A reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling

Abstract

This invention discloses a reservoir identification method that integrates seismic attribute inversion and paleokarst morphology modeling, belonging to the field of reservoir identification and modeling technology. The method includes: using an image sensor to perform multimodal fusion processing on a standardized image dataset to obtain multimodal fused image data; performing high-precision texture sampling and spatial alignment processing to generate a cross-modal image fusion matrix; inputting the cross-modal image fusion matrix into an image topology map construction process to establish image block node relationships and spatial connection edge weights, generating a paleokarst topology map; coupling the karst-seismic joint matching map with comprehensive reservoir attribute data to establish an effective reservoir evaluation model; and generating a multi-level reservoir prediction layer set through weighted fusion and multi-parameter spatial analysis. By constructing the paleokarst topology map and the karst-seismic joint matching map, spatial modeling of complex karst structures and cross-modal correlation analysis of seismic anomaly attributes are realized.

Description

Technical Field

[0001] This invention relates to the field of reservoir identification and modeling technology, and in particular to a reservoir identification method that integrates seismic attribute inversion and paleokarst morphology modeling. Background Technology

[0002] With the continuous exploration and development of unconventional oil and gas resources, reservoir identification technology under complex geological conditions has gradually become a research hotspot in the field of oil and gas geophysics and geological modeling. Among them, seismic attribute inversion technology, as an important bridge connecting seismic data and geological attribute parameters, can effectively extract various subsurface information such as lithology, porosity, and fracture development contained in seismic data, and has been widely used in reservoir prediction and fine structural interpretation.

[0003] Current technologies still have certain limitations in fusing paleokalite morphology and seismic attribute inversion data. Most current methods focus on extracting features from a single mode for modeling, lacking in-depth fusion mechanisms for cross-modal collaboration between multi-source geological data, especially image data, and seismic attribute volumes. This makes it difficult to fully represent the coupling relationship between karst structural features and seismic anomaly attributes. Furthermore, in establishing paleokalite spatial morphology maps, existing technologies largely rely on traditional geometric modeling methods, failing to fully incorporate graph structure theory to characterize the topological relationships between image patches. Consequently, they struggle to accurately represent the spatial connectivity and weight distribution between key nodes in the karst system. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a reservoir identification method that integrates seismic attribute inversion and paleokarst morphology modeling to solve the problems of insufficient cross-modal fusion between multi-source data and inaccurate expression of karst structure topology.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a reservoir identification method that integrates seismic attribute inversion and paleokarst morphology modeling, which includes acquiring a multi-source geological image dataset, performing format unification, spatial label binding, image enhancement and scale normalization processing to construct a standardized image dataset, and combining seismic inversion calculation to generate a structured seismic inversion attribute body.

[0008] Multimodal fusion processing is performed on a standardized image dataset using an image sensor to obtain multimodal fused image data. High-precision texture sampling and spatial alignment processing are then performed to generate a cross-modal image fusion matrix.

[0009] The cross-modal image fusion matrix is ​​input into the image topology graph construction process to establish the relationships between image block nodes and the weights of spatial connection edges, thereby generating an ancient karst topology map.

[0010] By spatially overlaying the paleokarst topology map with the structured seismic inversion attribute volume, and analyzing the correlation between topological structure and seismic anomaly attributes through spatial location correspondence and feature matching, a karst-seismic joint matching map is generated.

[0011] By coupling karst seismic joint matching maps with comprehensive reservoir attribute data, an effective reservoir evaluation model is established. Through weighted fusion and multi-parameter spatial analysis, a multi-level reservoir prediction layer set is generated.

[0012] As a preferred embodiment of the reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling described in this invention, the specific steps for constructing the standardized image dataset are as follows:

[0013] Collect core cross-section images, well profile images, geomorphic outcrop remote sensing images, well logging curve images, and seismic profile images of the target area to construct an original multi-source geological image dataset;

[0014] The original multi-source geological image dataset was uniformly converted into grayscale image format to obtain a unified format original geological grayscale image set;

[0015] The original geological grayscale image set in a unified format is subjected to format unification, spatial label binding, image enhancement and scale normalization processing to generate a standardized image dataset.

[0016] As a preferred embodiment of the reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling described in this invention, the specific steps for generating the structured seismic inversion attribute volume are as follows:

[0017] Three-dimensional seismic reflection data corresponding to the target area are extracted from the standardized image dataset, and spatial coordinate alignment is performed to obtain spatially aligned seismic data volume. Pixel position mapping and sampling point number matching operations are then performed to establish a structural index table.

[0018] Using the structural index table, reflection waveform extraction and physical property function calculation are performed on the spatially aligned seismic data volume to obtain the initial seismic attribute volume set.

[0019] The initial set of seismic attribute volumes is subjected to multi-attribute fusion processing to obtain fused seismic feature volumes, which are then resampled and archived according to the spatial voxel structure to form structured seismic inversion attribute volumes.

[0020] As a preferred embodiment of the reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling described in this invention, the specific steps of using an image sensor to perform multimodal fusion processing on a standardized image dataset to obtain multimodal fused image data are as follows.

[0021] Using image sensors, multi-source geological image information in standardized image datasets is digitally read and multimodal information is extracted to extract multi-scale texture features and spatial structure features;

[0022] Multi-scale texture features and spatial structure features are combined through spatial registration, feature extraction, and weighted fusion operations to generate multimodal fused image data.

[0023] As a preferred embodiment of the reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling described in this invention, the specific steps for performing high-precision texture sampling and spatial alignment processing to generate a cross-modal image fusion matrix are as follows.

[0024] Based on the spatial range of the multimodal fused image data, a unified reference grid is constructed through boundary coordinate extraction and regular grid division operations. Local resampling processing is then performed on the texture channels in the multimodal fused image data to obtain a multimodal texture resampled image set.

