A method and apparatus for interpreting oil and gas geological maps

Abstract

This invention provides a method and apparatus for analyzing oil and gas geological maps, belonging to the field of oil and gas exploration and development technology. The method includes: acquiring oil and gas geological map data; extracting features and their spatial location information from the oil and gas geological map data, wherein the features are in vector form and include text features and visual features; calculating the association weights between the features and their spatial location information, wherein the association weights between the text features and the visual features are calculated based on a cross-modal attention mechanism; constructing an association graph between the features based on the association weights; processing the association graph based on a graph neural network to obtain an updated association graph that incorporates global dependencies; and outputting the updated association graph to obtain a first analysis result of the oil and gas geological map. This method improves the comprehensive processing capability of graphic and textual information, thereby better meeting the accuracy requirements in practical applications.

Description

Technical Field

[0001] This invention relates to the field of oil and gas exploration and development technology, and specifically to a method and equipment for analyzing oil and gas geological maps. Background Technology

[0002] Oil and gas geological maps are among the most important data carriers in the oil and gas exploration and development process, containing rich geological information such as structural features, stratigraphic distribution, oil and gas reservoir locations, and fault systems. Accurate interpretation of oil and gas geological maps is of great significance for key decisions such as oil and gas resource assessment, exploration deployment, and development plan formulation.

[0003] Traditional oil and gas geological map interpretation relies primarily on manual reading, supplemented by simple computer-aided tools. Industry reality shows that oil and gas companies generate millions of geological maps annually, the vast majority of which require manual interpretation. This traditional manual interpretation method has the following significant limitations: slow speed and low efficiency; susceptibility to subjective factors, resulting in limited accuracy; high professional knowledge threshold, limiting the use of geological data by non-professionals; incomplete information extraction, failing to fully extract the complex geological relationships hidden within the map, only obtaining surface information; and insufficient knowledge transfer, making it difficult to effectively pass on and reuse the experience and knowledge of experts.

[0004] In recent years, artificial intelligence (AI) technology has made groundbreaking progress in fields such as computer vision and natural language processing, providing new technical approaches for the automatic interpretation of geological maps. Currently, there are some research and practices attempting to apply AI technology to geological map interpretation. For example, some studies use traditional computer vision methods to identify simple elements in geological maps; others try to extract textual information from the maps using OCR technology. However, these methods often only process one aspect of the geological map's information, lacking a comprehensive understanding of both graphic and textual information, and thus failing to meet the accuracy requirements of practical applications. Summary of the Invention

[0005] The purpose of this invention is to provide a method and equipment for analyzing oil and gas geological maps, which improves the comprehensive processing capability of graphic and textual information, thereby better meeting the accuracy requirements in practical applications.

[0006] To achieve the above objectives, in a first aspect, embodiments of the present invention provide a method for analyzing oil and gas geological maps. The method includes: acquiring oil and gas geological map data; extracting features and their spatial location information from the oil and gas geological map data, wherein the features are in vector form and include text features and visual features; calculating the association weights between the features and their spatial location information, wherein the association weights between the text features and the visual features are calculated based on a cross-modal attention mechanism; constructing an association graph between the features based on the association weights; processing the association graph based on a graph neural network to obtain an updated association graph that integrates global dependencies; and outputting the updated association graph to obtain a first analysis result of the oil and gas geological map.

[0007] Optionally, extracting text features from the oil and gas geological map data includes: performing text region detection on the oil and gas geological map data based on the EAST algorithm to obtain text regions; extracting text information from the text regions based on an OCR model; and extracting text features from the text information based on a pre-trained language model, including technical terms, parameter values, and / or sentence semantics.

[0008] Optionally, extracting visual features from the oil and gas geological map data based on a pre-trained convolutional neural network includes: identifying primitives from the oil and gas geological map data based on the YOLOv8 model, the primitives including well locations, faults, and contour lines; identifying lines from the oil and gas geological map data and classifying the lines based on Hough transform and support vector machine classification, the line classification including structural lines and fault lines; identifying different color blocks from the oil and gas geological map data and labeling their lithology and reservoir type based on the K-means clustering algorithm; and locating legend regions in the oil and gas geological map data based on object detection and text association, and obtaining the correspondence between the symbols and meanings of the legend regions.

