remote sensing identification system for mineral resource exploration as described in claim 1, characterized in that, The acquisition module includes the following execution: The linear frequency modulated pulse sequence is grouped and transmitted according to different polarization bases, including horizontal-horizontal polarization base, horizontal-vertical polarization base, vertical-horizontal polarization base, and vertical-vertical polarization base; The emission timestamp of each pulse under each polarization base is recorded synchronously, and the echo signal is compressed in the range direction according to the emission timestamp to obtain the data set of the fully polarized complex scattering matrix.

3. The intelligent remote sensing identification system for mineral resource exploration as described in claim 1, characterized in that, The tomography focusing module includes the following execution: Arrange the data of each pixel in the fully polarized complex scattering matrix data set in order of time delay to construct a complex scattering intensity profile along the radar line of sight. The complex scattering intensity profile is subjected to deskewing processing on a depth-by-depth basis to separate the surface scattering component corresponding to the surface roughness of the ground object and the deep body scattering component corresponding to the structure of the underground target body. The depth cell position corresponding to the energy peak in the deep volume scattering component is extracted and used as the depth positioning parameter of the mineralization anomaly.

4. The intelligent remote sensing identification system for mineral resource exploration as described in claim 1, characterized in that, The parameter extraction module includes the following execution: Phase solution is performed on the co-polarized components and cross-polarized components under the linear polarization basis in the fully polarized complex scattering matrix data set. The phase difference between the cross-polarized components and the co-polarized components is calculated and used as the cross-polarization phase difference parameter. The fully polarized complex scattering matrix data set is polarized to obtain the power proportions of odd scattering, even scattering, and volume scattering. The ratio of the power proportion of volume scattering to the total scattering power is used as the volume scattering depolarization parameter. The amplitude value of the backscattering coefficient of the deep volume scattering component at the system center frequency is extracted, and the amplitude value is inverted according to the pre-calibrated incident angle-dielectric constant mapping relationship. The inversion result is used as the real part parameter of the complex dielectric constant.

5. The intelligent remote sensing identification system for mineral resource exploration as described in claim 1, characterized in that, The step size determination module includes the following execution: Multiple historical ore-bearing indicator feature vectors and corresponding multiple historical variable-scale detection step sizes are obtained as retrieval samples; The multiple historical ore-bearing indicator feature vectors are mapped to a multidimensional space with electromagnetic scattering parameters as the axis, and the points in the multidimensional space are labeled with the corresponding multiple historical variable-scale detection step sizes to construct the remote sensing detection strategy library for the mining area. In the multidimensional space, a dynamic search neighborhood is defined with the spatial point corresponding to the current ore-bearing indicator feature vector as the center, and the median of all historical variable-scale detection step sizes in the dynamic search neighborhood is calculated as the variable-scale detection step size.

6. The intelligent remote sensing identification system for mineral resource exploration as described in claim 5, characterized in that, The step size determination module includes the following execution: Extract all historical mineral-bearing indicator feature vectors within the dynamic search neighborhood, and calculate the Euclidean distance between each historical mineral-bearing indicator feature vector and the current mineral-bearing indicator feature vector to obtain a distance set; The distance values ​​in the distance set are sorted from smallest to largest, and the distance located at a preset quantile value is selected as the radius of the dynamic search neighborhood.

7. The intelligent remote sensing identification system for mineral resource exploration as described in claim 1, characterized in that, The imaging and recognition module includes the following functions: Using the variable-scale detection step size as the pulse accumulation point of the azimuth matched filter, the signal at each azimuth position in the fully polarized complex scattering matrix data set is subjected to point-by-point focusing imaging to generate a focused single-view complex image. Extract the total polarization power value of each pixel in the single-view complex image, and select the pixels with the total polarization power value greater than a preset power threshold as candidate mineralized pixels.

8. The intelligent remote sensing identification system for mineral resource exploration as described in claim 7, characterized in that, The imaging and recognition module includes the following functions: Spatial connectivity analysis is performed on the candidate mineralized pixels to merge spatially adjacent candidate mineralized pixels into the same mineralization anomalous patch, and the geometric center coordinates and projected area of ​​each mineralization anomalous patch are calculated. The geometric center coordinates are matched with the depth positioning parameters to obtain the three-dimensional spatial coordinates of the mineralization anomaly.

9. The intelligent remote sensing identification system for mineral resource exploration as described in claim 8, characterized in that, The imaging and recognition module includes the following functions: The average feature vector of the patch is obtained by averaging the cross-polarization phase difference parameter, the volume scattering depolarization parameter, and the real part parameter of the complex permittivity of all pixels in the mineralized anomaly patch, and is used as the spatial location calibration result of the mineralized indicator feature vector.

10. The intelligent remote sensing identification system for mineral resource exploration as described in claim 9, characterized in that, The imaging and recognition module includes the following functions: The three-dimensional spatial coordinates are combined with the average feature vector to generate a composite data record containing spatial location information and mineralization attributes. The composite data record is then spatially overlaid with a pre-stored geological map of the mining area to output the spatial distribution identification result of the target mineralization body.

Images