[0025] Spatial structure orientation information is extracted from a multimodal texture resampled image set, and a main texture orientation field is constructed through multi-scale gradient tensor analysis and orientation consistency interpolation.

[0026] Align the main texture orientation field with the spatial normal distribution of the unified reference mesh to obtain the texture orientation field, and generate a cross-modal image fusion matrix through feature mapping and weighted fusion operations.

[0027] As a preferred embodiment of the reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling described in this invention, the specific steps for generating the paleokarst topographic map are as follows:

[0028] An adaptive region growing algorithm is used to spatially segment the cross-modal image fusion matrix, forming multiple spatially connected image block nodes. The mean texture intensity, local geometric morphology parameters, and spatial three-dimensional coordinates are extracted, and feature vector normalization and fusion processing are performed to construct a multi-dimensional feature vector.

[0029] Calculate the similarity and spatial adjacency between multidimensional feature vector nodes, generate weighted connection edges, combine image block nodes with weighted connection edges, and generate an image topology graph structure.

[0030] Low-weight edges and isolated nodes are removed from the image topology graph to generate an ancient karst topology map.

[0031] As a preferred embodiment of the reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling described in this invention, the method involves spatially overlaying the paleokarst topology map with the structured seismic inversion attribute volume, analyzing the correlation between the topological structure and seismic anomaly attributes through spatial location correspondence and feature matching, and generating a karst-seismic joint matching map. The specific steps are as follows:

[0032] The spatial coordinates of the ancient karst topology map and the structured seismic inversion attribute volume are uniformly transformed and rigidly registered to construct a spatial alignment and fusion model, and the topological node structural feature vectors and seismic attribute volume anomaly features are extracted.

[0033] The joint matching weights of the topological node structure feature vector and the seismic attribute volume anomaly features are calculated using normalized fractional functions to construct a karst earthquake matching map.

[0034] The karst earthquake matching map is input into a graph clustering algorithm. Based on the joint matching weight, topological node structure feature vector, and earthquake attribute volume anomaly features, the algorithm identifies karst anomaly high matching areas and generates a karst earthquake joint matching map.

[0035] As a preferred embodiment of the reservoir identification method that integrates seismic attribute inversion and paleokarst morphology modeling as described in this invention, the spatial alignment fusion model is a spatially consistent fusion representation constructed by rigid registration and multi-dimensional attribute fusion based on the core node coordinates of the karst-seismic joint matching map and the spatial voxels of the structured seismic inversion attribute volume.

[0036] As a preferred embodiment of the reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling described in this invention, the method involves coupling karst-seismic joint matching maps with comprehensive reservoir attribute data to establish an effective reservoir evaluation model, and generating a multi-level reservoir prediction layer set through weighted fusion and multi-parameter spatial analysis. The specific steps are as follows:

[0037] From the karst earthquake joint matching map, select the set of key nodes whose joint matching weight exceeds the preset matching strength threshold, extract the corresponding spatial coordinates, and construct a spatial index mapping table;

[0038] Collect comprehensive reservoir attribute data and sample attributes at key node spatial locations marked in the spatial index mapping table to form a multi-parameter attribute vector set;

[0039] By using a nonlinear multi-parameter fusion function, the multi-parameter attribute vector set is coupled with the node weights in the karst seismic joint matching graph to construct an effective reservoir evaluation model.

[0040] Spatial adaptive smoothing and multi-scale interpolation are applied to effective reservoir evaluation models at different spatial locations to obtain comprehensive reservoir evaluation results. A hierarchical threshold system is constructed to semantically classify high-quality, potential, and inefficient reservoir areas. Multi-level reservoir prediction layer sets are generated through three-dimensional slicing technology.

[0041] As a preferred embodiment of the reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling described in this invention, the comprehensive reservoir attribute data includes porosity, permeability, fluid saturation, lithology type, fracture characteristics, wave impedance, elastic parameters, and formation pressure and temperature multidimensional physical and structural properties.

[0042] The beneficial effects of this invention are as follows: by constructing paleokarst topological maps and karst-seismic joint matching maps, spatial modeling of complex karst structures and cross-modal correlation analysis of seismic anomaly attributes are realized. The former effectively improves the structural authenticity and connectivity expression of karst morphology recognition through multi-dimensional feature extraction and graph structure construction, while the latter establishes a coupling channel between structural information and seismic attributes through coordinate overlay and feature matching, enhancing the spatial consistency and physical property matching degree of reservoir identification. Finally, based on the integration of standardized image processing and reservoir attribute data analysis, a multi-level reservoir prediction layer set is formed, which significantly improves the accuracy and practicality of reservoir prediction. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 A flowchart of a reservoir identification method that integrates seismic attribute inversion and paleokarst morphology modeling;

[0045] Figure 2 Flowchart for building a standardized image dataset;

[0046] Figure 3 Flowchart for generating structured seismic inversion attribute volumes;

[0047] Figure 4 Flowchart for generating cross-modal image fusion matrix. Detailed Implementation

[0048] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0049] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0050] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0051] Reference Figures 1-4 This is one embodiment of the present invention, which provides a reservoir identification method that integrates seismic attribute inversion and paleokarst morphology modeling, including the following steps:

[0052] S1. Collect multi-source geological image datasets and perform format unification, spatial label binding, image enhancement and scale normalization processing to construct a standardized image dataset. Combined with seismic inversion calculations, generate a structured seismic inversion attribute body.

[0053] S1.1 Collect core cross-section images, well profile images, geomorphic outcrop remote sensing images, well logging curve images, and seismic profile images of the target area to construct an original multi-source geological image dataset.