[0009] Optionally, the method for extracting the spatial location information includes: for text features, using the position coordinates of the text region corresponding to the text feature as spatial location information; and for visual features, using the position of the visual element corresponding to the visual feature as spatial location information, wherein the visual element includes the primitive, line, color block and legend.

[0010] Optionally, the method for calculating the association weight between the text features and the visual features includes: unifying the dimensions of the text features, the visual features, and the spatial location information; calculating the spatial location attention weight between the text features and the visual features based on the unified spatial location information; calculating the similarity score between the unified text features and the visual features based on a pre-trained attention module, normalizing the similarity score, and using it as the cross-modal semantic attention weight between the two; and weightedly fusing the spatial location attention weight and the cross-modal semantic attention weight based on a preset fusion coefficient to obtain the association weight, wherein the attention module is trained based on a training set composed of text feature-visual feature data pairs.

[0011] Optionally, the formula for calculating the spatial location attention weight is:

[0012] Where W is the spatial location attention weight, d is the relative distance between the text feature and the visual feature, and α is an adjustable attenuation parameter.

[0013] Optionally, after calculating the similarity score between the text features and the visual features after unifying the dimension, the method further includes: determining the first geological entity type of the text features and the second geological entity type of the visual features; determining whether the first geological entity type and the second geological entity type can be matched based on a pre-constructed geological entity type mask; and if they cannot be matched, assigning the similarity score to zero.

[0014] Optionally, constructing the association graph of the visual features and text features based on the association weights includes: treating the visual features and text features as visual nodes and text nodes, respectively; traversing the association weights of all visual nodes and text nodes, and performing the following operations on any visual node and any text node to obtain the association graph: determining whether the association weight between the visual node and the text node is greater than a preset weight threshold; if it is greater, then taking the visual node and the text node as a pair of associated nodes in the association graph, and establishing an edge between the pair of associated nodes.

[0015] Optionally, the method further includes: obtaining all the visual features and text features from the first parsing result as geological entities; extracting the relationships between the geological entities and the attributes of the geological entities from the text annotations of the oil and gas geological map data; updating the knowledge graph based on the geological entities, the relationships between the geological entities, and the attributes of the geological entities; and outputting the knowledge graph to obtain the second parsing result of the oil and gas geological map.

[0016] Optionally, the method further includes: after extracting features and their spatial location information from the oil and gas geological map data, determining the magnitude relationship between the number of features and a first quantity threshold; if the number of features is less than the first quantity threshold, determining that the oil and gas geological map data is fuzzy data; discarding the features and their spatial location information extracted from the fuzzy data; and preprocessing the fuzzy data, and re-executing the oil and gas geological map parsing method based on the preprocessed fuzzy data, where the preprocessing includes image denoising, skew correction, contrast enhancement, resolution adjustment, and color space conversion.

[0017] On the other hand, the present invention provides a device for parsing an oil and gas geological map, the device includes a memory and a processor, the processor is used to run a program, where the program is used to execute the method when it runs.

[0018] Through the above technical solutions, it is possible to quickly and automatically parse an oil and gas geological map, which not only liberates human resources, but also greatly improves work efficiency and parsing accuracy, reduces the professional threshold, and avoids various problems existing in manual parsing. For the correlation weight between text features and visual features, the present invention improves the comprehensive processing ability of various information types based on a cross-modal attention mechanism. Then, based on the correlation weight, a correlation graph between features is constructed and the global dependency relationship is fused, which can effectively integrate global information, realize the global联动 understanding of multiple elements, and avoid information deviation caused by isolated parsing. This method simultaneously processes visual information and text information in the geological map and realizes a multi-modal automatic parsing technology of deep fusion understanding, which is of great significance for improving the efficiency of oil and gas exploration and development, reducing costs, and improving decision-making quality.

[0019] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent specific implementation part. Description of the Drawings

[0020] The drawings are used to provide a further understanding of the embodiments of the present invention, and constitute a part of the specification, and are used to explain the embodiments of the present invention together with the following specific implementation manners, but do not constitute a limitation to the embodiments of the present invention. In the drawings: Figure 1 is a schematic flowchart of the oil and gas geological map parsing method provided by the embodiment of the present invention; Figure 2 is a schematic structural diagram of the device provided by the embodiment of the present invention.

[0021] Description of the Reference Numerals 101 Processor 102 Memory 103 Bus 10 Device Specific Embodiments The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0022] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with relevant laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0023] Figure 1 This is a schematic flowchart of the oil and gas geological map analysis method provided in an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes steps S1 to S6.

