Geological exploration drilling quality detection method and system based on image technology

Abstract

The invention relates to the technical field of geological quality detection, and discloses a geological exploration borehole quality detection method and system based on an image technology, and the method comprises the steps: collecting a high-definition image of a geological exploration borehole, carrying out the feature progressive extraction of the high-definition image, and reconstructing the high-definition image, and obtaining a preliminary reconstruction image; extracting multi-scale features of the preliminary reconstructed image, carrying out wavelet frequency domain feature coding to obtain joint coding features, and carrying out progressive image decoding on the joint coding features to obtain a target reconstructed image; identifying an image structure, performing multi-agent deployment, identifying region information of the target reconstructed image, and performing region segmentation on the target reconstructed image to obtain a segmented image; carrying out binarization processing on the segmented image, constructing a persistent graph, identifying topological characteristics of the persistent graph, carrying out defect identification on the geological exploration drilling hole, and evaluating the quality of the geological exploration drilling hole based on an identification result of defect identification. According to the invention, the detection accuracy of the geological exploration drilling quality can be improved.

Description

Technical Field

[0001] The present invention relates to a geological exploration borehole quality detection method and system based on image technology, belonging to the technical field of geological quality detection. Background Art

[0002] Drillhole quality testing plays a key role in ensuring the accuracy and reliability of geological exploration work. Accurate drillhole quality testing ensures the authenticity and validity of acquired geological data, providing a solid foundation for subsequent resource assessment and engineering construction. Accurate drillhole quality testing can effectively prevent serious consequences such as resource misjudgment and engineering risks caused by drilling problems.

[0003] Currently, traditional physical measurement and manual inspection techniques are commonly used to inspect the quality of geological exploration boreholes. Traditional physical measurement primarily uses simple measuring tools to measure basic parameters such as borehole depth and diameter. Manual inspection relies on experienced personnel visually observing the borehole shape and core samples. However, these methods have many limitations. Traditional physical measurement makes it difficult to fully inspect the complex internal structure and subtle defects of the borehole. Manual inspection is limited by the subjective judgment and fatigue of the personnel. Faced with a large number of borehole inspection tasks, it is prone to omissions and misjudgments, resulting in insufficient accuracy in geological exploration borehole quality inspection. Summary of the Invention

[0004] The present invention provides a method and system for detecting the quality of geological exploration boreholes based on image technology, the main purpose of which is to improve the detection accuracy of geological exploration borehole quality.

[0005] To achieve the above-mentioned purpose, the present invention provides a method for detecting the quality of geological exploration boreholes based on image technology, comprising:

[0006] collecting high-definition images of geological exploration boreholes, constructing a feature pyramid for the high-definition images, progressively extracting features from the high-definition images using the feature pyramid to obtain multi-scale features, and performing high-level semantic reconstruction on the high-definition images based on the multi-scale features to obtain a preliminary reconstructed image;

[0007] Extracting multi-scale features of the preliminary reconstructed image, and jointly encoding the multi-scale features to obtain potential coding features, performing wavelet frequency domain feature encoding on the preliminary reconstructed image to obtain wavelet coding features, performing coding fusion on the potential coding features and the wavelet coding features to obtain joint coding features, performing progressive image decoding on the joint coding features to obtain a sub-band image, and performing inverse encoding on the sub-band image to obtain a target reconstructed image;

[0008] Identifying an image structure of the target reconstructed image, deploying multiple agents in the target reconstructed image based on the image structure to obtain deployed agents, identifying regional information of the target reconstructed image using the deployed agents, and performing regional segmentation on the target reconstructed image based on the regional information to obtain a segmented image;

[0009] The segmented image is binarized to obtain a binary image, a simplicial complex of the binary image is constructed, and a homology group of the simplicial complex is identified. Based on the homology group, a persistence graph of the binary image is constructed, and topological features of the persistence graph are identified. Defects of the geological exploration borehole are identified using the topological features, and the quality of the geological exploration borehole is evaluated based on the defect identification results.

[0010] Optionally, the step of progressively extracting features from the high-definition image using the feature pyramid to obtain multi-scale features includes:

[0011] The feature pyramid is layered up and down to obtain a layered feature pyramid;

[0012] Performing hierarchical feature detection operator deployment on the hierarchical feature pyramid to obtain a first deployment operator and a second deployment operator;

[0013] Using the first deployment operator to perform Gaussian difference processing on the corresponding image of the lower layer of the hierarchical feature pyramid to obtain a differential image;

[0014] Identifying image extreme points of the differential image and calculating the gradient histogram of the image extreme points to obtain lower-layer features;

[0015] Using the second deployment operator to perform unit division on the upper layer corresponding image of the hierarchical feature pyramid to obtain a unit image;

[0016] Calculating the gradient histogram of the unit image to obtain upper-layer features;

[0017] The lower layer features and the upper layer features are subjected to feature fusion, and the fused features are subjected to feature screening to obtain multi-scale features.

[0018] Optionally, performing high-level semantic reconstruction on the high-definition image based on the multi-scale features to obtain a preliminary reconstructed image includes:

[0019] Performing high-dimensional mapping on the multi-scale features to obtain high-dimensional multi-scale features;

[0020] Performing self-attention encoding on the high-dimensional multi-scale features to obtain encoded multi-scale features;

[0021] Constructing a semantic graph encoding multi-scale features;

[0022] Performing relationship mining on the encoded multi-scale features in the semantic graph to construct a relationship model of the encoded multi-scale features;

[0023] Based on the relationship model, image reconstruction is performed on the encoded multi-scale features to obtain a preliminary reconstructed image.