[0054] Furthermore, when acquiring core cross-section images of the target area, the core samples obtained from drilling are longitudinally cut, and high-resolution image acquisition equipment is used to acquire cross-section images, preserving lithological bedding and pore structure characteristics. When acquiring drilling profile images, the cross-sectional structure diagram formed during drilling is obtained, preserving stratigraphic boundaries and depth labeling information. When acquiring geomorphic outcrop remote sensing images, high-resolution remote sensing images covering the target area are extracted, preserving typical outcrop morphology and spatial resolution information. When acquiring well logging curve images, the curves of logging parameters such as resistivity, sonic transit time, and gamma are visualized and scaled uniformly according to well depth. When acquiring seismic profile images, two-dimensional seismic reflection images are exported and spatial coordinate information is preserved. The aforementioned core cross-section images, drilling profile images, geomorphic outcrop remote sensing images, well logging curve images, and seismic profile images together constitute the original multi-source geological image dataset.

[0055] S1.2. Convert the original multi-source geological image dataset into a unified grayscale image format to obtain a unified format original geological grayscale image set.

[0056] Furthermore, the process of unifying the original multi-source geological image dataset into a grayscale image format includes: using channel compression methods to convert color images into grayscale images for core cross-section images, preserving bedding and porosity contrast information; extracting color layer information from well profile images and converting them into grayscale value matrices, preserving depth and structural relationships; using spectral compression methods to convert geomorphic outcrop remote sensing images into grayscale, such as using the average weighting method or the maximum component method, preserving geomorphic boundary features; mapping the contrast between the curve and the background to grayscale levels for well logging curve images, maintaining clear curve trends; and using linear weighting methods to convert color gradients into grayscale intensity distributions for seismic profile images, preserving reflective interface features. After completing the above processing, the core cross-section images, well profile images, geomorphic outcrop remote sensing images, well logging curve images, and seismic profile images are unified into a single-channel grayscale image format, constituting a unified format original geological grayscale image set.

[0057] S1.3 Perform format unification, spatial label binding, image enhancement and scale normalization processing on the original geological grayscale image set in the unified format to generate a standardized image dataset.

[0058] Furthermore, the process of performing format unification, spatial label binding, image enhancement, and scale normalization on the unified format original geological grayscale image set includes: unifying the image encoding format, resolution, and bit depth; performing spatial label binding operations to bind the core cross-section image, well profile image, geomorphic outcrop remote sensing image, well logging curve image, and seismic profile image to their corresponding spatial coordinates, depth information, or projection information; performing image enhancement processing according to image type, such as edge enhancement for seismic profile images, contrast stretching for well logging curve images, and histogram equalization for core cross-section images; and performing scale normalization processing after image enhancement, using bilinear interpolation or cubic convolution interpolation methods to unify the images to a consistent spatial resolution or depth ratio, ultimately generating a standardized image dataset.

[0059] S1.4 Extract the three-dimensional seismic reflection data corresponding to the target area from the standardized image dataset, perform spatial coordinate alignment processing to obtain the spatially aligned seismic data volume, and perform pixel position mapping and sampling point number matching operations to establish a structural index table.

[0060] Furthermore, the process of extracting 3D seismic reflection data corresponding to the target area from the standardized image dataset includes: determining the geological coordinate range of the target area by reading the spatial label field in the standardized image dataset, extracting the 3D seismic reflection profile image of the corresponding area, and reconstructing it into a 3D seismic reflector; performing spatial coordinate alignment processing on the 3D seismic reflector using spatial affine transformation or rigid transformation methods, and maintaining spatial reference consistency with the standardized image dataset; after alignment, performing pixel position mapping operation to establish the spatial correspondence between voxels in the 3D seismic reflector and pixels in the standardized image dataset; finally, assigning sampling point numbers to each valid voxel according to the spatial correspondence, organizing voxel coordinates, pixel positions and sampling numbers, and constructing a structural index table.

[0061] It should also be noted that the target area refers to the geological spatial range in the standardized image dataset that is related to the reservoir to be identified, which is obtained by filtering spatial labels from core sections, well profiles, remote sensing images, well logging curves, and seismic profile images.

[0062] S1.5. Using the structural index table, perform reflection waveform extraction and physical attribute function calculation on the spatially aligned seismic data volume to obtain the initial seismic attribute volume set.

[0063] Furthermore, using the structural index table, the reflected waveform information is first extracted from each valid voxel in the spatially aligned seismic data volume. Specifically, this includes calculating the amplitude, phase, and frequency characteristics of the reflected waves to reflect the changes in the subsurface geological interface.

[0064] Then, based on the extracted reflection waveforms, the corresponding seismic physical properties, such as wave impedance, elastic modulus, and velocity, are calculated using physical property functions. These physical properties are obtained through the known correspondence between waveforms and seismic physical properties. All calculated reflection waveform features and physical property data are associated with voxel spatial coordinates defined in the structural index table to form an initial set of seismic property volumes, providing a foundation for subsequent multi-attribute fusion and spatial resampling.

[0065] S1.6 Perform multi-attribute fusion processing on the initial set of seismic attribute volumes to obtain fused seismic feature volumes, and resample and archive them according to the spatial voxel structure to form structured seismic inversion attribute volumes.

[0066] Furthermore, a weighted fusion method is employed to fuse multiple seismic attribute data from the initial seismic attribute volume set. This involves determining fusion weights based on attribute correlation and spatial consistency, and weighted averaging of attributes at the same spatial voxel locations to enhance the stability and accuracy of seismic features. After fusion, the fused seismic feature volume is resampled in three dimensions according to a pre-defined spatial voxel structure. Interpolation algorithms are used to adjust data resolution and consistency, ensuring continuous spatial distribution and uniform resolution. Finally, the resampled fused seismic feature volume is archived and managed according to spatial voxel coordinates, forming a structured seismic inversion attribute volume, providing a foundation for subsequent reservoir identification and analysis.

