Uranium mineralization zone identification method and system based on multi-source remote sensing data fusion

Abstract

The invention discloses a uranium mineralization zone identification method and system based on multi-source remote sensing data fusion, and relates to the technical field of geological exploration, and the method comprises a multi-source data obtaining module which obtains multi-platform and multi-band remote sensing data, and the remote sensing data comprises optical remote sensing data, microwave remote sensing data and auxiliary data; constructing a uranium ore zone feature database based on the remote sensing data; the data preprocessing module is used for carrying out noise elimination, atmospheric interference removal and geometric distortion correction on the remote sensing data; the feature extraction and fusion module is used for extracting uranium mineralization zone sensitive features from the preprocessed multi-source data and carrying out cross-modal fusion, and the cross-modal fusion comprises feature layer fusion and decision-making layer fusion; an anomaly detection and geological analysis module; a result output and verification module; according to the method, all-weather, total-factor and non-blind-area coverage of a complex terrain area can be achieved, the exploration range is not limited to roads and gentle zones any more, and the coverage rate of a potential metallogenic zone is increased.

Description

Technical Field

[0001] This invention belongs to the field of geological exploration technology, specifically relating to a method and system for identifying uranium mineralization zones based on multi-source remote sensing data fusion. Background Technology

[0002] Uranium mineralization zones are linear or banded geological structural units with mineral enrichment characteristics discovered during geological exploration. Their identification requires a combination of geological, geophysical, and geochemical methods. Currently, the identification of uranium mineralization zones mainly relies on a combination of traditional geological exploration and single remote sensing technology. That is, based on ground geological mapping, combined with artificial radiometric measurements (such as field detection by gamma spectrometer), soil geochemical sampling and analysis, and geophysical exploration (gravity and magnetic measurements), the mineralization anomaly area is directly delineated.

[0003] However, due to the difficulty of personnel and equipment access in complex terrain areas such as deep mountains and canyons, swamps and wetlands, and high-altitude glaciers, exploration areas can only be explored around roads and flat areas, resulting in a large number of blind spots and an inability to fully cover potential mineralized zones. Furthermore, geological mapping requires personnel to record lithology and structural orientation point by point, and manual radiometric measurement is inefficient. Therefore, we need to propose a method and system for identifying uranium mineralization zones based on multi-source remote sensing data fusion to solve the above-mentioned shortcomings of limited coverage and time-consuming and labor-intensive methods. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for identifying uranium mineralization zones based on multi-source remote sensing data fusion. Through multi-source remote sensing data fusion and intelligent processing, it can be applied to rapid exploration of complex terrains, improve the accuracy of exploration of different terrains, and solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A uranium mineralization zone identification system based on multi-source remote sensing data fusion includes: a multi-source data acquisition module, which acquires remote sensing data from multiple platforms and multiple bands, including optical remote sensing data, microwave remote sensing data and auxiliary data, and constructs a uranium mineralization zone feature database based on the remote sensing data; The data preprocessing module performs noise reduction, atmospheric interference removal, and geometric distortion correction on remote sensing data to achieve spatiotemporal registration of multi-source data. The feature extraction and fusion module extracts sensitive features of uranium mineralization zones from preprocessed multi-source data and performs cross-modal fusion, which includes feature-level fusion and decision-level fusion. The anomaly detection and geological analysis module identifies uranium mineralization anomaly areas and uses geological models to predict mineralization. The results output and verification module is used to generate a uranium mineralization zone distribution map and perform cross-validation and drilling data verification to correct the parameters of the geological model.

[0006] Preferably, the multi-source data acquisition module includes a spectral imager, a multispectral camera, a synthetic aperture radar, a stereo mapping camera, and an airborne gamma spectrometer. The spectral imager, multispectral camera, synthetic aperture radar, stereo mapping camera, and airborne gamma spectrometer are all connected to the data preprocessing module through a data interface.

