Geological survey method based on big data

Abstract

The invention relates to the technical field of computer technology and geological survey, in particular to a geological survey method based on big data. The method comprises the following steps: acquiring geological survey data of a sub-region, respectively preprocessing data of different data sources, and integrating basic feature vectors of the different data sources obtained by preprocessing; constructing a geologic structure map based on the fused feature vectors of the sub-regions obtained through integration, and introducing an edge weight construction mechanism combining a spatial adjacency relation and feature similarity to obtain edge weights among the sub-regions; and based on the edge weight between the sub-regions, introducing a directivity and edge weight joint modulation mechanism, carrying out node feature propagation, and predicting a geological structure category. The problems that a traditional geological survey method cannot accurately evaluate the credibility of each data source during fusion, neglects the correlation between the regions on the feature level, and is difficult to sense the composite action of the spatial direction difference between the regions and the edge connection strength on the propagation path are solved.

Description

Technical Field

[0001] This invention relates to the fields of computer technology and geological survey technology, and in particular to a geological survey method based on big data. Background Technology

[0002] Geological surveys, as a crucial foundational task in natural resource development and disaster early warning, have yielded complex geological data in recent years, characterized by diverse sources, rich types, and varying resolutions, thanks to advancements in remote sensing imaging, seismic wave detection, and underground radar. Against this backdrop, the core challenge in geological modeling and intelligent analysis has become how to integrate multi-source geological data, extract complementary information from different data sets, and accurately identify geological structural types.

[0003] Traditional geological survey methods suffer from the following problems: multi-source geological data fusion methods fail to accurately assess the reliability of each data source; regional graph structures are often built based on spatial adjacency relationships, ignoring the correlation between regions at the feature level. In areas with complex geological structures, regions may not be adjacent but may have related features, such as similar fault patterns. Such relationships cannot be effectively characterized in existing graph structures; traditional graph convolutional neural networks mostly propagate node features based on adjacency matrices or attention weights, making it difficult to perceive the combined effect of spatial orientation differences between regions and edge connection strength on the propagation path, thus limiting the model's expressive ability when identifying structural types with clear spatial structural orientations, such as faults, folds, and fracture zones. Summary of the Invention

[0004] This invention provides a big data-based geological survey method to address the shortcomings of traditional geological survey methods in accurately assessing the reliability of various data sources during multi-source geological data fusion. Furthermore, traditional methods often construct regional graph structures based on spatial adjacency relationships, neglecting the correlation between regions at the feature level. In geologically complex regions, regions may not be adjacent but may share related features, such as similar fracture patterns. Such relationships cannot be effectively characterized in existing graph structures. Traditional graph convolutional neural networks mostly rely on adjacency matrices or attention weights for node feature propagation, making it difficult to perceive the combined effect of spatial directional differences between regions and edge connection strength on the propagation path. This limits the model's expressive ability in identifying structural types with clear spatial orientations, such as faults, folds, and fracture zones.

[0005] The present invention provides a geological survey method based on big data, which specifically includes the following technical solutions: A geological survey method based on big data includes the following steps: S1. Divide the geological survey area into regions and obtain geological survey data for each sub-region; preprocess the data from different data sources in the sub-regions to obtain the basic feature vectors of different data sources; introduce a data source credibility weighting mechanism to integrate the basic feature vectors of different data sources to obtain the fused feature vector of the sub-region. S2. Based on the fusion feature vectors of sub-regions, a geological structure map is constructed, and an edge weight construction mechanism combining spatial adjacency and feature similarity is introduced to obtain the edge weights between sub-regions. Based on the edge weights between sub-regions, a joint modulation mechanism of directionality and edge weights is introduced to propagate node features, obtain the output features of sub-regions, and predict the geological structure category.

[0006] Preferably, S1 specifically includes: The preprocessing process includes spatial alignment, feature extraction, dimensionality unification, and feature standardization.

[0007] Preferably, S1 specifically includes: In the implementation of the data source credibility weighting mechanism, the original credibility score of the data source is introduced, and the basic feature vectors of different data sources are weighted and summed to obtain the fused feature vector of the sub-region.

[0008] Preferably, S2 specifically includes: A geological structure map is constructed by using sub-regions as nodes and the fused feature vectors of the sub-regions as node feature vectors. In the implementation of an edge weight construction mechanism that combines spatial adjacency and feature similarity, the edge weights between sub-regions are obtained by combining the spatial adjacency and feature similarity of nodes and introducing a linear transformation coefficient matrix.

[0009] Preferably, S2 specifically includes: In the implementation of the joint modulation mechanism of directionality and edge weight, a direction mapping matrix is ​​introduced, and the directionality modulation factor is calculated by combining the center coordinates of the sub-region.

