Mineral geological exploration system based on remote sensing image texture analysis

Abstract

The invention relates to the technical field of remote sensing geological information processing, and discloses a mineral geological exploration system based on remote sensing image texture analysis, which comprises a time sequence base line library construction unit, an ecological proxy texture channel processing module, a remote sensing image texture analysis unit and a remote sensing image texture analysis unit, the system comprises a normalization vegetation index texture purification module for purifying a normalization vegetation index texture of a current image according to a time sequence statistical baseline so as to generate an ecological proxy texture anomaly graph, a texture channel construction processing module for generating a texture anomaly graph, and a collaborative verification engine module for performing collaborative verification on the texture anomaly graph before executing spatial coupling judgment. According to the method, the information quality of two channels is evaluated according to image information entropy, verification logic is dynamically selected, earth surface covering information regarded as interference in traditional exploration is converted into an independent verification dimension, and an evidence chain of internal cross verification is constructed through double constraints of space and time and self-adaptive evaluation of the information quality.

Description

TECHNICAL FIELD

[0001] The present application relates to a mineral geological exploration system based on remote sensing image texture analysis, and belongs to the technical field of remote sensing geological information processing. BACKGROUND

[0002] In the field of mineral geological exploration services using high technology, remote sensing technology has become a common technical means in the early stage of general survey and target area screening due to its macroscopic, rapid and economical characteristics. The main application is to identify geological structures, lithological anomalies or alteration information related to mineralization by interpreting the spectral or texture information in the image.

[0003] However, in most areas of the earth's land covered by vegetation, soil and other coverings, the application of the above technology faces a fundamental dilemma. In order to obtain the information of the underlying bedrock, the current mainstream technical approach generally follows a direct reduction logic, that is, it tries to strip the influence of the surface coverings in order to directly observe the underlying geological information by using complex image algorithms or relying on high-cost sensors with certain penetration capabilities. In the exploration practice, especially in the general survey stage with strict economic requirements, this approach has become a technical trade-off, that is, either to invest in higher exploration costs for the pursuit of direct information, which violates the economic principle of early exploration, or to use universal remote sensing data, which inevitably introduces artifacts and uncertainties in the stripping process, resulting in a decrease in the reliability of the interpretation results.

[0004] In order to get rid of this dilemma, the industry has been committed to developing more advanced sensors or more sophisticated algorithms, but these efforts are still essentially improvements on the original direct reduction approach, and have not been able to break out of their inherent limitations. This technical inertia has caused the industry to ignore a key possibility for a long time, that is, the texture characteristics of the spatial distribution of surface coverings, which are obstacles to exploration, are actually a kind of surface ecological response under the long-term and comprehensive action of deep geological environment, and contain rich and interpretable indirect information. This response mechanism is mainly established through two ways: 1. Deep mineralization bodies will change the geochemical environment of the overlying soil through weathering leaching and other geological processes, forming element anomalies that are stressful or selective to plant growth. This long-term geochemical pressure can cause observable systematic differences in species composition, growth density and health status of surface vegetation; 2. Geological structures such as faults and fracture zones related to mineralization will directly affect the hydrological conditions and physical properties of local soil, and then regulate the type and growth of vegetation, forming linear or ring-shaped ecological anomalies along the structure. Therefore, how to create an information processing method that no longer tries to strip or penetrate the surface covering, but for the first time uses the texture characteristics of the covering itself as a proxy information channel independent of geological structure information, and through the coordinated verification of information from the two channels, to achieve reliable identification of concealed geological anomalies relying only on standard and universal remote sensing data, has become the technical problem to be solved by the present invention. SUMMARY

[0005] The present application provides a mineral geological exploration system based on remote sensing image texture analysis, which mainly aims to solve the dilemma that the existing technology is trapped in when dealing with remote sensing images in vegetation-covered areas, that is, the cost and reliability are difficult to balance due to adherence to the direct reduction logic, especially how to use universal remote sensing data to achieve high-precision and low-cost identification of concealed geological anomalies through an information processing method that converts surface covering information from noise to proxy signal.

[0006] To achieve the above purpose, the mineral geological exploration system based on remote sensing image texture analysis provided by the present application comprises:

[0007] A time series baseline library construction unit acquires multiple historical synchronous remote sensing images of the exploration area, and calculates the statistical parameters of the normalized vegetation index texture of the historical synchronous remote sensing images to generate a time series statistical baseline defining the normal fluctuation range;

[0008] An ecological proxy texture channel processing module calculates the normalized vegetation index map of the current remote sensing image, performs texture analysis on the normalized vegetation index map, and compares the texture analysis result with the time series statistical baseline to generate an ecological proxy texture anomaly map containing only the area where the texture value exceeds the normal fluctuation range;

[0009] A texture channel processing module is used to acquire shortwave infrared images that are insensitive to vegetation and perform texture analysis on the images to generate texture anomaly maps.