[0007] Preferably, when acquiring remote sensing data, the multi-source data acquisition module determines the satellite transit time and air flight route to be covered according to the preset monitoring range of the mining area. It simultaneously collects data through a spectral imager, a multispectral camera, a synthetic aperture radar, a stereo mapping camera, and an airborne gamma spectrometer. The data interface receives the collected raw data, parses the data format, extracts metadata, and packages the raw data and metadata into a uranium ore belt feature database.

[0008] Preferably, the data preprocessing module includes: a radiation correction unit, which performs radiation correction using an absolute calibration algorithm and a 6S atmospheric correction model; Geometric registration units are performed using a quadratic polynomial transformation algorithm. The quality control unit automatically detects the integrity and accuracy of the data processed by the data radiation correction unit and geometric registration unit, and transmits qualified data to the feature extraction and fusion module, while transmitting unqualified data back to the multi-source data acquisition module for re-acquisition.

[0009] Preferably, during data preprocessing, optical data from remote sensing data received by the multi-source data acquisition module undergoes radiometric correction, while microwave data and auxiliary data undergo denoising. The radiometrically corrected optical data, denoised microwave data, and auxiliary data are then geometrically registered using Gaofen-2 data as the base map. The quality control unit detects the integrity and accuracy of the processed data to eliminate noise, atmospheric interference, and geometric distortion in the original data, transforming the non-standardized original data into clean data that is spatiotemporally consistent and meets accuracy standards.

[0010] Preferably, the sensitive features of the uranium mineralization zone include spectral features, texture features and deep learning features. The spectral feature extraction uses the SAM algorithm to calculate the similarity angle between the spectrum to be classified and the reference spectrum of the uranium alteration mineral, and at the same time calculates the alteration intensity index. Texture feature extraction using GLCM The image contrast, which reflects the degree of structural fragmentation, and the entropy, which reflects the texture complexity, are calculated under the window. Deep learning feature extraction inputs preprocessed multi-source data into a fine-tuned ResNet-50 network and outputs a 2048-dimensional feature vector.

[0011] Preferably, during cross-modal fusion, the received preprocessed data is processed by extracting features from each channel in the order of spectral, texture, and deep learning. The extracted features are then concatenated to form a multidimensional feature vector. The feature state is then predicted using a state transition matrix, and the error is corrected using an observation matrix. Finally, a denoised and highly correlated fused feature set is output. The signal-to-noise ratio of the fused features is then calculated. If the signal-to-noise ratio is higher than 20dB, the feature is pushed to the anomaly detection and geological analysis module; otherwise, the parameters of the Kalman filter algorithm are readjusted.

[0012] Preferably, the anomaly detection and geological analysis module includes: an anomaly detection unit, which identifies uranium mineralization anomaly areas based on a local outlier factor algorithm; The geological analysis unit transforms preliminary anomalous areas into favorable mineralization target areas by overlaying geological background data and 3D modeling.

[0013] Preferably, the result output and verification module includes: a result generation unit, used for vector graphic generation, 3D model rendering, and anomaly probability map drawing; The validation unit uses cross-validation and drilling data validation to improve the overall accuracy of the geological model. During cross-validation, known uranium mineralization areas are selected as true value samples, and the confusion matrix of the model recognition results is calculated. The model optimization unit includes an error tracing process, a parameter iteration process, and a termination condition process.

[0014] Based on the above-described uranium mineralization zone identification system based on multi-source remote sensing data fusion, this invention also provides a uranium mineralization zone identification method based on multi-source remote sensing data fusion, comprising the following steps: S1. Utilize multiple platforms and multiple bands to acquire multi-source remote sensing data, including optical remote sensing data, microwave remote sensing data, and auxiliary data, and construct a database of uranium ore belt characteristics. S2. Perform radiometric correction and geometric registration on the acquired multi-source remote sensing data; S3. Extract sensitive features of uranium mineralization zones from the processed multi-source remote sensing data. Sensitive features of uranium mineralization zones include spectral features, texture features, and deep learning features. S4. Concatenate the spectral features, texture features, and deep learning features into a multi-dimensional feature vector, and then use the Kalman filter algorithm to perform decision layer fusion on the multi-dimensional feature vector to obtain the fused feature set. S5. Based on the fusion feature set, the local outlier factor algorithm is used to detect outliers in the multi-feature space, and then the outliers are classified to obtain the preliminary uranium mineralization anomaly region. S6. Perform GIS spatial overlay analysis on the preliminary uranium mineralization anomaly area, fault structure data, and lithological distribution data. Combine the three-dimensional terrain model constructed by DEM and the underground structural information inverted from SAR data to comprehensively judge the mineralization conditions and delineate favorable target areas for uranium mineralization. S7. Verify the accuracy of the delineated uranium mineralization target area by combining cross-validation and drilling verification. Adjust the feature extraction parameters in step S3 and the fusion algorithm parameters in step S4 based on the verification results.