[0010] Preferably, S2 specifically includes: In the implementation of the joint modulation mechanism of directionality and edge weight, the hidden features of nodes are initialized as node feature vectors. Based on the directional modulation factor, the hidden features of adjacent sub-region nodes are weighted element by element in the direction to obtain the hidden features of adjacent sub-region nodes after directional modulation.

[0011] Preferably, S2 specifically includes: In the implementation of the joint modulation mechanism of directionality and edge weight, the hidden features of adjacent sub-region nodes after directional modulation are weighted based on the edge weights between sub-regions, and a node state self-updating matrix is ​​introduced to update the hidden features of the sub-region nodes themselves, so as to obtain the output features of the sub-region.

[0012] Preferably, S2 specifically includes: Based on the output features of the sub-region, a linear classifier is used to predict the geological structure category of the sub-region.

[0013] The beneficial effects of the technical solution of the present invention are: 1. This invention introduces a data source credibility weighting mechanism, which can dynamically adjust the credibility coefficient according to the quality of different data sources in each sub-region when performing multi-source data feature fusion. This effectively suppresses the negative impact of data sources with problems such as missing, interference or insufficient coverage, and enhances the stability of the fused feature vector expression.

[0014] 2. This invention introduces an edge weight construction mechanism that combines spatial adjacency and feature similarity. For edge relationships between node pairs in geological structure maps, it suppresses distant sub-regions through spatial distance terms and strengthens spatial consistency by utilizing feature cosine similarity. This enables the constructed edge weights between nodes to have stronger spatial awareness and feature sensitivity, avoiding the problem of redundant information propagation caused by relying solely on spatial adjacency.

[0015] 3. This invention introduces a joint modulation mechanism of directionality and edge weights to model the spatial relative relationships between sub-regions. It modulates the node feature propagation process into a propagation with spatial direction awareness, strengthens the spatial extension trend of features, suppresses interference from irrelevant directions, and further considers the importance of the connection relationship between sub-regions. This enhances the graph convolutional neural network model's ability to model the consistency of spatial structure and directional changes between sub-regions, improves the accuracy of geological structure type identification, and avoids the problem that traditional graph convolutional neural network models have difficulty capturing directional features. Attached Figure Description

[0016] Figure 1 This is a flowchart of a geological survey method based on big data as described in this invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details a specific scheme for a geological survey method based on big data provided by this invention.

[0020] See attached document Figure 1 The diagram illustrates a flowchart of a big data-based geological survey method according to an embodiment of the present invention, which includes the following steps: S1. Divide the geological survey area into regions and obtain geological survey data for each sub-region; preprocess the data from different data sources in the sub-regions to obtain the basic feature vectors of different data sources; introduce a data source credibility weighting mechanism to integrate the basic feature vectors of different data sources to obtain the fused feature vector of the sub-region. Geological survey area Divided into Sub-regions , Indicates the first Each sub-region can be obtained from data sources such as remote sensing satellites, seismic instruments, and underground radar. Geological survey data, including remote sensing images, seismic wave data, and underground radar images, were collected and preprocessed for each data source within a sub-region. The preprocessing included spatial alignment, feature extraction, dimensionality unification, and feature standardization. Spatial alignment involved aligning the sub-region data... Data from each data source is projected onto a unified spatial coordinate system using geographic coordinates to obtain projected data. A unified spatial resolution is then set, and the projected data is interpolated to a unified spatial size using existing nearest neighbor interpolation or bilinear interpolation methods to obtain spatially aligned data. Feature extraction is performed on sub-regions using existing convolutional neural networks. After spatial alignment, features such as edges, textures, and region morphology are extracted from the data source data, and feature vectors for each data source are output. ,in Represents the set of real numbers. It is the first The channel dimension of each data source; the dimension is uniformly transformed into feature vectors through linear transformation. The channel dimensions are unified into This yields feature vectors with uniform dimensions. The feature standardization mentioned above refers to feature vectors with uniform dimensions. Standardization is performed using existing Min-Max normalization methods; finally, sub-regions are obtained. The Middle The basic feature vector of each data source ,in, , indicating the data source index, This indicates the number of data sources; the techniques used in the above preprocessing process are all well-known to those skilled in the art and will not be elaborated upon here. To enhance subregions The feature fusion process assesses the responsiveness of key geological regions such as faults, folds, and fracture zones. It introduces a data source credibility weighting mechanism, using a softmax function to normalize the original credibility scores of each data source in a sub-region into credibility coefficients. Based on these credibility coefficients, the basic feature vectors from different data sources are weighted and summed to obtain the sub-region's credibility coefficient. fused feature vector The formula is as follows: , in, Subregion fused feature vectors; , indicating the data source index, Indicates the number of data sources; Subregion The Middle The credibility coefficient of a data source; the higher the value, the more reliable it is. , indicates the data source index; Subregion The Middle The original credibility scores of each data source are determined by sub-regions. The Middle The score is obtained by weighting the signal-to-noise ratio, coverage integrity, and texture complexity of each data source. The weight of each indicator is set to 1. The signal-to-noise ratio is obtained by the ratio of signal power to noise power, the coverage integrity is obtained by the ratio of effective data pixels to total pixels, and the texture complexity is calculated by the gray-level co-occurrence matrix. The signal power, noise power, effective data pixels, and total pixels are all derived from geological survey data. Subregion The Middle The original credibility scores of each data source; Subregion The Middle The basic feature vector of each data source; The above formula introduces a data source credibility weighting mechanism, which can dynamically set the credibility coefficient according to the quality of different data sources in each sub-region. This can effectively suppress the negative impact of data sources with problems such as missing, interference, or insufficient coverage, and enhance the stability of the fused feature vector expression.