[0024] Step S1: Obtain oil and gas geological map data.

[0025] This invention supports geological maps in various formats, including but not limited to: image formats such as JPEG, PNG and TIFF; vector formats such as SVG, DXF and DWG; professional formats such as those generated by professional software such as GeoMap and Petrel; and PDF formats such as scanned or exported PDF geological maps.

[0026] For oil and gas geological map data in different formats, after acquiring the oil and gas geological map data, the format of the oil and gas geological map data is determined based on the file extension and file header information of the oil and gas geological map data; based on the format of the oil and gas geological map data, the corresponding parser is called to parse the oil and gas geological map data, such as a raster parser, vector parser, or PDF parser; and the output of the parser is converted into a unified, machine learning model-processable preset format multimodal document for subsequent processing.

[0027] Step S2: Extract features and their spatial location information from the oil and gas geological map data. The features are in vector form and include text features and visual features.

[0028] Step S3: Based on the features and their spatial location information, calculate the association weights between each feature, wherein the association weights between the text features and the visual features are calculated based on a cross-modal attention mechanism.

[0029] Step S4: Construct an association graph between the features based on the association weights.

[0030] Step S5: Process the association graph based on the graph neural network to obtain the updated association graph that incorporates global dependencies.

[0031] Step S6: Output the updated correlation map to obtain the first analytical result of the oil and gas geological map.

[0032] The method proposed in this invention enables rapid and automated analysis of oil and gas geological maps, freeing up human resources, significantly improving work efficiency and accuracy, lowering the professional threshold, and avoiding various problems associated with manual analysis. For the association weights between textual and visual features, this invention specifically designs a cross-modal attention mechanism, which effectively captures the associations between visual and textual information. For example, it automatically identifies the association between "reservoir" text and corresponding color blocks, improving the comprehensive processing capability for multiple information types. After obtaining the associated features, an association graph is constructed based on the association weights, transforming scattered elements in the geological map into a structured association graph. This enables effective integration of global information and establishes cross-modal associations between different information types. Finally, processing is performed using a graph neural network, which captures global dependencies. For example, the graph structure reveals the association between different well locations and the same reservoir, as well as the common characteristics of reservoir attributes, resulting in an updated association graph that integrates global dependencies. This achieves a global, interconnected understanding of multiple elements, avoiding information bias caused by isolated analysis. This method simultaneously processes visual and textual information in geological maps and achieves multimodal automatic parsing technology for deep fusion understanding, which is of great significance for improving the efficiency of oil and gas exploration and development, reducing costs, and enhancing decision-making quality.

[0033] Furthermore, extracting text features from the oil and gas geological map data includes: performing text region detection on the oil and gas geological map data based on the EAST algorithm to obtain text regions in different locations and directions, such as different types of text regions like "reservoir parameter table," "legend description," and "fault annotation" in the location map; extracting text information from the text regions based on the OCR model, such as recognizing technical terms like "interbedded sandstone and mudstone" and "permeability 50mD"; and extracting text features from the text information based on a pre-trained language model, including technical terms, parameter values, and / or sentence semantics.

[0034] The text information includes annotations, legends, coordinate data, and parameter descriptions, such as "Fault F1", "Porosity 15-20%", "X coordinate 118°E", and "Jurassic strata".

[0035] Furthermore, the extraction of visual features from the oil and gas geological map data based on a pre-trained convolutional neural network includes: identifying primitives from the oil and gas geological map data based on the YOLOv8 model, the primitives including well locations, faults, and contour lines; identifying lines from the oil and gas geological map data and classifying the lines based on Hough transform and support vector machine classification, the line classification including structural lines and fault lines; identifying different color blocks from the oil and gas geological map data and labeling their lithology and reservoir type based on the K-means clustering algorithm; and locating legend regions in the oil and gas geological map data based on object detection and text association, and obtaining the correspondence between the symbols and meanings of the legend regions.

[0036] Furthermore, the method for extracting spatial location information includes: for text features, using the position coordinates of the text region corresponding to the text feature as spatial location information; and for visual features, using the position of the visual element corresponding to the visual feature as spatial location information, wherein the visual element includes the primitive, line, color block and legend.

[0037] For any feature, its position coordinates can be derived based on a preset location. Specifically, for text features, a detection box is obtained during text region detection, and the center of the detection box or one of its vertices is used as the position coordinates. For visual features, a similar approach can be taken as for text features, using a specific location of the detection box, or other existing localization methods to determine the position coordinates.