[0067] It should also be noted that spatial voxel structure refers to dividing seismic reflection data and its corresponding attributes into several regular blocks at fixed intervals, based on the spatial coordinates and sampling point numbers in a standardized image dataset and following a unified three-dimensional raster partitioning rule. Each regular block has a clear spatial location and volume range, and there are clear adjacency relationships between them, thus constructing a structurally ordered three-dimensional voxel grid. Spatial voxel structure ensures the continuity and integrity of the fused seismic features in the spatial dimension, facilitating resampling, archiving, and subsequent spatial analysis operations.

[0068] S2. Use an image sensor to perform multimodal fusion processing on a standardized image dataset to obtain multimodal fused image data and perform high-precision texture sampling and spatial alignment processing to generate a cross-modal image fusion matrix.

[0069] S2.1 Using an image sensor, digital reading and multimodal information extraction are performed on multi-source geological image information in a standardized image dataset to extract multi-scale texture features and spatial structure features.

[0070] Furthermore, image sensors are used to digitize multi-source geological image information from standardized image datasets. Acquisition devices convert the image's grayscale values ​​and spatial coordinates into digital signals. Based on the digitized images, multi-scale texture analysis methods are employed to extract multi-scale texture features, including gray-level co-occurrence matrix, local binary patterns, and frequency domain transforms, to describe the image's detailed structure and texture distribution.

[0071] Simultaneously, spatial structure analysis methods are used to calculate edge direction, spatial frequency, and spatial correlation in the image, obtaining spatial structure features that reflect the tectonic morphology at different spatial scales in the geological image. Finally, multi-scale texture features and spatial structure features are used as the results of multimodal information extraction, providing a foundation for subsequent fusion processing.

[0072] S2.2. Multi-scale texture features and spatial structure features are combined through spatial registration, feature extraction and weighted fusion operations to generate multi-modal fused image data.

[0073] Furthermore, multi-scale texture features and spatial structure features are geometrically corrected using spatial registration methods based on unified spatial coordinates to eliminate spatial biases and scale differences between different modalities, ensuring accurate spatial correspondence of features. Subsequently, key features are extracted from the registered multi-scale texture features and spatial structure features respectively. Statistical calculations use a sliding window to extract feature values ​​such as mean, variance, and skewness of local regions. Spatial frequency analysis is based on Fourier transform to identify the frequency distribution patterns of texture and structural features to extract key characterization information. Based on the importance of features and their contribution to reservoir identification, a weighted fusion operation is used to weight and superimpose multi-scale texture features and spatial structure features at corresponding spatial locations, forming fused multi-modal image data, achieving effective integration and enhancement of information from different modalities.

[0074] S2.3. Based on the spatial range of the multimodal fused image data, a unified reference grid is constructed through boundary coordinate extraction and regular grid division operations. Local resampling processing is then performed on the texture channels in the multimodal fused image data to obtain a multimodal texture resampled image set.

[0075] Furthermore, based on the spatial extent of the multimodal fused image data, boundary coordinates are first extracted to clarify the spatial coverage of the fused image. Based on these boundary coordinates, a unified reference grid is constructed using a regular 3D raster division method. This unified reference grid consists of equally spaced spatial nodes, ensuring that each spatial location has a unique identifier and a definite position.

[0076] Subsequently, local resampling is performed on the texture channels in the multimodal fused image data. Interpolation algorithms are then used to spatially adjust and refine the texture information, ensuring the continuity and consistency of the texture data on a unified reference grid. Through these steps, a multimodal texture resampled image set with spatial uniformity and high resolution is generated.

[0077] It should also be noted that the spatial extent of the multimodal fused image data originates from the spatial label fields of various multi-source geological images in the standardized image dataset. These spatial labels include the spatial coordinates, well depth information, and geological unit boundary information recorded in core cross-section images, well profile images, geomorphic outcrop remote sensing images, well logging curve images, and seismic profile images. By unifying the spatial reference system of various images and registering and aligning the spatial coordinates, well depth information, and geological boundary information, the overall spatial coverage of the multimodal fused image data is determined, ensuring the consistency and integrity of the fused data with the original geological information in terms of spatial location.

[0078] S2.4 Extract spatial structure orientation information from the multimodal texture resampled image set, and construct the main texture orientation field through multi-scale gradient tensor analysis and orientation consistency interpolation.

[0079] Furthermore, from the multimodal texture resampled image set, a multi-scale gradient tensor field is constructed in the local region of the image using multi-scale gradient tensor analysis. The gradient magnitude and direction angle in each direction are calculated to extract local gradient information. By weighted averaging of the gradient tensor fields at different scales, the spatial structure's direction information is extracted. Based on this spatial structure direction information, direction consistency interpolation is used to smooth and compensate for the direction information of adjacent pixels, eliminating noise and local inconsistencies. Finally, the interpolated direction information is integrated to construct a master texture direction field reflecting the dominant direction of the multimodal texture, providing a precise direction reference for spatial alignment and fusion of texture directions.

[0080] S2.5 Align the main texture direction field with the spatial normal distribution of the unified reference mesh to obtain the texture direction field, and generate a cross-modal image fusion matrix through feature mapping and weighted fusion operations.