[0010] A collaborative verification engine module receives an ecological proxy texture anomaly map and a constructed texture anomaly map, calculates the image information entropy value of the two maps, and compares the information entropy value with a benchmark interval obtained from multi-sample statistics. When the information entropy values ​​of the two maps are both within the benchmark interval, strong verification logic is executed to identify the spatially coupled anomaly region in the two maps as the target area for mineral exploration.

[0011] Preferably, when the collaborative verification engine module executes strong verification logic, it uses a coupling index C determined by both spatial overlap and spatial proximity. index To determine whether anomaly regions are spatially coupled, the coupling index C is used. index The calculation rules are as follows: Among them, A overlap To construct the spatial overlap area between the anomalous connected clusters in the texture anomaly map and the anomalous connected clusters in the ecological proxy texture anomaly map, A T d is the sum of the areas of two connected clusters. min Let d be the minimum edge distance between two connected clusters. max To find the maximum search distance, w1 and w2 are non-negative weight coefficients that satisfy w1 + w2 = 1. The collaborative verification engine module will use the coupling index C. index Regions exceeding the judgment threshold are identified as spatially coupled abnormal regions.

[0012] Preferably, both the ecological proxy texture channel processing module and the constructed texture channel processing module use the gray-level co-occurrence matrix algorithm to perform texture analysis. Specifically, the constructed texture channel processing module calculates at least one texture parameter among the contrast or heterogeneity of the shortwave infrared band image, and the ecological proxy texture channel processing module calculates at least one texture parameter among the variance or entropy of the normalized vegetation index map.

[0013] Preferably, when either of the two graphs' information entropy values ​​is not within the baseline range, the collaborative verification engine module executes weak verification logic. It takes the abnormal region in the abnormal graph generated by the channel whose information entropy value is within the baseline range as the main target, and searches for edge responses in the other channel only within the bandwidth range of the main target to confirm the main target.

[0014] Preferably, when the exploration area is determined to be a homogeneous lithological cover area, the structural texture channel processing module selects remote sensing images with low solar altitude angles acquired at sunrise or sunset in that area, and performs edge detection on the remote sensing images to extract the shadow distribution formed by the slight topographic undulations to generate a structural texture anomaly map.

[0015] Preferably, the system further comprises a terrain context modulation module, which: calculates a terrain wetness index map of the survey area based on a digital elevation model; and the eco-proxy texture channel processing module, when generating the time-series statistical baseline, spatially modulates the normal fluctuation range according to the terrain wetness index map, expands the limit values of the normal fluctuation range in areas with high terrain wetness index values, and narrows the limit values of the normal fluctuation range in areas with low terrain wetness index values.

[0016] Preferably, the system further comprises a structural permeability detection module, which: after the survey area experiences a rainfall event, automatically obtains normalized difference water index images of the area at two different time points when the soil is close to saturation and when the soil is dry, generates an image representing the spatial difference of soil water decay rate by pixel-by-pixel subtraction of the two normalized difference water index images, and the verification engine module cooperates to determine the mineral exploration target area by taking the area with abnormal water decay rate in the image as a third information layer, and cooperating with the structural texture anomaly map and the eco-proxy texture anomaly map.

[0017] Preferably, the system further comprises a residual information interpretation module, which: performs spatial logical XOR operation on the structural texture anomaly map and the eco-proxy texture anomaly map to extract non-coupled abnormal areas that only appear in a single channel, and according to the source channel of the non-coupled abnormal area, marks the area from the structural texture channel processing module as a potential deep structure area, and marks the area from the eco-proxy texture channel processing module as a shallow environmental anomaly area, to generate auxiliary decision-making information.

[0018] Preferably, the time-series baseline library construction unit obtains all historical images of the same month as the current image within the past ten years of the survey area, and the eco-proxy texture channel processing module automatically determines the threshold value for segmenting the texture analysis result using the Otsu method or a statistical method based on the mean and standard deviation of historical samples, to preliminarily identify abnormal areas, and compares the preliminarily identified abnormal areas with the time-series statistical baseline.

[0019] Compared with the prior art, the present application has the following advantages:

[0020] 1. The present application provides a kind of mineral geological exploration information processing mode, it is by parallelly constructing a structure texture channel reflecting the physical landform of basement and an ecological proxy texture channel representing the state of surface ecosystem, then using collaborative verification engine to identify and determine the abnormal area coupled in space in two channels, this processing mode converts the information of surface covering material regarded as interference in traditional exploration into a independent verification dimension associated with the structure information in the cause of earth science, so that the system in target judgment, its basis is no longer single-dimensional texture anomaly signal, but the coupling signal that is simultaneously embodied in physical structure and ecological response two levels, thus, for the identification of real anomaly caused by deep geological reasons, it is no longer dependent on the improvement of single signal signal-to-noise ratio, but changes into the confirmation of a evidence chain with internal causal relationship, which provides a high-accuracy information screening path for early exploration using universal remote sensing data in wide coverage area.