[0015] The present invention proposes a method and system for identifying uranium mineralization zones based on multi-source remote sensing data fusion, which has the following advantages compared with existing technologies: 1. This invention achieves all-weather, all-element, and blind-spot-free coverage of complex terrain areas through the coordinated operation of a multi-source data acquisition module, a data preprocessing module, a feature extraction and fusion module, an anomaly detection and geological analysis module, and a result output and verification module. The exploration scope is no longer limited to roads and flat areas, thus improving the coverage of potential mineralized zones.

[0016] 2. The feature extraction and fusion module of this invention eliminates the limitations of single data by fusing the feature layer and the decision layer at two levels, enhances the identification of mineralization features under complex terrain, and improves the accuracy of feature identification.

[0017] 3. The data preprocessing module of this invention performs radiometric correction and geometric registration through a radiometric correction unit and a geometric registration unit, which can compress the in-form exploration cycle and achieve rapid response. Attached Figure Description

[0018] Figure 1 A system block diagram according to an embodiment of the present invention is shown; Figure 2 A block diagram of a data preprocessing module according to an embodiment of the present invention is shown; Figure 3 A block diagram of the result output and verification module according to an embodiment of the present invention is shown; Figure 4 A flowchart of a method according to an embodiment of the present invention is shown. Detailed Implementation

[0019] 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, and not all embodiments. The specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention. 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.

[0020] This invention provides, for example Figure 1-3 The system shown is a uranium mineralization zone identification system based on multi-source remote sensing data fusion, including a multi-source data acquisition module, a data preprocessing module, a feature extraction and fusion module, an anomaly detection and geological analysis module, and a result output and verification module. The multi-source data acquisition module is used to acquire remote sensing data from multiple platforms and multiple bands. The remote sensing data includes optical remote sensing data, microwave remote sensing data, and auxiliary data. A uranium mineralization zone feature database is constructed based on the remote sensing data. The multi-source data acquisition module includes a spectral imager, a multispectral camera, a synthetic aperture radar, a stereo mapping camera, and an airborne gamma spectrometer. All of these components are connected to the data preprocessing module via data interfaces. The spectral imager is a Gaofen-5 hyperspectral imager with a filter range of 0.4-2.5 μm and ≥330 bands. The multispectral camera is a Gaofen-2 panchromatic or multispectral camera with 1m panchromatic resolution and 4M multispectral resolution. Optical remote sensing data is acquired through the spectral imager and the multispectral camera. The synthetic aperture radar is the Gaofen-3 synthetic aperture radar. The synthetic aperture radar operates in the C-band and its resolution is adjustable in the range of 1-100m. It is used to acquire microwave remote sensing data. The stereo mapping camera is the Gaofen-7 stereo mapping camera, with an elevation accuracy of 1m; the airborne gamma spectrometer detects the radioactivity intensity of uranium, thorium, and potassium, and is used to collect auxiliary data. The data interface is compatible with HDF5 and GeoTIFF formats and supports automatic parsing of satellite orbit parameters. It captures raw data periodically through the satellite ground station, and the data interface automatically parses HDF5 and GeoTIFF files, extracts metadata, and temporarily stores it. When acquiring remote sensing data, the multi-source data acquisition module determines the satellite transit time and air flight route to be covered according to the preset mining area monitoring range (the preset mining area monitoring range is set based on the combination of geological exploration foundation, monitoring target requirements and data acquisition technology feasibility to ensure that the range covers key areas and can efficiently acquire effective data). Data is collected simultaneously through a spectral imager, multispectral camera, synthetic aperture radar, stereo mapping camera and airborne gamma spectrometer. The data interface receives the collected raw data, parses the data format and extracts metadata. The raw data and metadata are packaged and temporarily stored in the uranium ore belt feature database. The module acquires multi-source raw data required for uranium mineralization belt identification in a full range of ways and with high timeliness, while ensuring the integrity of the data and the accuracy of the metadata, providing a raw material basis for subsequent preprocessing and feature extraction. The uranium ore belt feature database is used to centrally store multi-type, multi-platform remote sensing data. It constructs a multi-dimensional index based on spatiotemporal dimensions, data types, and feature types to support fast retrieval. The data preprocessing module is used to perform noise reduction, atmospheric interference removal, and geometric distortion correction on remote sensing data, so as to achieve spatiotemporal registration of multi-source data. The data preprocessing module includes a radiation correction unit, a geometric registration unit, and a quality control unit. The radiation correction unit and the geometric registration unit are both electrically connected to the quality control unit. The radiation correction unit uses an absolute calibration algorithm and a 6S atmospheric correction model for radiation correction. The formula for the absolute scaling algorithm is: , in, Radiance (unit: W·m) -2 ·sr -1 ·μm -1 ), where DN is the digital quantization value, For sensor gain, This is the sensor offset coefficient; The formula for the 6S atmospheric correction model is: , in, Let L be the surface reflectance and L be the total radiance received by the sensor. For path radiation, Atmospheric transmittance, Downward irradiance, The value is 3.141529, used to correlate radiance L with irradiance. The physical dimensions are unified.