[0024] Optionally, performing wavelet frequency domain feature encoding on the preliminary reconstructed image to obtain wavelet coding features includes:

[0025] performing detail enhancement processing on the preliminary reconstructed image to obtain a preliminary processed image;

[0026] identifying image features of the preliminarily processed image to configure a wavelet basis function of the preliminarily processed image;

[0027] Performing multi-level sub-band decomposition on the preliminary processed image using the wavelet basis function to obtain multi-level image sub-bands;

[0028] Calculating statistical features, energy features, and texture features of each sub-band of the multi-level image to obtain multi-level quantized features;

[0029] Adaptively quantize the multi-level image subbands to obtain quantized features

[0030] Zerotree coding is performed on the quantized features to obtain wavelet coding features.

[0031] Optionally, the using the deployed intelligent agent to identify region information of the target reconstructed image includes:

[0032] Constructing a collaboration mechanism for the deployment agents and formulating a communication protocol for the deployment agents to obtain a communication agent;

[0033] Using the communication agent to perform multimodal perception on the target reconstructed image to obtain multimodal data;

[0034] querying a graph structure of the target reconstructed image, and performing regional modeling on the graph structure using the multimodal data to obtain a regional graph model;

[0035] The communication agent is used to perform integrated learning on the region graph model to obtain region information of the target reconstructed image.

[0036] Optionally, performing region segmentation on the target reconstructed image based on the region information to obtain a segmented image includes:

[0037] Segmenting the target reconstructed image into superpixels based on the region information;

[0038] Querying pixel features of the superpixel;

[0039] Calculating edge weights of the target reconstructed image based on the pixel features;

[0040] Based on the edge weights, construct the region labels of the target reconstructed image

[0041] Perform preliminary segmentation on the target reconstructed image based on the region label to obtain a preliminary segmented image

[0042] Inputting the preliminary segmented image and the target reconstructed image into a preconfigured graph learning model to identify an optimized segmentation boundary of the target reconstructed image;

[0043] Segmentation optimization is performed on the preliminary segmented image based on the optimized segmentation boundary to obtain a segmented image.

[0044] Optionally, calculating the edge weight of the target reconstructed image based on the pixel features includes:

[0045] Querying the color feature, texture feature, and spatial distance in the pixel feature to obtain a query feature;

[0046] Quantifying the query feature to obtain a quantitative feature;

[0047] Based on the quantitative features, edge weights of the target reconstructed image are calculated.

[0048] Optionally, identifying the topological features of the persistent graph includes:

[0049] Gridding the persistence graph to obtain a grid graph;

[0050] Identifying homology class points in the grid graph and calculating the persistence interval of the homology class points;

[0051] plotting a histogram of the persistence interval;

[0052] Identifying interval features of the persistence graph based on the histogram, and counting point densities of homology points in different grid intervals in the grid graph;

[0053] A topological feature of the persistence graph is identified based on the interval feature and the point density.

[0054] Optionally, the evaluating the quality of the geological exploration borehole based on the defect identification result includes:

[0055] Determining the defect category of the geological exploration borehole based on the defect identification result;

[0056] Analyzing the importance coefficients of different categories in the defect category;

[0057] A quality assessment value of the geological exploration borehole is calculated based on the defect category and the importance coefficient.

[0058] In order to solve the above problems, the present invention also provides a geological exploration drilling quality detection system based on image technology, the system comprising:

[0059] a preliminary image reconstruction module for collecting high-definition images of geological exploration boreholes, constructing a feature pyramid for the high-definition images, performing progressive feature extraction on the high-definition images using the feature pyramid to obtain multi-scale features, and performing high-level semantic reconstruction on the high-definition images based on the multi-scale features to obtain a preliminary reconstructed image;

[0060] an image secondary reconstruction module, configured to extract multi-scale features of the preliminary reconstructed image, perform joint encoding on the multi-scale features to obtain potential coding features, perform wavelet frequency domain feature encoding on the preliminary reconstructed image to obtain wavelet coding features, perform coding fusion on the potential coding features and the wavelet coding features to obtain joint coding features, perform progressive image decoding on the joint coding features to obtain a sub-band image, and perform inverse encoding on the sub-band image to obtain a target reconstructed image;

[0061] an image segmentation module, configured to identify an image structure of the target reconstructed image, deploy multiple agents in the target reconstructed image based on the image structure to obtain deployed agents, use the deployed agents to identify regional information of the target reconstructed image, and perform regional segmentation on the target reconstructed image based on the regional information to obtain a segmented image;

[0062] A quality assessment module is configured to binarize the segmented image to obtain a binary image, construct a simplicial complex of the binary image, identify a homology group of the simplicial complex, construct a persistence graph of the binary image based on the homology group, identify topological features of the persistence graph, perform defect identification on the geological exploration borehole using the topological features, and evaluate the quality of the geological exploration borehole based on the defect identification results.