[0021] 2, the present application further introduces a set of dynamic anomaly determination mechanism based on time series statistical baseline, which, when processing the ecological proxy texture channel, does not use a fixed global threshold, but automatically obtains the same period historical image of the exploration area in past years, establishes the normal texture fluctuation range of each plot in a particular season, and only when the texture state of the current image deviates from its own historical normal track, it is determined as an anomaly, this process enables the system to distinguish between periodic, seasonal surface changes and long-term persistent geological influences, so that before entering the dual-channel collaborative verification, those pseudo-anomaly signals caused by climate or regular human activities, which have no direct correlation with deep mineralization, are filtered out, ensuring the ecological proxy texture information sent to the final verification link.

[0022] 3, the present application also provides a working mode of information enhancement using microtopography shadow in the processing of structure texture channel, that is, in the area where the lithology is homogeneous and the spectral feature difference is weak, the system will preferentially select the low solar elevation angle image obtained at sunrise or sunset, and identify and outline the potential macro geological structure pattern by analyzing the shadow distribution elongated by weak topographic relief, which makes the extraction of structure information no longer completely limited to the internal spectral properties of surface material, but converts the external natural condition of sunlight into a detection method for amplifying microtopography, which enables the whole exploration system of the present application to obtain a structure texture input in areas where traditional optical remote sensing methods are poor in information, such as areas covered by large area single lithology or aeolian sediments, ensuring the operation of dual-channel collaborative verification mechanism. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 The present application is a system dual-channel information processing and collaborative verification technical flow chart.

[0024] Figure 2 Performance calibration curve diagram of the spatial coupling exponential weight parameter of the application;

[0025] Figure 3 Flow chart of the dynamic decision logic of the cooperative verification engine based on information entropy of the application;

[0026] Figure 4 Module cooperation timing diagram for the cooperative verification engine to execute double verification logic. DETAILED DESCRIPTION

[0027] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described in detail below with reference to the drawings. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0028] The mineral geological exploration system based on remote sensing image texture analysis provided by the present application comprises a time sequence baseline library construction unit, an ecological proxy texture channel processing module, a structure texture channel processing module and a cooperative verification engine module. The components work cooperatively to generate two independent texture anomaly information layers by parallel processing of remote sensing data, and identify the spatially coupled anomalies in the two information layers according to a preset cooperative verification logic to determine the exploration target area. In a specific application scenario, such as mineral exploration in a region widely covered by vegetation and red soil, the time sequence baseline library construction unit of the system is first started. The unit is configured to access the public remote sensing data archive and automatically obtain all historical images of the same month as the current image to be processed in the past ten years in the user-specified exploration area. The ground state of a specific area in a specific season follows a statistical periodicity, so that a quantitative reference baseline for the normal texture fluctuation range of the area can be established by obtaining long-time sequence contemporaneous images. For each obtained historical image, the ecological proxy texture channel processing module calculates the normalized difference vegetation index (NDVI) image thereof and calculates the variance or entropy of the NDVI image using the gray level co-occurrence matrix (GLCM) algorithm to represent the uniformity or complexity of vegetation coverage. The time sequence baseline library construction unit performs statistics on the texture parameters of all historical contemporaneous images, calculates and stores the mean and standard deviation of the texture parameters of each image unit, thereby generating a time sequence statistical baseline defining the normal fluctuation range.

[0029] After the time-series statistical baseline is established, the ecological proxy texture channel processing module processes the current remote sensing image, which calculates the NDVI map and its texture parameter map of the current image, and compares the current calculated texture parameter value with the normal fluctuation range of the corresponding geographical location and month in the time-series baseline library. The texture state of a region is only determined as an ecological proxy texture anomaly when it significantly deviates from its own historical normal track, for example, its texture value exceeds two standard deviations of the historical same period mean, and is recorded in an ecological proxy texture anomaly map. The processing method aims to filter out texture changes caused by regular seasonal changes or periodic human activities, and to improve the geological information purity of subsequent analysis. In parallel, the texture channel processing module is constructed to extract information reflecting the tectonic framework. The configuration of this module aims to avoid the interference of vegetation and soil moisture. Therefore, it automatically selects short-wave infrared (SWIR) band images that are not sensitive to vegetation, such as the 6th band image of Landsat 8 or 9, or the 11th band image of Sentinel 2. In such bands, the image brightness value mainly responds to the mineral composition of rocks and soil, reducing the interference of surface cover on the interpretation of underlying geological information. This module also uses the gray level co-occurrence matrix (GLCM) algorithm to calculate the contrast or heterogeneity texture parameters of the SWIR image. Such parameters are sensitive to linear or circular brightness mutations caused by fault structures, lithological contact zones, etc. The tectonic texture anomaly map generated after processing identifies potential regional-scale geological tectonic framework within the region.