[0021] The geometric registration unit uses a quadratic polynomial transformation algorithm for geometric registration. The formula for the quadratic polynomial transformation algorithm is: , Where (x,y) are the original coordinates, , () represents the corrected coordinates. , , , , , and , , , , , These are the polynomial coefficients, which are solved using the least squares method. The quality control unit automatically detects the integrity and accuracy of the data processed by the data radiation correction unit and geometric registration unit, and transmits qualified data to the feature extraction and fusion module, while transmitting unqualified data back to the multi-source data acquisition module for re-acquisition.

[0022] During data preprocessing, optical data from remote sensing data received by the multi-source data acquisition module undergoes radiometric correction, while microwave data and auxiliary data undergo denoising. Then, the radiometrically corrected optical data, denoised microwave data, and auxiliary data are geometrically registered using Gaofen-2 data as the base map. The quality control unit detects the integrity and accuracy of the processed data to eliminate noise, atmospheric interference, and geometric distortion in the original data, transforming non-standardized original data into clean data that is spatiotemporally consistent and meets accuracy standards, thereby improving the accuracy of feature extraction. The feature extraction and fusion module extracts sensitive features of uranium mineralization zones from preprocessed multi-source data and performs cross-modal fusion, which includes feature layer fusion and decision layer fusion. Sensitive features of uranium mineralization zones include spectral features, texture features, and deep learning features. Spectral feature extraction uses the SAM algorithm to calculate the similarity angle between the spectrum to be classified and the reference spectrum of the altered minerals of uranium, and at the same time calculates the alteration intensity index. The formula for calculating the similarity angle is: , in, The angle between the reference spectral vector and the spectral vector to be classified is denoted as n, where n is the total number of bands shared by the reference spectrum and the spectrum to be classified. The value of the i-th band of the reference spectrum. Let i be the value of the i-th band of the spectrum to be classified; Texture feature extraction using GLCM The image contrast, which reflects the degree of structural fragmentation, and the entropy, which reflects the texture complexity, are calculated under the window. Deep learning feature extraction inputs preprocessed multi-source data into a fine-tuned ResNet-50 network and outputs a 2048-dimensional spatial feature vector to capture multi-scale construction information.