[0063] Compared with the problems described in the background technology, the embodiments of the present invention collect high-definition images of geological exploration boreholes, and then extract multi-scale features of the high-definition images to perform high-dimensional mapping, self-attention coding, construct semantic maps, mine relationships, and image reconstruction, thereby restoring image details and structures to make the image closer to the actual geological conditions, providing a basis for subsequent analysis; further, the present invention extracts multi-scale features of the preliminary reconstructed image and jointly encodes them to obtain potential coding features, and performs wavelet frequency domain feature coding on the preliminary reconstructed image to obtain wavelet coding features, and then fuses the two features to obtain fused features that can integrate multi-scale features, capture image frequency information, and combine the advantages of different coding to obtain high-quality image data, which is convenient for subsequent identification and exploration; further, the present invention It is clear that by utilizing an intelligent agent to identify information in an image and then performing image segmentation, the speed and accuracy of image analysis can be improved, and different elements of the image can be separated for easy individual analysis and processing; further, the present invention binarizes the segmented image to obtain a binary image, constructs its simplicial complex and identifies the homology group, constructs a persistence graph based on the homology group, and identifies the topological features of the persistence graph, which can convert the image geometric structure into a topological object, capture the stability and changes of the topological features, and provide a basis for defect identification; further, the present invention utilizes topological features to perform defect identification on geological exploration boreholes to understand whether there are defects, their types, and locations in the borehole image, and finally uses the formula to calculate the quality assessment value to comprehensively evaluate the borehole quality, thereby improving the detection accuracy of the geological exploration borehole quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 A flowchart of a method for detecting the quality of geological exploration drilling holes based on image technology provided by one embodiment of the present invention;

[0065] Figure 2 A schematic diagram of modules for implementing a geological exploration drilling quality detection method based on image technology provided in one embodiment of the present invention.

[0066] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0067] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0068] The embodiments of the present application provide a method for detecting the quality of geological exploration drilling holes based on image technology. The execution subject of the method includes, but is not limited to, at least one of electronic devices such as a server or a terminal that can be configured to execute the method provided in the embodiments of the present application. In other words, the method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0069] Example 1:

[0070] Reference Figure 1 FIG. 1 is a flow chart of a method for detecting the quality of geological exploration boreholes based on image technology according to an embodiment of the present invention. In this embodiment, the method for detecting the quality of geological exploration boreholes based on image technology includes:

[0071] S1. Collect high-definition images of geological exploration boreholes, construct a feature pyramid for the high-definition images, use the feature pyramid to progressively extract features from the high-definition images to obtain multi-scale features, and perform high-level semantic reconstruction on the high-definition images based on the multi-scale features to obtain a preliminary reconstructed image.

[0072] The embodiment of the present invention can understand the underground geological conditions, evaluate the resource quality, and provide a basic analysis basis for subsequent geological exploration drilling quality analysis by collecting high-definition images of geological exploration boreholes.

[0073] The geological exploration drilling hole refers to a cylindrical hole with a certain depth and diameter formed by drilling underground at the ground or other specific locations using professional drilling equipment during the geological exploration process.

[0074] Optionally, the high-definition image can be collected by a high-definition camera.

[0075] Furthermore, embodiments of the present invention facilitate analysis and processing of features of high-definition images at different scales by constructing a feature pyramid for the high-definition image, thereby obtaining more accurate image information. The feature pyramid refers to a data structure that represents an image at multiple scales.

[0076] Optionally, the feature pyramid may be constructed by a Gaussian pyramid.

[0077] Furthermore, the embodiment of the present invention uses the feature pyramid to progressively extract features from the high-definition image to obtain multi-scale features, which can gradually extract image features from different levels, thereby obtaining richer geological information.

[0078] As an embodiment of the present invention, the method of progressively extracting features from the high-definition image using the feature pyramid to obtain multi-scale features includes: layering the feature pyramid up and down to obtain a layered feature pyramid, performing hierarchical feature detection operator deployment on the layered feature pyramid to obtain a first deployment operator and a second deployment operator, using the first deployment operator to perform Gaussian difference processing on the lower layer corresponding image of the layered feature pyramid to obtain a differential image, identifying image extreme points of the differential image, and calculating the gradient histogram of the image extreme points to obtain lower-layer features, using the second deployment operator to perform unit division on the upper layer corresponding image of the layered feature pyramid to obtain a unit image, calculating the gradient histogram of the unit image to obtain upper-layer features, performing feature fusion on the lower-layer features and the upper-layer features, and performing feature screening on the fused features to obtain multi-scale features.

[0079] Among them, the feature detection operator refers to a mathematical algorithm or function used to extract specific features in an image, such as the Sobel operator, the Harris corner detection operator, etc., and the image extreme point refers to the point where the pixel value has the maximum or minimum value in a local area of the image.

[0080] Optionally, the hierarchical feature pyramid can be divided by analyzing the distribution of image details in the feature pyramid. For example, if the pyramid has five layers, the bottom 1 to 2 layers with multiple image details can be divided into the lower layer, and the 3 to 5 layers with image details can be divided into the upper layer. The specific setting needs to be based on the actual application. The first deployment operator can be obtained by deploying the SIFT operator on the lower layer of the hierarchical feature pyramid, and the second deployment operator can be obtained by deploying the HOG operator on the upper layer of the hierarchical feature pyramid. The image extreme points can be obtained by comparing each pixel in the difference image with the pixels in its neighborhood to find points with values greater or less than those of all surrounding pixels. The gradient histogram can be calculated using a gradient histogram algorithm. The unit image can be obtained by dividing the upper layer image into multiple non-overlapping small units, for example, by dividing the image into small squares of a certain pixel size (e.g., 8x8 pixels). The multi-scale features can be obtained by fusing the lower layer features and the upper layer features using a splicing method to obtain preliminary fused features, which are then reduced in dimension using principal component analysis.