[0030] When the survey area is determined to be a homogeneous lithology covered area, such as a large area of desert, the texture contrast in the SWIR band is weak at this time. The tectonic texture channel processing module is configured in a special working mode. In this mode, the system preferentially selects remote sensing images acquired at sunrise or sunset and with low solar elevation angles in the region. By using the phenomenon that weak topographic relief will cast elongated shadows under low-angle lighting, the system extracts shadow patterns with specific morphological distribution by performing edge detection algorithms on the image, and uses the extracted shadow patterns as the tectonic texture anomaly map. This method uses sunlight as a means of enhancing micro-topographic patterns, ensuring that the collaborative verification mechanism can obtain tectonic information input under different surface conditions. The ecological proxy texture anomaly map and the tectonic texture anomaly map are sent to the collaborative verification engine module together. Before performing spatial coupling determination, the module first calculates the image information entropy of the two anomaly maps to quantify the complexity of image information, and compares the calculated information entropy values with a reference interval obtained from multiple sample statistics. When the information entropy values of the two maps are within the reference interval, the engine performs strong verification logic to determine whether the anomaly region is spatially coupled through a coupling index C index where A​overlap A is the spatial overlap area of two abnormally connected clusters T d is the sum of the areas of two connected clusters min d is the minimum edge distance of two connected clusters max w1 and w2 are non-negative weight coefficients satisfying w1 + w2 = 1, when C index When the value is greater than the determination threshold, the pair of abnormal regions is determined as a high-confidence mineral exploration target area. If the information entropy value of any input graph is not located in the reference interval, indicating that the quality of the information source is reduced, the engine switches to weak verification logic, and the abnormal graph generated by the channel with a normal information entropy value is taken as the main target. Within a certain bandwidth range of the main target, it is searched whether there is an edge response in another channel, which is used as an auxiliary confirmation of the main target.

[0031] To correct the modulation effect of terrain on the surface ecological response signal, the system can also integrate a terrain context modulation module. The module calculates a terrain wetness index (TWI) map of the exploration area based on a digital elevation model (DEM), and modulates the limit values of the normal fluctuation range according to the TWI map when the time series statistical baseline is generated in the ecological proxy texture channel processing module. In areas with high terrain wetness index values, the limit values of the normal fluctuation range are expanded, and vice versa. This modulation allows the system to adjust the determination criteria according to the local terrain conditions to adapt to the differential expression of the signal in different terrain units. To evaluate the water-conducting and ore-conducting potential of structures, the system can be equipped with a structure permeability detection module. The module automatically obtains normalized difference water index (NDWI) images of the region at two different times when the soil is near saturation and when it is dry using a regional rainfall event. By subtracting the two NDWI images pixel by pixel, an image representing the spatial difference in soil moisture decay rate is generated. In this difference image, areas with abnormal moisture decay rates indicate underground areas with high permeability. This image is used as the third information layer and sent to the collaborative verification engine module for three-way collaborative verification with the structure texture anomaly map and the ecological proxy texture anomaly map to identify functionally active exploration target areas. In addition, the system also includes a residual information interpretation module. The module performs a spatial logical XOR operation on the structure texture anomaly map and the ecological proxy texture anomaly map to extract non-coupled abnormal areas that only appear in a single channel and classify them according to their source channels. Areas derived from the structure texture channel are marked as potential deep structure areas, and areas derived from the ecological proxy texture channel are marked as shallow environmental anomaly areas. The module finally generates an auxiliary decision information layer, which is presented to the user together with the core mineral exploration target area.

[0032] In a gold-bearing quartz vein survey project in Southeast Asia, the objective obstacle of the exploration work is that the surface is widely covered by vegetation and laterite layer, the overlying soil geochemical anomaly caused by deep structure is weak, which limits the application of interpretation method relying on the spectral characteristics of surface materials. If only the texture analysis of single-stage remote sensing image of the region is carried out, a large number of texture anomaly map spots will be generated due to seasonal vegetation growth, agricultural activities and other surface state changes, so that the subsequent field verification work is not feasible due to the large target base. In this application scenario, the system is deployed to perform target screening. The structural texture channel processing module of the system first acquires and processes the short-wave infrared image of the sixth band of Landsat 8 covering the exploration area, calculates the contrast texture parameters, and generates a structural texture anomaly map. The map identifies several northeast-trending linear structural belts and several ring structures, which are potential ore-hosting structural units under the regional geological background. At the same time, the ecological proxy texture channel processing module of the system processes the normalized vegetation index map acquired at the same time as the short-wave infrared image, and the time series baseline library construction unit automatically processes all historical images of the same month in the past ten years in the exploration area, establishing a time series statistical baseline for the normalized vegetation index texture of the region. When the ecological proxy texture channel processing module compares the texture analysis results of the current image with the time series statistical baseline, a large number of texture change areas caused by seasonal factors are filtered out, and only a few areas that are spatially in block form and deviate from their historical normal fluctuation range in the time dimension are retained in the final generated ecological proxy texture anomaly map. The ecological proxy texture anomaly map after time series purification provides a screened input for subsequent collaborative verification.