[0023] Feature layer fusion is to concatenate spectral features, texture features and deep learning features into a multi-dimensional feature vector. For example, spectral features are 2-dimensional, texture features are 2-dimensional, and deep learning features are 2048-dimensional, which are concatenated into a 2052-dimensional feature vector. The decision-level fusion uses the Kalman filter algorithm to estimate the state of the multi-dimensional feature vectors, suppresses feature noise through the noise covariance matrix, and outputs a fused feature set. , in, Let A be the prior state estimate at time k, and H be the state transition matrix and the observation matrix. Let B be the posterior state estimate of time k, and let B be the control input matrix. Let k be the control input vector at the previous time step. Let k be the prior covariance matrix at time k. Let k be the posterior covariance matrix at time k. Let Q be the transpose of the state transition matrix, and let Q be the process noise covariance matrix. Here is the Kalman gain matrix. For the transpose of the observation matrix, Let R be the observation prediction error covariance, and R be the observation noise covariance matrix. The operation of inverting a matrix. This is the vector of observations at time k. Let I be the observed prediction based on the prior state, and let I be the identity matrix. During cross-modal fusion, the received preprocessed data is processed by channel-wise feature extraction in the order of spectrum, texture, and deep learning. The extracted features are then concatenated to form a multi-dimensional feature vector. The feature state is then predicted using a state transition matrix, and the error is corrected using an observation matrix. Finally, the denoised and highly correlated fused feature set is output. The signal-to-noise ratio of the fused features is calculated. If the signal-to-noise ratio is higher than 20dB, it is pushed to the anomaly detection and geological analysis module; otherwise, the parameters of the Kalman filter algorithm are readjusted. The anomaly detection and geological analysis module identifies uranium mineralization anomaly areas and uses geological models to predict mineralization. The anomaly detection and geological analysis module includes an anomaly detection unit and a geological analysis unit. The anomaly detection unit identifies uranium mineralization anomaly regions based on the local outlier factor algorithm. The procedure for identifying uranium mineralization anomalies is as follows: A1. Normalize the multidimensional feature vectors to eliminate differences in the dimensions of different features; A2. Calculate the multi-feature spatial distance between each test point p and all other points, and select the q points closest to the test point to form a nearest neighbor set; The formula for calculating the spatial distance of multiple features is: , in, Let m be the Euclidean distance between point p and point o, and m be the total number of feature dimensions. Let be the i-th eigenvalue of the point p to be measured. Let be the i-th eigenvalue of the nearest neighbor point o; A3. Calculate the reachability distance of any point o in the nearest neighbor set, the local reachability density of the point p to be tested, and the local outlier factor of the q nearest neighbors of point p; The formula for calculating reachable distance is: , in, Let q be the reachable distance from point p to point o. Let q be the nearest neighbor distance of any point o, where q is the number of nearest neighbors. Let be the Euclidean distance between point p and point o; The formula for calculating the local reachability density of the point p to be measured is: , in, Let q be the local reachability density of the nearest neighbors of the point p to be measured. Let q be the reachable distance from point p to point o. Let q be the set of the nearest neighbors of the point p to be tested; The formula for calculating the local outlier factor of point p's q nearest neighbors is: , in, Let q be the local outlier factor of the test point p. This is the density ratio; A4. Compare the q-nearest neighbor local outlier factor of the test point p with the standard threshold. If the q-nearest neighbor local outlier factor of point p is greater than the standard threshold, then mark the test point p as a uranium mineralization anomaly point; otherwise, mark it as a uranium mineralization normal point. The standard threshold can be calibrated using known mineralization zone data. A5. Morphological opening operations are used to remove isolated noise points from all outliers, and then connected component analysis is used to merge discrete outliers into preliminary uranium mineralization anomaly regions. The geological analysis unit transforms preliminary anomalous areas into favorable mineralization target areas by overlaying geological background data and three-dimensional modeling. The conversion process of favorable target areas for mineralization is as follows: B1. Access geological data and perform standardized processing; The vector map of the fault structure in the mining area, the lithological distribution map, auxiliary data and remote sensing data were integrated, and all data were unified to the WGS84 coordinate system to ensure the consistency of the spatial reference. B2. Spatial overlay of anomalous areas with geological elements; The spatial relationship between the preliminary anomaly area and the fault structure is statistically analyzed. If the anomaly area is located within 500m on both sides of the fault zone, it is a mineralization potential enhancement zone; otherwise, it is a non-mineralization favorable target area. Then, it is determined whether the preliminary anomaly area falls within the known ore-bearing rock series. If the matching degree exceeds 70%, it is marked as a lithological favorable area; otherwise, it is marked as a lithological unfavorable area. Then, the slope and aspect of the preliminary anomaly area are calculated based on auxiliary data, and a shape suitability score is assigned. B3. Three-dimensional structural modeling and mineralization potential assessment to delineate favorable target areas for uranium mineralization; Based on microwave remote sensing data, concealed structures within 50m underground are retrieved to determine whether a continuous tectonic fracture zone exists beneath the preliminary anomaly area. A weighted scoring model is constructed based on structure, lithology, topography, and anomaly intensity. The formula for the weighted scoring model is as follows: , Where S is the total score for mineralization potential, and 0.4, 0.3, 0.1, and 0.2 are the weighting coefficients, respectively. To construct a matching score, For lithological suitability scoring, For terrain suitability scoring, Anomaly intensity scoring is used; when the total mineralization potential score exceeds 80 points, the area is delineated as a favorable target area for uranium mineralization. The result output and verification module is used to generate a uranium mineralization zone distribution map and perform cross-validation and drilling data verification to correct the parameters of the geological model.