[0021] S2. Based on the fusion feature vectors of sub-regions, a geological structure map is constructed, and an edge weight construction mechanism combining spatial adjacency and feature similarity is introduced to obtain the edge weights between sub-regions. Based on the edge weights between sub-regions, a joint modulation mechanism of directionality and edge weights is introduced to propagate node features, obtain the output features of sub-regions, and predict the geological structure category.

[0022] get After fusing the feature vectors of each sub-region, the sub-regions Abstracted as a node in a geological structure diagram The set of nodes is denoted as The set of node feature vectors is denoted as Geological structure map is constructed based on node set and node feature vector set. ; Targeting node pairs in geological structure maps To address the edge relationships between nodes, a mechanism combining feature similarity and spatial adjacency is introduced. This mechanism considers both the spatial adjacency and feature similarity between two nodes, and obtains node weights through a nonlinear activation function. The edge weights between them are given by the following formula: , in, Represents node pairs The edge weights between them, i.e., sub-regions and Edge weights between them; Using the Sigmoid activation function, the edge weights can be controlled within a certain range. ; The linear transformation coefficient matrix is ​​obtained through the backpropagation algorithm, which is a well-known technique in the art and will not be described in detail here. It is a column vector containing feature similarity and spatial adjacency; Subregion The fused feature vector, i.e., the first Nodes Node feature vectors; Subregion The fused feature vector, i.e., the first Nodes Node feature vectors; Represents the node feature vector The modulus length; Represents the node feature vector The modulus length; Represents the node feature vector and node feature vectors The feature similarity between the two regions is calculated using cosine similarity. The larger the value, the more similar the geological structures of the two sub-regions. For distance attenuation factor, This is used to control the influence of the distance between adjacent sub-regions on the edge weights, and is learned through the backpropagation algorithm; , Representing sub-regions and The coordinates of the center point; This is used to calculate the straight-line distance between two sub-regions; Used to measure sub-regions and The adjacency relationship is such that the larger the value, the closer the two sub-regions are. The bias term is learned through the backpropagation algorithm; To enhance the directionality of geological structure propagation between sub-regions, and addressing the limitation of traditional graph convolutional neural network models in capturing directional features, a joint modulation mechanism of directionality and edge weights is introduced to initialize the hidden features of nodes, based on the traditional graph convolutional neural network model. Node feature vectors The hidden features of adjacent sub-regions are weighted element-wise by a directional modulation factor, while also incorporating edge weights. Perform overall weighting based on the hidden features of the sub-region nodes themselves. Dynamic propagation is implemented; the update rules for each convolutional layer are improved as follows: , in, Indicates the first Layer nodes Hidden features; This represents the index of the convolutional layer in a graph convolutional neural network model. This indicates the number of convolutional layers in the graph convolutional neural network model; It is the ReLU activation function; The node state self-updating matrix is ​​obtained through joint optimization using gradient descent algorithm and loss function based on multi-class cross-entropy. Indicates the first Layer nodes Hidden features; For nodes In the The self-updating process of the layers ensures that each node retains its own feature memory path during the multi-layer convolution process, and does not completely rely on the propagation of neighboring nodes; Represents a node The set of adjacent nodes; Represents node pairs Edge weights between them; Indicates the first Layer nodes Hidden features; For nodes and nodes Directional modulation factors are used to enhance orientation perception capabilities; This is a direction mapping matrix used to map the direction unit vector to a direction that is analogous to the given direction. The same feature dimensions are learned through the backpropagation algorithm; , Representing sub-regions and The center coordinates, For the node Pointing to node The direction vector represents the subregion. and The relative spatial direction, Indicates will The direction vector is normalized to a unit direction vector; This indicates element-wise multiplication, used to amplify or suppress the hidden features of each node in a specific direction without changing the dimension of the node features. These are the hidden features of adjacent sub-region nodes after directional modulation, representing the features of neighboring nodes. The Hidden features of layers Transmitted to node During the process, according to the nodes and nodes The relative spatial direction for nodes The Hidden features of layers To achieve directional modulation, modulation can be strengthened or weakened. This represents the edge weighting of the hidden features of adjacent sub-region nodes after directional modulation. Weighted; The above formula introduces a joint modulation mechanism of directionality and edge weights to model the spatial relative relationship between sub-regions. To address the issue that adjacent sub-regions in different spatial directions have different effects on the node features of the current sub-region, the node feature propagation process is modulated into a propagation with spatial direction awareness, which strengthens the spatial extension trend of features and suppresses interference from irrelevant directions. At the same time, it further considers the importance of the connection relationship between sub-regions, enhances the graph convolutional neural network model's ability to model the consistency of spatial structure and directional changes between sub-regions, and improves the accuracy of geological structure type identification. go through After layer graph convolution calculation, the output features of each sub-region are finally obtained. The softmax linear classifier is used to predict the geological structure type of the sub-region, such as faults, folds, and fracture zones. The softmax linear classifier is a well-known technique in the field and will not be described in detail here.