[0033] This processing path does not try to remove the surface cover to explore the underlying geological information, but regards the spatial texture characteristics of the surface cover itself as a proxy signal. By introducing the constraint of the time dimension, it distinguishes between the periodic changes of the surface and the persistent anomalies that may be caused by the long-term influence of the deep geological environment. Then, the core of the exploration problem changes from how to eliminate surface interference to how to verify whether these persistent surface ecological anomalies have a spatial coupling relationship with deep geological structures. Finally, the collaborative verification engine module receives the structural texture anomaly map and the ecological proxy texture anomaly map, confirms that the information entropy values of the two maps are within the reference interval, and executes the strong verification logic. The analysis result shows that among the several northeast-trending linear structural belts, only the specific sections of two structural belts show overlapping or adjacent relationship with the abnormal blocks in the ecological proxy texture anomaly map in space, and their coupling index C indexThe value greater than the preset determination threshold, other texture anomalies and ecological proxy texture anomalies, are all in a spatially unrelated isolated state; the final output of the mineral exploration target area of the system only contains the spatial coupling area that is simultaneously embodied in the two dimensions of physical structure and ecological response, the number of exploration target areas is reduced from thousands to single digits, and subsequent field geological survey work can be focused on these high-potential target areas, and alteration phenomena related to gold mineralization are found therein. This provides a high-precision information screening path for early exploration of a wide coverage area under the condition of relying only on universal remote sensing data, and improves the efficiency of subsequent exploration work.

[0034] Example 2: To quantitatively evaluate the effectiveness of the system of the present application in screening exploration targets in a complex coverage area, the following comparative test was carried out. The test selected a 100 km 2 area in the survey area, which has a ground truth map indicating 8 known mineralization points established based on past detailed geological investigation and engineering verification data, which is used as a benchmark for evaluating test results. The test sets up one test group and three control groups, all using the same scene Landsat 8 image covering the test area and its historical archives as data sources. The test group uses the complete technical solution of the present application, the control group A only uses the output of the structure texture channel processing module, the control group B only uses the output of the ecological proxy texture channel processing module purified by the time series baseline library, and the control group C uses the alteration mineral extraction method based on spectral angle mapping. For the test group, the determination threshold of the coupling index C index in the verification engine module is set to 0.7. This value is determined by balancing the recall rate and accuracy rate in the early stage of exploration through parameter testing in other known areas. The value is an engineering setting that can better balance the two. After parallel processing of each group, the control group C fails to identify any effective target due to spectral interference of vegetation and laterite. The control group A outputs 25 structure texture anomalies, of which 5 are true targets and 20 are false targets, and the identification accuracy is 20.0%. The control group B outputs 12 ecological proxy texture anomalies purified by time series, of which 4 are true targets and 8 are false targets, and the identification accuracy is 33.3%. The test group verifies the structure texture anomaly map and the ecological proxy texture anomaly map, and finally outputs 7 mineral exploration target areas, of which 6 are true targets and 1 is a false target, and the identification accuracy is 85.7%.

[0035] The test data shows that although the control groups A and B can identify some real targets, they cannot exclude a large number of anomalies caused by non-mineralized geological structures or non-structural environmental factors due to the reliance on only a single dimension of information, resulting in a large number of false targets and low accuracy. The dual-channel collaborative verification method used in the test group spatially couples the structural information and the chronologically purified ecological anomaly information, and only the anomalies that respond in both independent information dimensions are confirmed as final targets. This cross-validation mechanism eliminates the ambiguity of single-source information, intercepts most real targets while controlling the number of false targets at a low level. The results of this test show that compared with the exploration method relying on a single information source, the collaborative verification method used in the present application can improve the identification accuracy of concealed mineralization targets in vegetation-covered areas and provides a dual-channel collaborative verification technical path for remote sensing geological exploration.

[0036] Embodiment 3: This embodiment combines Figures 1 to 4 to illustrate a mineral geological exploration system based on remote sensing image texture analysis, as shown in Figure 1 , the data input end includes a plurality of historical synchronous remote sensing images for establishing a historical reference benchmark, a digital elevation model DEM for correcting the influence of topography, a current remote sensing image for generating ecological proxy information, and a short-wave infrared band image for extracting structural information. The time series baseline library construction unit processes the historical images to calculate and output the time series statistical baseline, and the topographic context modulation module generates a topographic humidity index based on the DEM to spatially modulate the baseline. The modulated baseline is sent to the ecological proxy texture channel processing module, which simultaneously processes the normalized vegetation index NDVI of the current image to purify and identify areas deviating from the historical trajectory, generates an ecological proxy texture anomaly map, and in parallel, the structural texture channel processing module analyzes the short-wave infrared SWIR band image that is not sensitive to vegetation to extract the basement structural pattern information and generate a structural texture anomaly map. These two independent anomaly maps are simultaneously sent to the collaborative verification engine module, which evaluates the quality of the two-channel information, executes strong or weak verification logic to determine spatial coupling, and finally delineates a high-confidence mineral exploration target area. At the same time, a residual information interpretation module can extract non-coupled anomaly areas from the two anomaly maps to generate auxiliary decision-making information containing potential deep structural areas and shallow environmental anomaly areas.