[0024] The result output and verification module includes a result generation unit, a verification analysis unit, and a model optimization unit. The result generation unit is used for vector map generation, 3D model rendering, and anomaly probability map drawing. Vector map generation converts the favorable target area into an ESRIShapefile format vector map, which includes target area number, score, and core mineral control elements, such as target area A: fault zone + granite + high uranium anomaly. The 3D model rendering is based on the Unity engine, which integrates DEM terrain data, target area vector boundaries, and underground structure inversion results to generate an interactive 3D scene. The anomaly probability map is drawn using the Kriging interpolation method to interpolate the LOF values ​​of discrete anomalies into a continuous uranium mineralization probability surface, with areas having a probability exceeding 70% highlighted in red. The verification unit uses cross-validation and drilling data verification to improve the overall accuracy of the geological model. During cross-validation, known uranium mineralization areas are selected as true samples, and the confusion matrix of the model identification results is calculated. That is, the model identification accuracy is confirmed by calculating the accuracy and recall of the model identification. Cross-validation is used to quantify the overall accuracy of the model and avoid misjudgment of unknown areas. Drilling data validation involves deploying validation boreholes in high-probability target areas and collecting core samples. This involves analyzing the uranium grade and alteration mineral composition of the core samples in the laboratory to calculate the target area drilling coincidence rate, which is the percentage of the number of mineralized boreholes to the total number of validation boreholes. Drilling data validation, combined with physical samples, directly verifies the actual mineralization potential of the target area, ensuring that the results are applicable. The model optimization unit includes an error tracing process, a parameter iteration process, and a termination condition process, which are used to optimize the geological model parameters. The error tracing process analyzes the source of error when the cross-validation accuracy is less than 80% or the drilling consistency is less than 50%. If the false negative rate is high, the q value of the nearest local outlier factor calculation formula is adjusted. If the false positive rate is high, the standard threshold of the nearest local outlier factor calculation formula is increased or the radioactivity intensity is increased. The parameter iteration process is the process of retraining the feature extraction model and optimizing the Kalman filter algorithm based on validation data or observing the noise covariance matrix. The termination condition process stops optimization when the accuracy improvement is less than 2% after three consecutive iterations or the drilling match rate exceeds 70%.