[0038] Because different geological maps have different scales, this invention performs coordinate normalization on the oil and gas geological maps before extracting spatial location information. The normalized coordinate axes are obtained by using the diagonal length of the oil and gas geological map as the denominator and the length or width of the map as the numerator. In this coordinate axis, the distance between any two locations is called the relative distance, with a value range of [0,1]. This ensures that the spatial location information is not affected by the original image resolution, improving the robustness of the analysis results.

[0039] Furthermore, the method for calculating the association weight between the text features and the visual features includes: unifying the dimensions of the text features, the visual features, and the spatial location information; calculating the spatial location attention weight between the text features and the visual features based on the unified spatial location information; calculating the similarity score between the unified text features and the visual features based on a pre-trained attention module, normalizing the similarity score, and using it as the cross-modal semantic attention weight between the two; and weightedly fusing the spatial location attention weight with the cross-modal semantic attention weight based on a preset fusion coefficient to obtain the association weight.

[0040] The attention module is trained on a training set consisting of text feature-visual feature data pairs. When constructing the training set, a large number of text features and visual features are first acquired, then labeled, and text feature-visual feature data pairs are established. For example, the text feature of "Fault F1" and the visual feature of "Fault Line" are considered as one text feature-visual feature data pair; similarly, the text feature of "Sandstone Reservoir" and the visual feature of "Red Color Block" are considered as another text feature-visual feature data pair, thereby establishing the association between text features and visual features.

[0041] Furthermore, the formula for calculating the spatial location attention weight is as follows:

[0042] Where W is the spatial location attention weight, and the calculated value ranges from (0,1); d is the relative distance between the text feature and the visual feature, and the value ranges from [0,1], where 0 represents complete overlap and 1 represents the farthest boundary; α is an adjustable decay parameter used to control the rate at which the spatial location attention weight decays with distance, and can be set based on experience.

[0043] Calculating spatial location attention weights using this formula can improve the correlation weight scores between near-field text features and visual features, thereby increasing the accuracy of subsequent analysis results.

[0044] Furthermore, after calculating the similarity score between the text features and the visual features after unifying the dimension, the method further includes: determining the first geological entity type of the text features and the second geological entity type of the visual features; determining whether the first geological entity type and the second geological entity type can be matched based on a pre-constructed geological entity type mask; and if they cannot be matched, assigning the similarity score to zero.

[0045] In practical implementation, entity types are first labeled with textual or visual features. Entity types can be set according to actual conditions, such as faults, folds, traps, or other types. A classification model is then trained based on the labeled textual or visual features. For any two entity types, the correspondence rules between them are determined based on geological expertise, and a geological entity type mask is assigned to either 0 or 1. For two entity types with a geological entity type mask of 0, they are considered unmatched, and their similarity score is set to zero; for two entity types with a geological entity type mask of 1, they are considered matched, and their similarity score is not modified. This method avoids meaningless matching in general algorithms, adapts to the professional needs of geological map analysis, and involves entirely numerical calculations without complex logic. It can be directly implemented using PyTorch, and the mask matrix can be flexibly adjusted through configuration files to adapt to different geological entity types.

[0046] Furthermore, constructing the association graph of the visual features and text features based on the association weights includes: treating the visual features and text features as visual nodes and text nodes, respectively; traversing the association weights of all visual nodes and text nodes, and performing the following operations on any visual node and any text node to obtain the association graph: determining whether the association weight between the visual node and the text node is greater than a preset weight threshold; if it is greater, then taking the visual node and the text node as a pair of associated nodes in the association graph, and establishing an edge between the pair of associated nodes.

[0047] As an example, the weight threshold can be set to 0.2. For the text feature "F1 fault" and the visual feature "fault line", it was found that the association weight between the two is greater than the weight threshold, so they are regarded as a pair of associated nodes in the association graph.

[0048] Furthermore, the method further includes: obtaining all the visual features and text features from the first parsing result as geological entities; extracting the relationships between the geological entities and the attributes of the geological entities from the text annotations of the oil and gas geological map data; updating the knowledge graph based on the geological entities, the relationships between the geological entities, and the attributes of the geological entities; and outputting the knowledge graph to obtain the second parsing result of the oil and gas geological map.