[0081] The embodiment of the present invention performs high-level semantic reconstruction on the high-definition image based on the multi-scale features to obtain a preliminary reconstructed image that can restore the details and structure of the image, while incorporating semantic information extracted from the multi-scale features, making the preliminary reconstructed image closer to the actual geological situation.

[0082] As an embodiment of the present invention, the high-level semantic reconstruction of the high-definition image based on the multi-scale features to obtain a preliminary reconstructed image includes: performing high-dimensional mapping on the multi-scale features to obtain high-dimensional multi-scale features, performing self-attention encoding on the high-dimensional multi-scale features to obtain encoded multi-scale features, constructing a semantic graph of the encoded multi-scale features, performing relationship mining on the encoded multi-scale features in the semantic graph to construct a relationship model of the encoded multi-scale features, and reconstructing the encoded multi-scale features based on the relationship model to obtain a preliminary reconstructed image.

[0083] Among them, the semantic graph refers to a data structure that represents semantic information in the form of a graph. It uses various semantic elements in the image (such as objects, scenes, attributes, etc.) as nodes, and the semantic relationships between nodes (such as spatial relationships, belonging relationships, causal relationships, etc.) as edges. The relationship model refers to a mathematical model or computational framework used to describe and analyze the relationship between nodes in the semantic graph, which can be constructed through probabilistic graph models such as Bayesian networks, Markov random fields, etc.

[0084] Optionally, the high-dimensional multi-scale features can be obtained by mapping the multi-scale features to a multi-layer perceptron network. The encoded multi-scale features can be obtained by encoding the high-dimensional multi-scale features using the self-attention mechanism module in the Transformer. The semantic graph can be constructed by treating each semantic element in the encoded multi-scale features as a node, and then determining the connection relationship of the edges by calculating indicators such as semantic similarity and correlation between the nodes. The relationship mining of the encoded multi-scale features can be carried out using a graph neural network. The preliminary reconstructed image can be obtained by using a generative adversarial network to take the relationship model and the encoded multi-scale features as input conditions to guide the generator through operations such as convolution and deconvolution.

[0085] S2. Extract multi-scale features of the preliminary reconstructed image, and jointly encode the multi-scale features to obtain potential coding features, perform wavelet frequency domain feature encoding on the preliminary reconstructed image to obtain wavelet coding features, perform coding fusion on the potential coding features and the wavelet coding features to obtain joint coding features, perform progressive image decoding on the joint coding features to obtain a sub-band image, and inversely encode the sub-band image to obtain a target reconstructed image.

[0086] The embodiment of the present invention extracts multi-scale features from the preliminary reconstructed image and jointly encodes the multi-scale features to obtain potential coding features, which can integrate and encode the multi-scale features and convert them into a more compact and abstract representation, thereby reducing the dimension of the data while retaining important information and improving information processing efficiency.

[0087] Optionally, the latent coding features can be obtained by jointly encoding multi-scale features through a variational autoencoder.

[0088] In the embodiment of the present invention, the wavelet frequency domain feature encoding is performed on the preliminary reconstructed image to obtain wavelet coding features that can capture the frequency information of the image, so as to better handle the problems of image noise, compression and enhancement.

[0089] As an embodiment of the present invention, the wavelet frequency domain feature encoding of the preliminary reconstructed image to obtain wavelet coding features includes: performing detail enhancement processing on the preliminary reconstructed image to obtain a preliminary processed image, identifying image features of the preliminary processed image to configure a wavelet basis function of the preliminary processed image, performing multi-level sub-band decomposition on the preliminary processed image using the wavelet basis function to obtain multi-level image sub-bands, calculating statistical features, energy features and texture features of each sub-band of the multi-level image sub-bands to obtain multi-level quantization features, performing adaptive quantization on the multi-level image sub-bands to obtain quantization features, and performing zero-tree coding on the quantization features to obtain wavelet coding features.

[0090] Among them, the wavelet basis function refers to a function with special properties, which is used to perform multi-resolution analysis and representation of data such as signals or images. The zerotree coding refers to an efficient coding technology for image and signal compression. The main method is to organize and encode the wavelet coefficients based on the characteristics of the coefficients after wavelet transformation, that is, the wavelet coefficients have correlation at different scales and positions.

[0091] Optionally, the preliminary processed image can be obtained by performing RGB multi-channel processing on the preliminary reconstructed image and then performing normalization processing. The image features can be identified by a convolution model, and the wavelet basis functions can be configured by Daubechies (dbN), Symlets (symN) and Coiflets (coifN). The specific configuration needs to be combined with actual applications and is performed by analyzing the smoothness, texture and compression characteristics of the image. The statistical features can be obtained by calculating the mean, variance, skewness and kurtosis of each sub-band; the energy features can be obtained by calculating the sum of the squares of all wavelet coefficients in the sub-band; and the texture features can be calculated using the gray-level co-occurrence matrix.

[0092] The embodiment of the present invention obtains a joint coding feature by encoding and fusing the latent coding feature and the wavelet coding feature, which can combine the advantages of two different types of coding features to obtain a more powerful feature representation, thereby being able to more comprehensively describe the content and structure of the image.