[0081] Furthermore, the main texture direction field is mapped to the spatial normal distribution of a unified reference mesh. By calculating the angle difference between the main texture direction and the spatial normal vector, spatial alignment is achieved, ensuring that the texture direction is consistent with the geological spatial structure. Based on the alignment results, two key features—the main direction angle and direction consistency—are extracted from the texture direction field. Combined with the geometric properties of the spatial normal, a weighted fusion method is used to superimpose these features, enhancing direction consistency and spatial continuity. Finally, a cross-modal image fusion matrix is ​​generated, reflecting the comprehensive information of texture direction and spatial structure in different modal images, supporting subsequent topology construction and spatial analysis.

[0082] It should also be noted that the spatial normal distribution originates from the spatial coordinates of a unified reference grid and 3D geological structure information, and is obtained by calculating the geological interface normal vectors at grid nodes. Specifically, based on the spatial range and regular grid division of multimodal fused image data, the spatial variation gradient of geological structures within local neighborhoods is utilized. The finite difference method is used to calculate the attribute differences between neighboring nodes to obtain the normal direction, forming a spatial normal distribution covering the entire reference grid, which is used to assist in the spatial alignment and fusion processing of texture directions.

[0083] S3. Input the cross-modal image fusion matrix into the image topology graph structure construction process, establish the image block node relationship and spatial connection edge weight, and generate the paleokarst topology map.

[0084] S3.1. An adaptive region growing algorithm is used to spatially segment the cross-modal image fusion matrix to form multiple spatially connected image block nodes. The mean texture intensity, local geometric morphology parameters, and spatial three-dimensional coordinates are extracted, and feature vector normalization and fusion processing are performed to construct a multi-dimensional feature vector.

[0085] Furthermore, an adaptive region growing algorithm is employed to spatially segment the cross-modal image fusion matrix. By setting a similarity threshold and neighborhood connectivity rules, pixels with similar texture and structural features within the space are recursively merged to form multiple spatially connected image block nodes. For each image block node, the mean texture intensity is calculated, local geometric parameters such as shape compactness and surface curvature are extracted, and the node's spatial three-dimensional coordinates are obtained. After feature normalization processing of the mean texture intensity, local geometric parameters, and spatial three-dimensional coordinates to eliminate dimensional differences, a multi-dimensional feature vector is synthesized through feature fusion to comprehensively characterize the integrated spatial attributes and texture features of each image block node.

[0086] It should also be noted that the adaptive region growing algorithm is based on pixel texture intensity and spatial adjacency in the cross-modal image fusion matrix. First, it selects pixels with small variations in texture intensity and spatial proximity as initial growth seeds. Then, according to a preset similarity threshold, it gradually merges neighboring pixels that satisfy similar texture intensity and spatial continuity into the corresponding regions. The adaptive region growing algorithm dynamically adjusts the similarity threshold to adapt to the texture complexity of different regions, avoiding over-segmentation or under-segmentation, and ensuring the spatial coherence and texture uniformity of the segmented regions. Through iterative growth, multiple spatially connected image block nodes with consistent texture features are ultimately formed, supporting subsequent feature extraction and multi-dimensional feature vector construction.

[0087] The similarity threshold is set based on the statistical characteristics of the matching weights in the karst-seismic joint matching map. By combining the overall balance level and dispersion of the matching weights, a representative weight boundary is determined. The similarity threshold is used to identify key nodes with high matching weights and significant correlations in seismic anomaly manifestations and topological structures, thereby ensuring the accuracy and stability of subsequent spatial indexing mapping and reservoir attribute sampling processes.

[0088] S3.2 Calculate the similarity and spatial adjacency between multidimensional feature vector nodes, generate weighted connecting edges, combine image block nodes with weighted connecting edges, and generate image topology graph structure.

[0089] Furthermore, based on the multidimensional feature vectors, the similarity between nodes of each image patch is calculated, expressed as:

[0090]

[0091] Among them, S mn Let v represent the similarity between image patch node m and image patch node n, where m represents the index of the first image patch node, n represents the index of the second image patch node, and v m V is represented as the multidimensional feature vector of the m-th image patch node. n Represented as the multidimensional feature vector of the nth image patch node, ||vm -v n || represents the Euclidean distance between the feature vectors of image block nodes m and n.

[0092] The similarity of node features is quantified using methods such as Euclidean distance or cosine similarity. By combining the spatial 3D coordinates of image patch nodes, the spatial adjacency relationship between nodes is determined, identifying adjacent node pairs. Based on similarity and adjacency, the weight values ​​of connecting edges are calculated; these weight values ​​comprehensively reflect the similarity of node features and spatial proximity. Image patch nodes are treated as nodes in the topology graph, and connecting edges and their weights are treated as weighted connecting edges. Combining nodes with weighted connecting edges constructs an image topology graph structure that reflects the spatial structure and feature relationships of image patches, supporting subsequent topology analysis and node selection.

[0093] S3.3 Remove low-weight edges and isolated nodes from the image topology graph structure to generate an ancient karst topology map.

[0094] Furthermore, edge weight and node distribution analysis are performed on the image topology graph structure graph. Edges with weights lower than the weight distribution threshold in the image topology graph structure graph are removed, and isolated nodes that have not formed a connection relationship in the image topology graph structure graph are removed. Stable topological connection structures and main texture paths are retained, and finally, an ancient karst topology map is generated.

[0095] It should also be noted that the weight distribution threshold is set based on the statistical characteristics of all edge weights in the image topology graph. It is usually determined by calculating the mean and standard deviation of the edge weights and combining them with the adjustment coefficient. This threshold is used to distinguish between important connections and weak connections, ensuring that key topological structures and texture paths are preserved.

[0096] S4. Spatial coordinates are superimposed on the paleokarst topology map and the structured seismic inversion attribute volume. Through spatial location correspondence and feature matching, the correlation between topological structure and seismic anomaly attributes is analyzed to generate a karst-seismic joint matching map.