[0023] In summary, a geological survey method based on big data has been developed.

[0024] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0025] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0026] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method of geological surveying based on big data, characterized by, The method comprises the following steps: S1. Dividing a geological survey area into sub-regions and obtaining geological survey data of the sub-regions; respectively preprocessing data of different data sources in the geological survey data of the sub-regions to obtain basic feature vectors of the different data sources; introducing a data source credibility weighting mechanism to integrate the basic feature vectors of the different data sources to obtain fusion feature vectors of the sub-regions; S2. Based on the fusion feature vectors of the sub-regions, constructing a geological structure graph and introducing an edge weight construction mechanism combining spatial adjacency relationship and feature similarity to obtain edge weights between the sub-regions; based on the edge weights between the sub-regions, introducing a directional and edge weight joint modulation mechanism to perform node feature propagation to obtain output features of the sub-regions and predict a geological structure category.

2. The method of claim 1, wherein, The S1 specifically comprises: The preprocessing process comprises spatial alignment, feature extraction, dimension unification and feature standardization. 3.The geological survey method based on big data according to claim 1, wherein, The S1 specifically comprises: In the implementation process of the data source credibility weighting mechanism, the original credibility scores of the data sources are introduced to weight and sum the basic feature vectors of the different data sources to obtain the fusion feature vectors of the sub-regions. 4.The geological survey method based on big data according to claim 1, wherein, The S2 specifically comprises: The sub-regions are taken as nodes, and the fusion feature vectors of the sub-regions are taken as node feature vectors to construct a geological structure graph; in the implementation process of the edge weight construction mechanism combining spatial adjacency relationship and feature similarity, the spatial adjacency relationship and the feature similarity of the nodes are combined, and a linear transformation coefficient matrix is introduced to obtain edge weights between the sub-regions. 5.The geological survey method based on big data according to claim 1, wherein, The S2 specifically comprises: In the implementation process of the directional and edge weight joint modulation mechanism, a direction mapping matrix is introduced, and the center coordinates of the sub-regions are combined to calculate a directional modulation factor.

6. The method of geological investigation based on big data according to claim 5, characterized in that, The S2 specifically comprises: In the implementation process of the directional and edge weight joint modulation mechanism, the hidden features of the nodes are initialized as the node feature vectors, and based on the directional modulation factor, the hidden features of the adjacent sub-region nodes are directionally weighted element by element to obtain the hidden features of the adjacent sub-region nodes after directional modulation. 7.The geological survey method based on big data according to claim 6, wherein, The S2 specifically comprises: In the implementation process of the directional and edge weight joint modulation mechanism, based on the edge weights between the sub-regions, the hidden features of the adjacent sub-region nodes after directional modulation are weighted, and a node state self-updating matrix is introduced to update the hidden features of the sub-region nodes themselves to obtain the output features of the sub-regions. 8.The geological survey method based on big data according to claim 7, wherein, The S2 specifically comprises: Based on the output features of the sub-regions, a linear classifier is used to predict a geological structure category of the sub-regions.

Images