[0093] Optionally, the joint coding feature can be obtained by a weighted fusion method, by assigning different weights to the latent coding feature and the wavelet coding feature according to their importance in a specific task, and then adding the weighted features.

[0094] Furthermore, in the embodiment of the present invention, by performing progressive image decoding on the joint coding feature, sub-band images are obtained, which can gradually restore image information.

[0095] Optionally, the progressive image decoding of the joint coding feature to obtain the sub-band image can be achieved through Hierarchical Set Partitioning Trees (SPIHT) decoding technology.

[0096] Furthermore, the embodiment of the present invention can obtain high-quality image data by inversely encoding the sub-band image to obtain the target reconstructed image, which is convenient for subsequent image recognition and geological exploration tasks.

[0097] Optionally, the inverse encoding of the sub-band image to obtain the target reconstructed image can be achieved by using an inverse wavelet transform technique.

[0098] S3. Identify the image structure of the target reconstructed image, deploy multiple agents in the target reconstructed image based on the image structure to obtain deployed agents, use the deployed agents to identify regional information of the target reconstructed image, and segment the target reconstructed image based on the regional information to obtain a segmented image.

[0099] The embodiment of the present invention can help users understand the overall layout, texture, edge and other feature information of the image by identifying the image structure of the target reconstructed image, and further determine the importance and relevance of different parts in the image.

[0100] The image structure refers to the organization and mutual relationship of various elements in the image, which is an abstract description of the image content, such as geometric shape, hierarchical structure, etc.

[0101] Optionally, the image structure may be identified using spectral clustering algorithms.

[0102] Furthermore, the embodiment of the present invention deploys multiple agents in the target reconstructed image based on the image structure, so that the deployed agents can fully utilize the parallel processing capabilities of the agents, improve the speed and accuracy of image analysis, and complete complex image recognition and analysis tasks.

[0103] Optionally, the deployed intelligent agent can first clarify the key parts of the image through the image structure, and then deploy appropriate intelligent agents in the key parts. For example, in the area representing the core object in the image, if the object is the focus of the recognition task, an intelligent agent with high-precision recognition capabilities can be deployed. For example, in the object boundary or transition area in the image, an intelligent agent focusing on edge detection and information fusion can be deployed. For example, in the area with high dynamic changes or prone to interference in the image, an intelligent agent with adaptive processing capabilities can be deployed.

[0104] The embodiment of the present invention can provide an important basis for understanding and interpreting the image by utilizing the deployed intelligent agent to identify the regional information of the target reconstructed image, for example, identifying the object area, background area, and categories and boundaries of different objects in the image.

[0105] As an embodiment of the present invention, the use of the deployed intelligent agent to identify the regional information of the target reconstructed image includes: constructing a collaborative mechanism of the deployed intelligent agent, and formulating a communication protocol for the deployed intelligent agent to obtain a communication intelligent agent, using the communication intelligent agent to perform multimodal perception on the target reconstructed image to obtain multimodal data, querying the graph structure of the target reconstructed image, and using the multimodal data to perform regional modeling on the graph structure to obtain a regional graph model, and using the communication intelligent agent to perform integrated learning on the regional graph model to obtain the regional information of the target reconstructed image.

[0106] Among them, the collaboration mechanism refers to the working method and interaction mode used to coordinate multiple intelligent agents in the context of using deployed intelligent agents to identify targets and reconstruct image area information, and the communication protocol refers to the rules and standards followed when information is transmitted and interacted between intelligent agents.

[0107] Optionally, the collaboration mechanism can be designed based on the capabilities and task characteristics of the agent through a distributed consensus algorithm to design a reasonable task allocation and collaboration strategy, such as adopting a master-slave collaboration, designating some agents as the leaders, responsible for coordinating tasks, and the others as subordinates. The communication protocol can be formulated through the message passing interface (MPI). The regional graph model can be obtained by extracting features related to the nodes and edges of the graph structure from multimodal data, such as the color, texture features of the nodes and the connection strength of the edges, and then using a graph neural network combined with the extracted features for modeling. The regional information can be obtained by using a learner in the communicating agent, such as a convolutional neural network or a recurrent neural network, to perform image learning on the regional graph model.

[0108] In the embodiment of the present invention, by performing region segmentation on the target reconstructed image based on the region information, the segmented image obtained can separate different objects or scene elements in the image, thereby facilitating separate analysis and processing of each region.

[0109] As an embodiment of the present invention, the target reconstructed image is subjected to regional segmentation based on the regional information to obtain a segmented image, including: segmenting the target reconstructed image into superpixels based on the regional information, querying pixel features of the superpixels, calculating edge weights of the target reconstructed image based on the pixel features, constructing regional labels of the target reconstructed image based on the edge weights, performing preliminary segmentation on the target reconstructed image based on the regional labels to obtain a preliminary segmented image, inputting the preliminary segmented image and the target reconstructed image into a preconfigured graph learning model to identify an optimized segmentation boundary of the target reconstructed image, and performing segmentation optimization on the preliminary segmented image based on the optimized segmentation boundary to obtain a segmented image.

[0110] Among them, the superpixel refers to a set of pixels with certain semantic information formed by clustering adjacent pixels with similar features (such as color, texture, brightness, etc.) in the image, and the edge weight refers to a quantitative value used to measure a certain attribute or relationship of the edge connecting two superpixels.