[0037] As Figure 2As shown in the figure, the horizontal axis represents the value of the spatial overlap weight w1, and the vertical axis represents the index value. The figure shows the trend of three curves with the value of w1, namely the average coupling index of positive samples, the average coupling index of negative samples, and the comprehensive evaluation index F1 score. It can be seen from the figure that when the value of w1 is around 0.65, the average coupling index of positive samples and the F1 score both reach their peak, while the average coupling index of negative samples is at a relatively low level. This indicates that the weight configuration can most effectively identify real targets and suppress false targets.

[0038] like Figure 3 As shown, after receiving dual-channel anomaly maps in standby / waiting-for-input mode, the engine first enters the information quality evaluation stage, calculating the information entropy values ​​of the two anomaly maps. Then, it enters a core decision node. If it is determined that the information entropy is within the baseline range, the process switches to executing strong verification logic, calculating the coupling index by judging spatial overlap and proximity. If it is determined that either information entropy is not within the baseline range, it indicates that the quality of at least one information source has deteriorated, and the process switches to executing weak verification logic. At this time, the high-quality channel is the main target, and only the edge response of the other channel is searched. Regardless of the logic executed, the results are eventually incorporated into the target generation stage to delineate the target area or confirm the main target. After the judgment is completed, the engine outputs the results and resets, returning to the standby state.

[0039] like Figure 4 As shown, the process begins with the collaborative verification engine acquiring data from the ecological anomaly map and the constructed anomaly map respectively. Then, it calculates the information entropy of the two maps and returns the entropy value. In an alt selection combo box, two different paths are displayed based on the entropy value judgment result: If the entropy values ​​of both maps are within the baseline range, strong verification logic is triggered. The engine requests the analyzer to calculate the spatial overlap and spatial proximity, and the analyzer returns the generated coupling index. If either entropy value is abnormal, weak verification logic is triggered. After selecting the main target, the engine performs edge response search and returns the verification result. Finally, the engine determines the exploration target area based on the result returned by either path and calls the output module to generate a result report.

[0040] Example 4: When applying the system of the present invention to a province with geological and ecological environmental characteristics different from those of previous areas, in order to adapt the decision logic of the system's core algorithm module to the specific conditions of the region, a systematic offline parameter calibration procedure needs to be performed. This procedure aims to establish a statistical deviation threshold for determining ecological anomalies, a benchmark interval for image information entropy to assess the quality of input information, and a coupling index C for defining spatial coupling strength. index The calibration procedure first selects several areas with a total area of ​​approximately 100 km² within the province to be explored, based on existing regional geological data, using key parameters such as weighting coefficients. 2The training area needs to meet two conditions: first, the geology and surface cover type can represent the general characteristics of the whole province to be explored; second, the area contains known positive samples such as small ore spots or fault fracture zones verified by previous exploration work and known negative samples such as large areas or simple structures without mineralization records.

[0041] The first step of the procedure is to establish the image information entropy reference interval applicable to the province. The system automatically acquires and processes all available images covering all training areas in the past 15 years, runs the texture channel processing module and the ecological proxy texture channel processing module, generates a sample library containing hundreds of structure texture anomaly maps and ecological proxy texture anomaly maps, and calculates the image information entropy of each map in the sample library. The range between the 10th percentile and the 90th percentile of the statistical distribution is determined as the information entropy reference interval applicable to the exploration province. The second step of the procedure is to calibrate the ecological proxy texture anomaly threshold. The system uses known positive samples and known negative samples in the training area as a reference to iteratively test the statistical deviation threshold in the ecological proxy texture channel processing module. The test process traverses the standard deviation multiple from 1.0 to 3.0 with a step size of 0.1. At each step, the system calculates the recognition results of positive and negative samples in the training area and evaluates the overall classification performance. Finally, the standard deviation multiple that maximizes the sum of true positive rate and true negative rate is determined as the optimal threshold. The calibration result in this particular province is 2.1 times the standard deviation, which is set as the business parameter for the system to determine ecological proxy texture anomalies in this area. The third step of the procedure is to optimize the weight coefficients w1 and w2 of the coupling index C index . These two coefficients are used to balance the relative importance of spatial overlap and spatial proximity in determining the coupling relationship. The system uses the calibrated anomaly map from the previous step to traverse the value of w1 from 0 to 1 with a step size of 0.05 in the training area, and sets w2 = 1-w1. Under each weight combination, the system calculates the average value of C index for all known positive sample areas. The weight coefficient combination that maximizes the average value of C index for positive sample areas is w1 = 0.65 and w2 = 0.35, which is determined as the optimal configuration and applied to subsequent processing.

[0042] By performing the above procedure, the system completes the targeted localization calibration of all key algorithm parameters before conducting large-scale exploration operations. Each link of the system's information processing and decision logic has a setting basis based on objective data verification in the region. The system then enters a standby state, ready to perform mineral exploration target area screening operations in the region.