[0049] When updating the knowledge graph, each geological entity is linked to the existing knowledge graph based on the geological entities, the relationships between them, and their attributes. Incremental updates to the existing knowledge graph are performed based on newly parsed geological map data, eliminating the need to rebuild the entire graph. For example, after adding a new geological map, only newly discovered entities, relationships, and attributes are added, and the association information of related nodes is updated, improving update efficiency.

[0050] Existing methods can only extract information explicitly expressed in the map and lack the ability to reason based on existing knowledge and discover implicit information. The method proposed in this invention can construct a geological knowledge map based on the analysis results, support the accumulation, sharing and reasoning of knowledge, and discover the implicit geological features and patterns in the geological map.

[0051] Once the knowledge graph is obtained, knowledge reasoning can be performed based on the graph to discover new relationships. During reasoning, existing models can be used, such as the Qianwen model. For example, by analyzing a geological map, the "cutting" relationship between "Fault F2" (entity 1) and "Permian strata" (entity 2) can be extracted (visually, the fault line runs through the stratigraphic block, and the text label reads "F2 cuts the Permian"). Based on the rule in the knowledge graph that "cutting relationship means the fault formed later than the cut strata," a new relationship of "Fault F2 formed after the Permian" can be inferred. Similarly, after extracting the "occurrence" relationship between "oil and gas reservoir A" and "sandstone reservoir B," combined with the common knowledge in the knowledge graph that "sandstone reservoirs are the main storage space for oil and gas reservoirs," the implicit relationship of "the physical properties of sandstone reservoir B affect the reserves of oil and gas reservoir A" can be inferred.

[0052] Furthermore, the method further includes: extracting features and their spatial location information from the oil and gas geological map data, determining the number of features and the size of a first quantity threshold; if the number of features is less than the first quantity threshold, determining that the oil and gas geological map data is fuzzy data; discarding the features and their spatial location information extracted from the fuzzy data; and preprocessing the fuzzy data, and re-executing the oil and gas geological map parsing method based on the preprocessed fuzzy data, wherein the preprocessing includes image denoising, tilt correction, contrast enhancement, resolution adjustment, and color space conversion.

[0053] The first quantity threshold can be set according to the actual situation. If the number of extracted features is less than the first quantity threshold, the image is considered to have potential image quality problems. In this case, image preprocessing can improve the accuracy of subsequent recognition. Image denoising is used to remove speckle noise or interfering pixels in the scanned image or during electronic transmission, such as Gaussian filtering for denoising scanned images or smoothing for vector graphics. Tilt correction is used to correct the angle of the scanned image, such as correcting the tilt angle of the scanned image through Hough transform, ensuring a maximum tilt angle correction range of ±15°. Contrast enhancement is used to improve the contrast of the image, strengthen the boundaries of color blocks, and highlight detailed features, such as adaptive histogram equalization, which enhances details in images with unclear dark areas. Resolution adjustment is used to unify the image resolution for easier subsequent processing, such as unifying it to 300 DPI for easier subsequent recognition. Color space conversion is used to convert the image to a color space suitable for analysis, such as converting from RGB to HSV, improving the stability of color recognition.

[0054] The oil and gas geological map analysis method provided by this invention integrates computer vision, natural language processing and knowledge graph technologies to achieve automatic analysis and understanding of oil and gas geological maps, improve analysis efficiency and accuracy, lower the professional threshold, and provide intelligent data processing and analysis means for oil and gas exploration and development.

[0055] This invention also provides an apparatus for analyzing oil and gas geological maps, for example, such as... Figure 2 As shown, the device includes a memory and a processor, the processor being used to run a program, wherein the program, when run, is used to execute the oil and gas geological map analysis method described above.

[0056] The processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured, and the analysis of oil and gas geological maps can be achieved by adjusting kernel parameters.

[0057] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0058] This invention provides a storage medium storing a program that, when executed by a processor, implements the oil and gas geological map analysis method.

[0059] This invention provides a processor for running a program, wherein the program executes the oil and gas geological map analysis method.

[0060] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0061] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0062] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0063] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0064] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0065] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0066] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0067] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0068] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for analyzing oil and gas geological maps, characterized in that, The method includes: Obtain oil and gas geological map data; Features and their spatial location information are extracted from the oil and gas geological map data. The features are in vector form and include text features and visual features. Based on the features and their spatial location information, the association weights between the features are calculated, wherein the association weights between the text features and the visual features are calculated based on a cross-modal attention mechanism; Construct an association graph among the features based on the association weights; The association graph is processed using a graph neural network to obtain an updated association graph that incorporates global dependencies; and The updated correlation map is output to obtain the first analytical result of the oil and gas geological map.