[0097] S4.1. The spatial coordinates of the paleokarst topology map and the structured seismic inversion attribute volume are uniformly transformed and rigidly registered to construct a spatial alignment fusion model, and the topological node structural feature vectors and seismic attribute volume anomaly features are extracted.

[0098] Furthermore, a unified coordinate datum transformation was performed on the spatial coordinates of the paleokarst topology map and the structured seismic inversion attribute volume. A rigid transformation method was used to complete the registration correction, achieving consistency in their spatial positions. Based on the registration results, topological node structural feature vectors were extracted from the paleokarst topology map, including the nodes' geometric attributes, connectivity relationships, and texture features. Simultaneously, anomalous seismic attribute features, such as reflection intensity anomalies and wave impedance changes, were extracted from the structured seismic inversion attribute volume. These operations form a spatial alignment and fusion model, providing a spatial and feature basis for subsequent joint matching analysis.

[0099] It should also be noted that the spatial alignment and fusion model is set up as follows: using the coordinates of key nodes in the paleokarst topology map and the spatial voxel coordinates of the structured seismic inversion attribute volume as a reference, rigid transformation is used to rotate, translate and scale to achieve coordinate registration; extracting the topological node structural feature vectors and seismic attribute anomaly features, performing normalized weighted fusion to form a spatially consistent fusion representation, providing a basis for karst-seismic joint matching maps.

[0100] S4.2 Calculate the joint matching weight of the topological node structure feature vector and the seismic attribute volume anomaly feature using the normalized fractional function, and construct the karst seismic matching map.

[0101] Furthermore, using a normalized fractional function, the topological node structural feature vector and the seismic attribute volume anomaly features are quantified separately, and a weighted matching score between the two is calculated, expressed as:

[0102]

[0103] Where P represents the total weighted matching score, i represents the feature dimension index, Z represents the total number of feature dimensions, and w i p represents the weight coefficient of the i-th feature dimension. i This represents the matching score on the i-th feature dimension;

[0104] By adjusting the contribution ratio of structural features and anomaly features through weighting coefficients, a joint matching weight value is obtained. Based on the joint matching weight value, the topological nodes are connected to the corresponding seismic attribute volumes to form a karst seismic matching map, which reflects the matching strength and spatial correlation between nodes.

[0105] S4.3 Input the karst earthquake matching map into the graph clustering algorithm. Based on the joint matching weight, topological node structure feature vector and earthquake attribute volume anomaly features, identify karst anomaly high matching areas and generate karst earthquake joint matching map.

[0106] Furthermore, the karst-seismic matching map is input into a graph clustering algorithm. Based on the joint matching weight, topological node structural feature vectors, and seismic attribute volume anomaly features, the similarity between nodes is calculated through the weighted distance of multi-dimensional feature vectors, and the adjacency relationship is determined by the spatial distance and topological connection of nodes, forming the basis for clustering. Through iterative partitioning and aggregation, nodes with high joint matching weights and similar features are grouped into the same cluster, identifying high-matching karst anomaly areas. The graph clustering algorithm divides the nodes with high matching intensity into feature similarity clusters, and constructs subgraph structures based on the topological connection relationships of nodes within each cluster. Finally, the high-matching subgraphs are merged to form a complete karst-seismic joint matching map, reflecting the spatial distribution characteristics of karst anomaly areas and providing a basis for subsequent reservoir identification and evaluation.

[0107] S5. Couple the karst seismic joint matching map with the comprehensive reservoir attribute data to establish an effective reservoir evaluation model, and generate a multi-level reservoir prediction layer set through weighted fusion and multi-parameter spatial analysis.

[0108] S5.1 Select the set of key nodes whose joint matching weight exceeds the preset matching strength threshold from the karst earthquake joint matching map, extract the corresponding spatial coordinates, and construct a spatial index mapping table.

[0109] Furthermore, from the karst-earthquake joint matching map, a preset matching strength threshold is first set to filter out the set of key nodes whose joint matching weight exceeds the preset matching strength threshold; then, the spatial coordinates of the selected key node set are extracted to obtain the three-dimensional geospatial location corresponding to each node; based on the spatial coordinates of the key nodes, a spatial index mapping table is constructed to realize the registration of the correspondence between key nodes and spatial locations, providing basic support for subsequent spatial retrieval and positioning.

[0110] The preset matching strength threshold is based on the statistical distribution of the matching weights of all nodes in the karst-seismic joint matching map. By combining the average matching weight with the standard deviation and adjusting it according to the adjustment coefficient, a deterministic adaptive matching strength threshold is determined to screen nodes with high spatial and feature correlation. This ensures that key nodes with high matching strength are representative and improves the accuracy of subsequent analysis.

[0111] S5.2 Collect comprehensive reservoir attribute data and sample attributes at the key node spatial locations marked in the spatial index mapping table to form a multi-parameter attribute vector set.

[0112] Furthermore, comprehensive reservoir attribute data, encompassing multi-dimensional parameters such as porosity, permeability, rock density, and acoustic impedance, is collected. Based on the spatial locations of key nodes marked in the spatial index mapping table, the corresponding geological spatial coordinates are precisely located for each high-matching-intensity node, and attribute sampling is performed. Reservoir attributes surrounding the high-matching-intensity key nodes are estimated using spatial interpolation methods, and multi-dimensional attributes are fused according to weights to form a comprehensive attribute vector accurately representing the key nodes. This process obtains multi-parameter attribute information for each key node, ensuring the completeness and accuracy of the attribute data. The sampled porosity, permeability, acoustic impedance, and other reservoir attribute data are mapped one-to-one according to spatial location and fused to form a multi-parameter attribute vector set corresponding to the high-matching-intensity nodes, providing multi-dimensional attribute foundational support for subsequent karst reservoir identification and analysis.