[0043] Example 5: When the system of the present application is applied to an area with a single composition of surface material, such as a gobi region covered by a large area of wind-blown sand, the application of short-wave infrared band texture analysis relying on lithological differences will be limited; in such working conditions, before executing the main process, the system first performs a global variance calculation on the standard short-wave infrared band image covering the survey area, and if the calculated variance value is lower than a low variance threshold value obtained by statistical calibration of multiple geomorphic samples, the system determines that the area is a homogeneous lithology covered area, and adjusts the working mode of the structural texture channel processing module; in this mode, the system no longer performs the gray level co-occurrence matrix algorithm on the original image, but automatically accesses the satellite data archives, filters and calls the remote sensing image acquired at sunrise or sunset with a solar elevation angle of less than 15 degrees in the area, and performs edge detection algorithm on the image, and generates a structural texture anomaly map by extracting the shadow distribution formed by weak topographic relief.

[0044] To evaluate the structural permeability of the target area, the structural permeability detection module is activated to analyze the hydrological dynamic response of the area after a rainfall event, which accesses the meteorological data service to monitor the survey area, and is triggered when a rainfall event with rainfall exceeding 20 mm and covering more than 80% of the area within 24 hours is detected, the system immediately retrieves and acquires the first cloud-free remote sensing image sensitive to soil moisture after the rainfall ends as the saturated phase image, at the same time, the system accesses the geoscience database to obtain the dominant soil type of the target area, and calculates an optimal soil moisture decay rate difference observation time window according to the built-in soil moisture evaporation model combined with the average weather of the current season, such as for sandy loam in Central Asia, the window is determined to be 72 to 96 hours after the rainfall stops, the system further acquires the second cloud-free remote sensing image within this dynamically calculated time window as the decay phase image, and generates an information layer representing the structural permeability by processing the two images, which is used for subsequent collaborative verification.

[0045] Example 6: Before being deployed in a new survey project area, the system of the present application needs to perform a set of standardized localization adaptation procedures to ensure that its internal key parameters match the geoscience characteristics and data environment of the area, the first step of the procedure is to set the geometric parameters in the collaborative verification engine module, the procedure requires inputting the average influence radius of the typical geochemical halo or alteration zone related to the target mineral species and effective in the area, the system will set the maximum search distance d max and the bandwidth range in the weak verification logic to 1.5 times the radius value, which serves to associate the search scale of spatial coupling analysis with the objective geological influence range of mineralization in the area.

[0046] The second step of the procedure is to adaptively configure the algorithm for preliminary identification of abnormal areas in the ecological proxy texture channel processing module. When processing the normalized vegetation index texture map, the system first analyzes the gray histogram of the image. If the histogram presents a bimodal shape, indicating that there are two distinguishable populations of background and potential anomalies in the image, the system selects the Otsu method for threshold segmentation. If the histogram presents a unimodal or complex multi-peak shape, the system automatically switches to a data-driven algorithm selection mechanism based on the mean and standard deviation statistics of historical samples. This mechanism aims to select the most suitable processing path based on the statistical distribution characteristics of the input data. The third step of the procedure is to perform self-checking and fault-tolerant processing of historical data environment. The time series baseline library construction unit counts the number of available historical images within the target month and year range. If the number of images is less than 10, the system determines that the area is a data sparse region. In this working condition, the system automatically relaxes the data retrieval conditions to adjacent months and performs weighted averaging of images from different months to construct the time series statistical baseline, with the target month image weight set to the highest. After the completion of this procedure, all core parameters of the system are locally configured, and the system enters an optimized running state for the specific project area.

[0047] In Example 7, when exploring a subalpine area with complex terrain and recent local forest fires, some functional modules of the system face the challenge of asymmetric decline in data source quality. Specifically, large areas of burned land covered with homogeneous carbon vegetation make it difficult for the ecological proxy texture channel processing module to process the images in this area. The output of the ecological proxy texture anomaly map contains a large area of low information, and the image information entropy is lower than the lower limit of the preset baseline interval.

[0048] In this case, the collaborative verification engine module automatically switches to weak verification logic, and the information entropy normal texture anomaly map is used as the main basis for identification. At the same time, the terrain context modulation module modulates the abnormal judgment logic of the ecological proxy texture channel according to the significant elevation of the digital elevation model in this area. This modulation enables the system to identify several weak but persistent ecological proxy texture anomalies caused by groundwater seepage along structural fissures in steep and shady areas with high terrain humidity index with higher sensitivity. When executing weak verification logic, the system takes several north-west structural belts marked in the structural texture anomaly map as the main target, and successfully searches for the weak ecological response identified after terrain modulation within the specified bandwidth of these structural belts as confirmation evidence, thereby delineating the core exploration target area and completing the exploration task. After the exploration task is completed, the residual information interpretation module of the system processes the uncoupled anomalies in the two channels. A linear structural texture anomaly covered completely by the burned trace is marked as a potential deep structure area because it lacks any ecological response signal and cannot enter the collaborative verification, thus providing independent auxiliary decision-making information for subsequent deep physical exploration work.

[0049] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0050] Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application.