[0111] Optionally, the superpixels can be obtained by segmenting the target reconstructed image using a linear iterative clustering algorithm based on regional information. The region labels can be obtained by applying a spectral clustering algorithm, taking edge weights as input, analyzing the connection relationship between superpixels, classifying closely connected superpixels into one category, and assigning corresponding region labels. The preliminary segmented image can be obtained by merging superpixels with the same region label in the image and preliminarily dividing different regions. The optimized segmentation boundary can input the preliminary segmentation image and the target reconstructed image into a pre-trained graph learning model, and output a more accurate optimized segmentation boundary in combination with a graph normalization segmentation algorithm.

[0112] Furthermore, as another optional embodiment of the present invention, the calculating the edge weight of the target reconstructed image based on the pixel features includes: querying the color features, texture features, and spatial distances in the pixel features to obtain query features, quantizing the query features to obtain quantized features, and calculating the edge weight of the target reconstructed image based on the quantized features using the following formula:

[0113]

[0114] Among them, ω represents the edge weight, C a Represents the color vector of superpixel a in the quantized feature, C b Represents the color vector of the superpixel point b in the quantized feature, T a Represents the texture feature of super pixel a in the quantized feature, T b Represents the texture feature of the super pixel b in the quantized feature, d abrepresents the spatial distance between a and b, a represents the color feature weight coefficient, β represents the texture feature weight coefficient, γ represents the spatial distance weight coefficient, σ c Represents the similarity attenuation coefficient of color features, σ T Represents the similarity attenuation coefficient of texture features, σ d Represents the similarity decay coefficient of spatial distance.

[0115] It should be further explained that the color feature weight coefficient ranges from 0.3 to 0.5. In geological exploration borehole images, color can reflect important information such as rock composition and weathering. For example, fresh and weathered rock can have distinct color differences, and different mineral compositions can also appear different colors. However, color information can be affected by factors such as lighting, so it should not be overly weighted. A value of 0.3 to 0.5 allows for effective use of color information to determine rock properties of the borehole wall while comprehensively considering other features. The feature weight coefficient can be set by first statistically analyzing the color distribution of a large number of geological borehole images to determine the importance of color features in distinguishing different geological conditions. The texture feature weight coefficient ranges from 0.4 to 0.6. Texture in geological borehole images can reflect rock structure and cracks, and is crucial for assessing borehole quality and geological structure. For example, rock texture features such as bedding and cracks can help identify potential geological risks and borehole stability. Therefore, a relatively high weight for texture features, such as 0.4 to 0.6, highlights the importance of texture in quality inspection. The spatial distance weight coefficient ranges from 0.1 to 0.3. In geological exploration borehole quality testing, spatial distance is primarily used to consider the relationship between different locations within the borehole and its association with the surrounding geological environment. For example, it can be used to determine changes in rock characteristics at different depths within the borehole and the distance between the borehole and surrounding known geological structures. However, compared to color and texture features, the direct impact of spatial distance is relatively small, so a lower value, 0.1 to 0.3, can reasonably reflect the role of spatial factors in comprehensive assessments. The spatial distance weight coefficient can be determined based on the regional characteristics of the geological exploration and the distribution of boreholes. For example, if the boreholes are densely distributed and the mutual influence between boreholes needs to be considered, a value of 0.2 can be used. The similarity attenuation coefficient of the color feature ranges from 0.3 to 0.8. A smaller value means that color differences have a relatively small impact on similarity. The degree of color variation can be understood by performing statistical analysis such as standard deviation on the color data of a large number of geological borehole images. If the color variation is relatively uniform, a value of 0.6 or 0.7 can be used. The similarity attenuation coefficient of the texture feature is between 0.5 and 1.2, and can be based on the statistical analysis of the texture features of the geological image, such as calculating the changes in the texture roughness, directionality and other features. If the texture feature changes significantly and has a great impact on the quality detection, the value is larger, such as 1.0. If the texture feature changes significantly and has a small impact on the quality detection, the value is smaller, such as 0.6. The similarity attenuation coefficient of the spatial distance is between 0.6 and 1.5. When the value is larger, it means that the spatial distance has a greater impact on the similarity. The similarity attenuation coefficient of the spatial distance can be determined according to the structural characteristics of the geological area and the spatial distribution of the boreholes. For areas with complex geological structures and large spatial changes, it can be 1.3 or 1.4. For areas with simple geological structures and small spatial changes, it can be 0.8 to 0.9.

[0116] S4. Binarize the segmented image to obtain a binary image, construct a simplicial complex of the binary image, and identify the homology group of the simplicial complex. Based on the homology group, construct a persistence graph of the binary image, identify the topological features of the persistence graph, use the topological features to identify defects in the geological exploration borehole, and evaluate the quality of the geological exploration borehole based on the defect identification results.

[0117] The embodiment of the present invention performs binarization processing on the segmented image to obtain a binary image, which can divide pixels in the image into two categories: foreground and background, highlight key information in the image, and remove unnecessary grayscale change details.

[0118] Optionally, the binary image can be obtained by processing the segmented image using a grayscale histogram method.

[0119] The embodiment of the present invention can convert the geometric structure of the image into a mathematical object in topology by constructing the simplicial complex of the binary image and identifying the homology group of the simplicial complex, which makes it easier to mine the potential topological structure information in the image and then analyze the image using topological methods and theories.

[0120] Among them, the simplicial complex refers to a topological space composed of some simple geometric objects "simplices" according to certain rules, and the homology group refers to a certain algebraic property used to describe the topological space. It is an algebraic structure obtained by algebraically processing the topological space.