[0025] Based on the above-described uranium mineralization zone identification system based on multi-source remote sensing data fusion, this invention also provides a uranium mineralization zone identification method based on multi-source remote sensing data fusion, comprising the following steps: S1. Utilize multiple platforms and multiple bands to acquire multi-source remote sensing data, including optical remote sensing data, microwave remote sensing data, and auxiliary data, and construct a database of uranium ore belt characteristics. S2. Perform radiometric correction and geometric registration on the acquired multi-source remote sensing data; S3. Extract sensitive features of uranium mineralization zones from the processed multi-source remote sensing data. Sensitive features of uranium mineralization zones include spectral features, texture features, and deep learning features. S4. Concatenate the spectral features, texture features, and deep learning features into a multi-dimensional feature vector, and then use the Kalman filter algorithm to perform decision layer fusion on the multi-dimensional feature vector to obtain the fused feature set. S5. Based on the fusion feature set, the local outlier factor algorithm is used to detect outliers in the multi-feature space, and then the outliers are classified to obtain the preliminary uranium mineralization anomaly region. S6. Perform GIS spatial overlay analysis on the preliminary uranium mineralization anomaly area, fault structure data, and lithological distribution data. Combine the three-dimensional terrain model constructed by DEM and the underground structural information inverted from SAR data to comprehensively judge the mineralization conditions and delineate favorable target areas for uranium mineralization. S7. Verify the accuracy of the delineated uranium mineralization target area by combining cross-validation and drilling verification. Adjust the feature extraction parameters in step S3 and the fusion algorithm parameters in step S4 based on the verification results.

[0026] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A uranium mineralization zone identification system based on multi-source remote sensing data fusion, characterized in that: include: The multi-source data acquisition module acquires remote sensing data from multiple platforms and multiple bands, including optical remote sensing data, microwave remote sensing data, and auxiliary data. Based on the remote sensing data, a database of uranium ore belt characteristics is constructed. The data preprocessing module performs noise reduction, atmospheric interference removal, and geometric distortion correction on remote sensing data to achieve spatiotemporal registration of multi-source data. The feature extraction and fusion module extracts sensitive features of uranium mineralization zones from preprocessed multi-source data and performs cross-modal fusion, which includes feature-level fusion and decision-level fusion. The anomaly detection and geological analysis module identifies uranium mineralization anomaly areas and uses geological models to predict mineralization. The results output and verification module is used to generate a uranium mineralization zone distribution map and perform cross-validation and drilling data verification to correct the parameters of the geological model.

2. The uranium mineralization zone identification system based on multi-source remote sensing data fusion according to claim 1, characterized in that: The multi-source data acquisition module includes a spectral imager, a multispectral camera, a synthetic aperture radar, a stereo mapping camera, and an airborne gamma spectrometer. The spectral imager, multispectral camera, synthetic aperture radar, stereo mapping camera, and airborne gamma spectrometer are all connected to the data preprocessing module through a data interface.

3. The uranium mineralization zone identification system based on multi-source remote sensing data fusion according to claim 2, characterized in that: When acquiring remote sensing data, the multi-source data acquisition module determines the satellite transit time and air flight route to be covered according to the preset monitoring range of the mining area. It simultaneously collects data through a spectral imager, multispectral camera, synthetic aperture radar, stereo mapping camera, and airborne gamma spectrometer. The data interface receives the collected raw data, parses the data format, extracts metadata, and packages the raw data and metadata into a uranium ore belt feature database.

4. The uranium mineralization zone identification system based on multi-source remote sensing data fusion according to claim 3, characterized in that: The data preprocessing module includes a radiation correction unit, which performs radiation correction using an absolute calibration algorithm and a 6S atmospheric correction model. Geometric registration units are performed using a quadratic polynomial transformation algorithm. The quality control unit automatically detects the integrity and accuracy of the data processed by the data radiation correction unit and geometric registration unit, and transmits qualified data to the feature extraction and fusion module, while transmitting unqualified data back to the multi-source data acquisition module for re-acquisition.

5. A uranium mineralization zone identification system based on multi-source remote sensing data fusion according to claim 4, characterized in that: During data preprocessing, optical data from remote sensing data received by the multi-source data acquisition module undergoes radiometric correction, while microwave and auxiliary data undergo denoising. The radiometrically corrected optical data, denoised microwave and auxiliary data are then geometrically registered using Gaofen-2 data as the base map. The quality control unit checks the integrity and accuracy of the processed data to eliminate noise, atmospheric interference, and geometric distortion in the original data, transforming the non-standardized original data into clean data that is spatiotemporally consistent and meets accuracy standards.