[0113] S5.3. Using a nonlinear multi-parameter fusion function, the multi-parameter attribute vector set is coupled with the node weights in the karst seismic joint matching diagram to construct an effective reservoir evaluation model.

[0114] Furthermore, a nonlinear multi-parameter fusion function is used to couple the joint matching weights of corresponding nodes in the karst-seismic joint matching map with the multi-parameter attribute vector set. First, each reservoir attribute in the multi-parameter attribute vector is weighted according to a preset weight. Then, combined with the joint matching weights of the nodes, a nonlinear multi-parameter fusion function is used. The weighted multi-parameter attribute vector and the node joint matching weights are taken as input, and nonlinear coupling is achieved through function mapping. This comprehensively reflects the correlation strength between reservoir attributes and matching features, thereby generating a unified comprehensive evaluation score. The coupling process emphasizes the mutual influence of reservoir attributes and the correlation with seismic matching features. Nonlinear mapping enhances the discriminative power and accuracy of reservoir evaluation, forming an effective reservoir evaluation model that reflects the characteristics of karst reservoirs. The fusion result generates corresponding comprehensive reservoir evaluation indicators for each key node, supporting subsequent reservoir distribution analysis and optimization decisions.

[0115] It should also be noted that the effective reservoir evaluation model uses a nonlinear multi-parameter fusion function to couple the multi-parameter attribute vector with the node weights in the karst seismic joint matching map. By comprehensively considering the correlation between reservoir physical parameters and joint matching weights, the model obtains the comprehensive reservoir evaluation score by performing nonlinear mapping calculation on the weighted multi-parameter attributes and node weights, thereby achieving a quantitative evaluation of reservoir effectiveness.

[0116] S5.4. Implement spatial adaptive smoothing and multi-scale interpolation processing on effective reservoir evaluation models in different spatial locations to obtain comprehensive reservoir evaluation results. Construct a hierarchical threshold system to semantically classify high-quality, potential, and inefficient reservoir areas. Generate multi-level reservoir prediction layer sets through three-dimensional slicing technology.

[0117] Furthermore, for effective reservoir evaluation models in different spatial locations, spatial adaptive smoothing is first implemented to adjust the continuity of the comprehensive reservoir evaluation score in adjacent spatial locations. Then, multi-scale interpolation is used to fill in missing values ​​in the spatial distribution and refine the score distribution to obtain the comprehensive reservoir evaluation results. Based on the comprehensive reservoir evaluation results, a hierarchical threshold system is constructed to divide the reservoir into high-quality areas, potential areas, and inefficient areas to achieve semantic hierarchical classification. Finally, the hierarchical results are spatially displayed using three-dimensional slicing technology to generate a multi-level reservoir prediction layer set, supporting intuitive analysis of reservoir spatial characteristics and development planning.

[0118] It should also be noted that the stratification threshold system is set based on the distribution characteristics of the reservoir comprehensive evaluation results to determine the statistical range of the evaluation scores. Secondly, multiple stratification thresholds are set in combination with the experience of geological experts and the needs of reservoir development to distinguish between high-quality, potential, and inefficient reservoir areas. Then, the reservoir comprehensive evaluation results after spatial adaptive smoothing and multi-scale interpolation are used to classify and label each spatial unit to ensure that the stratification thresholds match the spatial characteristics of the reservoir. Finally, the stratification results are visualized through three-dimensional slicing technology to form a multi-level reservoir prediction layer set, which facilitates subsequent reservoir development analysis and decision-making.

[0119] This embodiment also provides a computer device applicable to reservoir identification methods that integrate seismic attribute inversion and paleokarst morphology modeling, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the reservoir identification method that integrates seismic attribute inversion and paleokarst morphology modeling as proposed in the above embodiment.

[0120] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0121] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the reservoir identification method proposed in the above embodiment, which integrates seismic attribute inversion and paleokarst morphology modeling. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0122] In summary, this invention achieves spatial modeling of complex karst structures and cross-modal correlation analysis of seismic anomaly attributes by constructing paleokarst topographic maps and karst-seismic joint matching maps. The former, through multi-dimensional feature extraction and graph structure construction, effectively enhances the structural realism and connectivity representation of karst morphology identification, while the latter, through coordinate overlay and feature matching, establishes a coupling channel between structural information and seismic attributes, enhancing the spatial consistency and physical property matching degree of reservoir identification. Finally, based on the integration of standardized image processing and reservoir attribute data analysis, a multi-level reservoir prediction layer set is formed, significantly improving the accuracy and practicality of reservoir prediction.

[0123] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling, characterized in that: include, A multi-source geological image dataset was collected, and standardized image datasets were constructed by performing format unification, spatial label binding, image enhancement and scale normalization processing. Combined with seismic inversion calculations, a structured seismic inversion attribute volume was generated. Multimodal fusion processing is performed on a standardized image dataset using an image sensor to obtain multimodal fused image data. High-precision texture sampling and spatial alignment processing are then performed to generate a cross-modal image fusion matrix. The cross-modal image fusion matrix is ​​input into the image topology graph construction process to establish the relationships between image block nodes and the weights of spatial connection edges, thereby generating an ancient karst topology map. By spatially overlaying the paleokarst topology map with the structured seismic inversion attribute volume, and analyzing the correlation between topological structure and seismic anomaly attributes through spatial location correspondence and feature matching, a karst-seismic joint matching map is generated. By coupling karst seismic joint matching maps with comprehensive reservoir attribute data, an effective reservoir evaluation model is established. Through weighted fusion and multi-parameter spatial analysis, a multi-level reservoir prediction layer set is generated.