[0121] Optionally, the simplicial complex can regard the foreground pixels in the binary image as 0-dimensional simplexes, the connecting edges of adjacent foreground pixels as 1-dimensional simplexes, and the triangular area formed by three adjacent foreground pixels as 2-dimensional simplexes, and these simplices are combined according to position and connection relationships to construct a simplicial complex structure; the homology group can first define the chain group generated by the simplicies of each dimension, then determine the edge operator, and map the high-dimensional chain to a low-dimensional chain, and then calculate the edge operator kernel to obtain the closed chain group, and finally obtain the homology group by the closed chain group modulo the edge chain group.

[0122] Furthermore, the embodiment of the present invention can capture the stability and changes of the topological features of the binary image at different scales by constructing the persistence graph of the binary image based on the homology group.

[0123] The persistence graph is an important tool for visualizing and analyzing the persistence of topological features in topological data analysis.

[0124] Optionally, the persistence map can first sort the binary image according to features such as pixel value and distance from a reference point, and add simplices to the simplicial complex in sequence as the feature value increases to form a filter sequence. Then, the homology group is calculated for each simplicial complex in the filter sequence, and the birth value generated and the death value disappeared by the homology class are recorded. The difference between the two is the persistence. Finally, a two-dimensional coordinate map is constructed with the birth value as the horizontal coordinate and the death value as the vertical coordinate.

[0125] The embodiment of the present invention can convert the topological information of the image into a feature vector that can be used for analysis and comparison by identifying the topological features of the persistence graph, thereby providing a specific basis for subsequent defect identification.

[0126] As an embodiment of the present invention, the identifying of the topological features of the persistence graph includes: gridding the persistence graph to obtain a grid graph, identifying homology class points in the grid graph, and calculating the persistence intervals of the homology class points, drawing a histogram of the persistence intervals, identifying the interval features of the persistence graph based on the histogram, counting the point density of the homology class points in different grid intervals in the grid graph, and identifying the topological features of the persistence graph based on the interval features and the point density.

[0127] Homology points are points in the persistence graph that correspond to a homology class of a simplicial complex at different filtering stages. These points are called homology points. The persistence interval measures the stability and importance of the topological feature represented by the homology class. A larger persistence interval indicates that the topological feature persists longer during the filtering process and has a greater impact on the overall topological structure. For example, in the persistence graph analysis of geological exploration borehole data, a homology point with a larger persistence interval may indicate that the corresponding geological structure feature is stable and significant.

[0128] Optionally, the grid map can be obtained by constructing a regular grid in the persistence map and dividing the persistence map into multiple small grid areas, such as a 10×10 grid. The persistence interval is calculated for each homology class point by subtracting its horizontal coordinate (birth value) from its vertical coordinate (death value). The interval characteristics can be obtained by observing the histogram peak and distribution shape to determine the main persistence interval range and distribution concentration. The point density can be obtained by dividing the homology class points in each grid interval by the grid area. The topological characteristics can be obtained by comprehensively analyzing the interval characteristics and point density to judge the stability of the topological characteristics in the persistence map, the density of the distribution, etc.

[0129] In the embodiment of the present invention, by using the topological features to perform defect recognition on the geological exploration borehole, it is possible to determine whether there are defects in the geological exploration borehole image and the type and location of the defects.

[0130] Optionally, the use of the topological features to identify defects in the geological exploration borehole can be achieved by collecting historical defect identification images of the geological exploration borehole, inputting the historical defect identification images and the topological features into a preconfigured convolution model, using the preconfigured convolution model to extract the historical topological features of the historical defect identification images, and then performing feature matching and identification with the topological features.

[0131] The embodiment of the present invention evaluates the quality of the geological exploration borehole by using the recognition result based on the defect recognition. The quality of the geological exploration borehole can be comprehensively evaluated according to the result of the defect recognition.

[0132] As an embodiment of the present invention, the evaluating the quality of the geological exploration borehole based on the recognition result of the defect recognition includes: determining the defect category of the geological exploration borehole based on the recognition result of the defect recognition, analyzing the importance coefficients of different categories in the defect category, and calculating the quality evaluation value of the geological exploration borehole based on the defect category and the importance coefficient using the following formula:

[0133]

[0134] Among them, Q represents the quality assessment value, m represents the number of defect categories, and R i represents the importance coefficient of the defect in the i-th position, f i Indicates the frequency factor of the i-th defect, n i represents the number of defects of type i, s ij represents the severity score of the jth defect in the i-th defect, d ij represents the distribution influence factor of the jth defect in the i-th defect, and I represents the exploration result impact index;

[0135] Based on the quality assessment value, the quality of the geological exploration borehole is determined.