6. A uranium mineralization zone identification system based on multi-source remote sensing data fusion according to claim 5, characterized in that: The sensitive features of the uranium mineralization zone include spectral features, texture features and deep learning features. The spectral feature extraction uses the SAM algorithm to calculate the similarity angle between the spectrum to be classified and the reference spectrum of the uranium alteration mineral, and at the same time calculates the alteration intensity index. Texture feature extraction using GLCM The image contrast, which reflects the degree of structural fragmentation, and the entropy, which reflects the texture complexity, are calculated under the window. Deep learning feature extraction inputs preprocessed multi-source data into a fine-tuned ResNet-50 network and outputs a 2048-dimensional feature vector.

7. A uranium mineralization zone identification system based on multi-source remote sensing data fusion according to claim 6, characterized in that: During cross-modal fusion, the received preprocessed data is processed by channel-wise feature extraction in the order of spectral, texture, and deep learning. The extracted features are then concatenated to form a multi-dimensional feature vector. The feature state is then predicted using a state transition matrix, and the error is corrected using an observation matrix. Finally, a denoised and highly correlated fused feature set is output. The signal-to-noise ratio of the fused features is calculated. If the signal-to-noise ratio is higher than 20dB, it is pushed to the anomaly detection and geological analysis module; otherwise, the parameters of the Kalman filter algorithm are readjusted.

8. A uranium mineralization zone identification system based on multi-source remote sensing data fusion according to claim 7, characterized in that: The anomaly detection and geological analysis module includes: an anomaly detection unit, which identifies uranium mineralization anomaly areas based on the local outlier factor algorithm; The geological analysis unit transforms preliminary anomalous areas into favorable mineralization target areas by overlaying geological background data and 3D modeling.

9. A uranium mineralization zone identification system based on multi-source remote sensing data fusion according to claim 8, characterized in that: The result output and verification module includes: a result generation unit, used for vector graphic generation, 3D model rendering, and anomaly probability map drawing; The validation unit uses cross-validation and drilling data validation to improve the overall accuracy of the geological model. During cross-validation, known uranium mineralization areas are selected as true value samples, and the confusion matrix of the model recognition results is calculated. The model optimization unit includes an error tracing process, a parameter iteration process, and a termination condition process.

10. A method for identifying uranium mineralization zones based on multi-source remote sensing data fusion, based on the uranium mineralization zone identification system based on multi-source remote sensing data fusion as described in any one of claims 1-9, characterized in that: Includes the following steps: S1. Utilize multiple platforms and multiple bands to acquire multi-source remote sensing data, including optical remote sensing data, microwave remote sensing data, and auxiliary data, and construct a database of uranium ore belt characteristics. S2. Perform radiometric correction and geometric registration on the acquired multi-source remote sensing data; S3. Extract sensitive features of uranium mineralization zones from the processed multi-source remote sensing data. Sensitive features of uranium mineralization zones include spectral features, texture features, and deep learning features. S4. Concatenate the spectral features, texture features, and deep learning features into a multi-dimensional feature vector, and then use the Kalman filter algorithm to perform decision layer fusion on the multi-dimensional feature vector to obtain the fused feature set. S5. Based on the fusion feature set, the local outlier factor algorithm is used to detect outliers in the multi-feature space, and then the outliers are classified to obtain the preliminary uranium mineralization anomaly region. S6. Perform GIS spatial overlay analysis on the preliminary uranium mineralization anomaly area, fault structure data, and lithological distribution data. Combine the three-dimensional terrain model constructed by DEM and the underground structural information inverted from SAR data to comprehensively judge the mineralization conditions and delineate favorable target areas for uranium mineralization. S7. Verify the accuracy of the delineated uranium mineralization target area by combining cross-validation and drilling verification. Adjust the feature extraction parameters in step S3 and the fusion algorithm parameters in step S4 based on the verification results.

Images