2. The reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling according to claim 1, characterized in that: The standardized image dataset is constructed using the following steps. Collect core cross-section images, well profile images, geomorphic outcrop remote sensing images, well logging curve images, and seismic profile images of the target area to construct an original multi-source geological image dataset; The original multi-source geological image dataset was uniformly converted into grayscale image format to obtain a unified format original geological grayscale image set; The original geological grayscale image set in a unified format is subjected to format unification, spatial label binding, image enhancement and scale normalization processing to generate a standardized image dataset.

3. The reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling according to claim 1, characterized in that: The specific steps for generating the structured seismic inversion attribute volume are as follows: Three-dimensional seismic reflection data corresponding to the target area are extracted from the standardized image dataset, and spatial coordinate alignment is performed to obtain spatially aligned seismic data volume. Pixel position mapping and sampling point number matching operations are then performed to establish a structural index table. Using the structural index table, reflection waveform extraction and physical property function calculation are performed on the spatially aligned seismic data volume to obtain the initial seismic attribute volume set. The initial set of seismic attribute volumes is subjected to multi-attribute fusion processing to obtain fused seismic feature volumes, which are then resampled and archived according to the spatial voxel structure to form structured seismic inversion attribute volumes.

4. The reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling according to claim 3, characterized in that: The process of using an image sensor to perform multimodal fusion processing on a standardized image dataset to obtain multimodal fused image data involves the following specific steps. Using image sensors, multi-source geological image information in standardized image datasets is digitally read and multimodal information is extracted to extract multi-scale texture features and spatial structure features; Multi-scale texture features and spatial structure features are combined through spatial registration, feature extraction, and weighted fusion operations to generate multimodal fused image data.

5. The reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling according to claim 4, characterized in that: The process of performing high-precision texture sampling and spatial alignment to generate a cross-modal image fusion matrix involves the following steps: Based on the spatial range of the multimodal fused image data, a unified reference grid is constructed through boundary coordinate extraction and regular grid division operations. Local resampling processing is then performed on the texture channels in the multimodal fused image data to obtain a multimodal texture resampled image set. Spatial structure orientation information is extracted from a multimodal texture resampled image set, and a main texture orientation field is constructed through multi-scale gradient tensor analysis and orientation consistency interpolation. Align the main texture orientation field with the spatial normal distribution of the unified reference mesh to obtain the texture orientation field, and generate a cross-modal image fusion matrix through feature mapping and weighted fusion operations.

6. The reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling according to claim 1, characterized in that: The specific steps for generating the paleokarst topographic map are as follows. An adaptive region growing algorithm is used to spatially segment the cross-modal image fusion matrix, forming multiple spatially connected image block nodes. The mean texture intensity, local geometric morphology parameters, and spatial three-dimensional coordinates are extracted, and feature vector normalization and fusion processing are performed to construct a multi-dimensional feature vector. Calculate the similarity and spatial adjacency between multidimensional feature vector nodes, generate weighted connection edges, combine image block nodes with weighted connection edges, and generate an image topology graph structure. Low-weight edges and isolated nodes are removed from the image topology graph to generate an ancient karst topology map.

7. The reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling according to claim 6, characterized in that: The process involves spatially overlaying paleokarst topology maps with structured seismic inversion attribute volumes. Through spatial location correspondence and feature matching, the correlation between topological structure and seismic anomaly attributes is analyzed to generate a karst-seismic joint matching map. The specific steps are as follows: The spatial coordinates of the ancient karst topology map and the structured seismic inversion attribute volume are uniformly transformed and rigidly registered to construct a spatial alignment and fusion model, and the topological node structural feature vectors and seismic attribute volume anomaly features are extracted. The joint matching weights of the topological node structure feature vector and the seismic attribute volume anomaly features are calculated using normalized fractional functions to construct a karst earthquake matching map. The karst earthquake matching map is input into a graph clustering algorithm. Based on the joint matching weight, topological node structure feature vector, and earthquake attribute volume anomaly features, the algorithm identifies karst anomaly high matching areas and generates a karst earthquake joint matching map.

8. The reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling according to claim 7, characterized in that: The spatial alignment and fusion model is a spatially consistent fusion representation constructed based on the coordinates of core nodes in the karst-seismic joint matching map and the spatial voxels of the structured seismic inversion attribute volume, through rigid registration and multi-dimensional attribute fusion.

9. The reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling according to claim 7, characterized in that: The process involves coupling karst seismic joint matching maps with comprehensive reservoir attribute data to establish an effective reservoir evaluation model. Furthermore, through weighted fusion and multi-parameter spatial analysis, a multi-level reservoir prediction layer set is generated. The specific steps are as follows: From the karst earthquake joint matching map, select the set of key nodes whose joint matching weight exceeds the preset matching strength threshold, extract the corresponding spatial coordinates, and construct a spatial index mapping table; Collect comprehensive reservoir attribute data and sample attributes at key node spatial locations marked in the spatial index mapping table to form a multi-parameter attribute vector set; By using a nonlinear multi-parameter fusion function, the multi-parameter attribute vector set is coupled with the node weights in the karst seismic joint matching graph to construct an effective reservoir evaluation model. Spatial adaptive smoothing and multi-scale interpolation are applied to effective reservoir evaluation models at different spatial locations to obtain comprehensive reservoir evaluation results. A hierarchical threshold system is constructed to semantically classify high-quality, potential, and inefficient reservoir areas. Multi-level reservoir prediction layer sets are generated through three-dimensional slicing technology.

10. The reservoir identification method integrating seismic attribute inversion and paleokarst morphology modeling according to claim 9, characterized in that: The comprehensive reservoir attribute data includes porosity, permeability, fluid saturation, lithology, fracture characteristics, wave impedance, elastic parameters, and formation pressure and temperature multidimensional physical and structural properties.

Images