Claims

1. A mineral geological exploration system based on remote sensing image texture analysis, characterized in that, The system comprises: a time-series baseline library construction unit, which acquires a plurality of historical co-temporal remote sensing images of the survey area and calculates statistical parameters of normalized vegetation index texture of the historical co-temporal remote sensing images to generate a time-series statistical baseline defining a normal fluctuation range; an ecological proxy texture channel processing module, which calculates a normalized vegetation index map of a current remote sensing image, performs texture analysis on the normalized vegetation index map, and compares the texture analysis result with the time-series statistical baseline to generate an ecological proxy texture anomaly map containing only regions with texture values beyond the normal fluctuation range; a tectonic texture channel processing module, which acquires a short-wave infrared band image insensitive to vegetation and performs texture analysis on the image to generate a tectonic texture anomaly map; a collaborative verification engine module, which receives the ecological proxy texture anomaly map and the tectonic texture anomaly map, calculates image information entropy values of the two maps, and compares the information entropy values with a reference interval obtained according to multi-sample statistics, and when the information entropy values of the two maps are both within the reference interval, performs strong verification logic to determine abnormal regions spatially coupled in the two maps as a mineral exploration target area.

2. The mineral geological exploration system based on remote sensing image texture analysis according to claim 1, characterized in that, When the collaborative verification engine module executes strong verification logic, it uses a coupling index C determined by both spatial overlap and spatial proximity. index To determine whether anomaly regions are spatially coupled, the coupling index C is used. index The calculation rules are as follows: Among them, A overlap To construct the spatial overlap area between the anomalous connected clusters in the texture anomaly map and the anomalous connected clusters in the ecological proxy texture anomaly map, A T d is the sum of the areas of two connected clusters. min Let d be the minimum edge distance between two connected clusters. max To find the maximum search distance, w1 and w2 are non-negative weight coefficients that satisfy w1 + w2 = 1. The collaborative verification engine module will use the coupling index C. index Regions exceeding the judgment threshold are identified as spatially coupled abnormal regions.

3. The mineral geological exploration system based on remote sensing image texture analysis according to claim 1, characterized in that, The ecological proxy texture channel processing module and the tectonic texture channel processing module both perform texture analysis using a gray level co-occurrence matrix algorithm, wherein the tectonic texture channel processing module calculates at least one of contrast or heterogeneity as a texture parameter of the short-wave infrared band image, and the ecological proxy texture channel processing module calculates at least one of variance or entropy as a texture parameter of the normalized vegetation index map.

4. The mineral geological exploration system based on remote sensing image texture analysis according to claim 1, characterized in that, The collaborative verification engine module performs weak verification logic when either of the information entropy values of the two maps is not within the reference interval, which takes abnormal regions in the anomaly map generated by a channel with an information entropy value within the reference interval as a main target and searches for edge responses of another channel within a bandwidth range of the main target.

5. The mineral geological exploration system based on remote sensing image texture analysis according to claim 1, characterized in that, The tectonic texture channel processing module selects a remote sensing image with a low solar elevation angle acquired at sunrise or sunset when the survey area is determined to be a homogeneous lithology coverage area, and generates the tectonic texture anomaly map by performing edge detection on the remote sensing image to extract shadow distribution formed by weak topographic relief.

6. The mineral geological exploration system based on remote sensing image texture analysis according to claim 1, characterized in that, The system further comprises a terrain context modulation module, which: calculates a terrain humidity index map of the survey area based on a digital elevation model; and the ecological proxy texture channel processing module spatially modulates the normal fluctuation range according to the terrain humidity index map when generating the time-series statistical baseline, expands the limit value of the normal fluctuation range in regions with high terrain humidity index values, and narrows the limit value of the normal fluctuation range in regions with low terrain humidity index values. 7.The mineral geological exploration system based on remote sensing image texture analysis of claim 1, wherein, The system further comprises a tectonic permeability detection module which: after a rainfall event in the survey area, automatically obtains normalized difference water index images of the area at two different times when the soil is close to saturation and when it is dry, generates an image representing the spatial difference of soil moisture attenuation rate by pixel-by-pixel subtraction of the two normalized difference water index images, and, in conjunction with the verification engine module, takes the area with abnormal moisture attenuation rate in the image as the third information layer, and verifies it in conjunction with the tectonic texture anomaly map and the ecological proxy texture anomaly map to determine the mineral exploration target area. 8.The mineral geological exploration system based on remote sensing image texture analysis of claim 1, wherein, The system further comprises a residual information interpretation module which: performs spatial logical XOR operation on the tectonic texture anomaly map and the ecological proxy texture anomaly map to extract non-coupled abnormal areas appearing only in a single channel, and according to the source channel of the non-coupled abnormal area, marks the area from the tectonic texture channel processing module as a potential deep tectonic area, and marks the area from the ecological proxy texture channel processing module as a shallow environmental anomaly area, to generate auxiliary decision-making information. 9.The mineral geological exploration system based on remote sensing image texture analysis of claim 1, wherein, The time series baseline library construction unit obtains all historical images of the same month as the current image within the past ten years in the survey area, and the ecological proxy texture channel processing module uses the Otsu method or a statistical method based on the mean and standard deviation of historical samples to automatically determine the threshold for segmenting the texture analysis results to preliminarily identify abnormal areas, and compares the preliminarily identified abnormal areas with the time series statistical baseline.

Images