2. The method according to claim 1, characterized in that, Extracting text features from the oil and gas geological map data includes: Based on the EAST algorithm, text region detection is performed on the oil and gas geological map data to obtain text regions; Based on the OCR model, text information is extracted from the text region; and Based on a pre-trained language model, text features are extracted from the text information, including technical terms, parameter values, and / or sentence semantics.

3. The method according to claim 2, characterized in that, Extracting visual features from the oil and gas geological map data based on a pre-trained convolutional neural network includes: Based on the YOLOv8 model, map elements are identified from the oil and gas geological map data, including well locations, faults, and contour lines; Based on Hough transform and support vector machine classification, lines are identified and classified from the oil and gas geological map data. The classification of lines includes structural lines and fault lines. Based on the K-means clustering algorithm, different color blocks are identified from the oil and gas geological map data, and their lithology and reservoir types are labeled; and Based on target detection and text association, the legend area is located in the oil and gas geological map data, and the correspondence between the symbols and meanings of the legend area is obtained.

4. The method according to claim 3, characterized in that, The method for extracting the spatial location information includes: For text features, the position coordinates of the text region corresponding to the text feature are used as spatial location information; and For visual features, the position of the visual element corresponding to the visual feature is used as spatial position information. The visual element includes the primitive, line, color block and legend.

5. The method according to claim 3, characterized in that, The method for calculating the association weight between the text features and the visual features includes: The text features, visual features, and spatial location information are unified in terms of dimensions; Based on the unified spatial location information, calculate the spatial location attention weight between the text features and the visual features; Based on a pre-trained attention module, a similarity score is calculated between the unified-dimensional text features and the visual features. This similarity score is then normalized and used as the cross-modal semantic attention weight between the two features. Based on a preset fusion coefficient, the spatial location attention weight and the cross-modal semantic attention weight are weighted and fused to obtain the association weight. The attention module is trained on a training set consisting of text feature-visual feature data pairs.

6. The method according to claim 5, characterized in that, The formula for calculating the spatial location attention weight is as follows: Where W is the spatial location attention weight, d is the relative distance between the text feature and the visual feature, and α is an adjustable attenuation parameter.

7. The method according to claim 5, characterized in that, After calculating the similarity score between the text features and the visual features after unifying the dimension, the method further includes: Determine the first geological entity type of the text features and the second geological entity type of the visual features; Based on a pre-constructed geological entity type mask, determine whether the first geological entity type and the second geological entity type can be matched; and If no match is found, the similarity score is set to zero.

8. The method according to claim 1, characterized in that, Constructing the association graph of the visual features and text features based on the association weights includes: The visual features and text features are respectively used as visual nodes and text nodes; Iterate through all the association weights of the visual nodes and text nodes, and perform the following operations on any visual node and any text node to obtain the association graph: Determine whether the association weight between the visual node and the text node is greater than a preset weight threshold; If the value is greater than the given value, then the visual node and the text node are considered as a pair of related nodes in the association graph, and an edge is established between the pair of related nodes.

9. The method according to claim 1, characterized in that, The method further includes: All the visual and textual features are obtained as geological entities from the first analysis result; Extract the relationships between geological entities and the attributes of the geological entities from the text annotations of the oil and gas geological map data; The knowledge graph is updated based on the geological entities, the relationships between them, and their attributes; and The knowledge graph is output to obtain the second analytical result of the oil and gas geological map.

10. The method according to claim 1, characterized in that, The method further includes: After extracting features and their spatial location information from the oil and gas geological map data, the number of features is compared with a first quantity threshold. If the number of features is less than the first number threshold, then the oil and gas geological map data is determined to be fuzzy data; Discard the features and their spatial location information extracted from the fuzzy data; and The fuzzy data is preprocessed, and the oil and gas geological map analysis method is re-executed based on the preprocessed fuzzy data. The preprocessing includes image denoising, tilt correction, contrast enhancement, resolution adjustment, and color space conversion.

11. An apparatus for analyzing oil and gas geological maps, the apparatus comprising a memory and a processor, characterized in that, The processor is used to run a program, wherein the program is run to execute the method as described in any one of claims 1-10.

Images