[0136] It should be further explained that the importance coefficient reflects the relative importance of the i-th defect to the quality of geological exploration drilling, and its value is between 0 and 1. It is generally determined by geological experts based on geological exploration experience and relevant standards and specifications, combined with specific geological conditions and exploration goals. It can also be determined by analyzing a large amount of historical drilling data and the corresponding quality conditions, and using mathematical methods such as hierarchical analysis method and fuzzy comprehensive evaluation method to calculate the importance ranking of each defect. The frequency factor represents the frequency of occurrence of the i-th defect in geological exploration drilling. If a defect often appears in multiple boreholes, f i The value of f will be higher. For example, under certain geological conditions, the probability of drilling shrinkage defects is higher, and f i It may be between 0.6 and 0.8; and some rare defects, fi The number of defects is a non-negative integer, which directly represents the number of defects of type i in the current geological exploration borehole. For example, if three cracks are found in a borehole, the number of defects of type i is n. i =3. The larger the value of the severity score, the higher the severity of the defect. For example, for a crack defect, if the crack width is small and the length is short, the severity score may be s ij The severity score is 3 to 5. If the crack width is large and runs through the entire borehole wall, the severity score may be 8 to 10. It can be determined based on the relevant geological exploration standards and specifications, combined with the specific characteristics of the defects in the image. For example, for crack defects, corresponding scoring standards can be formulated based on the geometric characteristics of the crack width, length, depth, and whether there is an expansion trend. A severity assessment model is trained using a machine learning algorithm to automatically score the identified defects. The distribution influence factor is used to measure the degree of influence of the distribution position of the jth defect in the borehole on the drilling quality. If the defect is distributed in the key part of the borehole, such as near the hole mouth or in the area where special treatment is expected, d ij The value is between 0.8 and 1; if the defects are distributed in relatively unimportant locations, d ij The value is between 0.2 and 0.5. The distribution impact factor can determine the importance of different areas of the borehole to the quality based on the design requirements of the geological exploration borehole and the actual geological conditions, and then determine the value of \(d_{ij}\) according to the specific location of the defect in the borehole. The borehole can be divided into different areas by establishing a three-dimensional model of the borehole, and each area is assigned a corresponding influence weight, and then the specific distribution impact factor is determined according to the area where the defect is located. The exploration result impact index comprehensively reflects the degree of influence of the results of the entire geological exploration borehole on subsequent exploration work or engineering applications. If the exploration results of the borehole play a key role in determining the geological structure, mineral resource distribution, etc., the E value is between 0.8 and 1; if the borehole is only used as a general reference, the E value is between 0.3 and 0.6. It can be determined by comprehensive evaluation based on factors such as the objectives of the exploration project, the location and role of the borehole in the entire exploration area, and the expected subsequent applications.

[0137] Optionally, the defect category can be obtained by classifying the defect identification results, such as borehole deviation, hole wall collapse, diameter reduction and blockage. The importance coefficient can be constructed by constructing a hierarchical model through the hierarchical analysis method, taking the borehole quality as the target layer and the defect category as the criterion layer, and calculating the weight of each defect category relative to the target layer by establishing a judgment matrix. The quality of the geological exploration borehole based on the quality assessment value can be determined according to a preset quality range, such as Q≥90 can be set as high quality, 80≤Q<90 can be set as good, 70≤Q<80 can be set as qualified, and Q<70 can be set as unqualified.

[0138] Example 2:

[0139] like Figure 2 The figure shows a functional module diagram of a geological exploration drilling quality detection system based on image technology according to the present invention.

[0140] The image-based geological exploration borehole quality inspection system 200 described in the present invention can be installed in an electronic device. Depending on the functionality implemented, the image-based geological exploration borehole quality inspection system can include a preliminary image reconstruction module 201, a secondary image reconstruction module 202, an image segmentation module 203, and a quality assessment module 204. A module, also referred to as a unit, is a series of computer program segments that can be executed by an electronic device processor and perform a fixed function. These modules are stored in the electronic device's memory.

[0141] In the embodiment of the present invention, the functions of each module / unit are as follows:

[0142] The image preliminary reconstruction module 201 is used to collect high-definition images of geological exploration boreholes, construct a feature pyramid for the high-definition images, progressively extract features from the high-definition images using the feature pyramid to obtain multi-scale features, and perform high-level semantic reconstruction on the high-definition images based on the multi-scale features to obtain a preliminary reconstructed image;

[0143] The image secondary reconstruction module 202 is configured to extract multi-scale features from the preliminary reconstructed image, perform joint encoding on the multi-scale features to obtain potential coding features, perform wavelet frequency domain feature encoding on the preliminary reconstructed image to obtain wavelet coding features, perform coding fusion on the potential coding features and the wavelet coding features to obtain joint coding features, perform progressive image decoding on the joint coding features to obtain sub-band images, and perform inverse encoding on the sub-band images to obtain a target reconstructed image.

[0144] The image segmentation module 203 is used to identify the image structure of the target reconstructed image, deploy multiple agents in the target reconstructed image based on the image structure to obtain deployed agents, use the deployed agents to identify regional information of the target reconstructed image, and perform regional segmentation on the target reconstructed image based on the regional information to obtain a segmented image;

[0145] The quality assessment module 204 is configured to binarize the segmented image to obtain a binary image, construct a simplicial complex of the binary image, identify the homology group of the simplicial complex, construct a persistence graph of the binary image based on the homology group, identify topological features of the persistence graph, use the topological features to identify defects in the geological exploration borehole, and evaluate the quality of the geological exploration borehole based on the defect identification results.

[0146] In detail, the modules in the geological exploration drilling quality detection system 200 based on image technology in the embodiment of the present invention are used in the same manner as above. Figure 1 The same technical means as the geological exploration drilling quality detection method based on image technology described in the above are able to produce the same technical effects, so they will not be repeated here.

[0147] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for detecting the quality of geological exploration boreholes based on image technology, characterized in that: The method comprises: collecting high-definition images of geological exploration boreholes, constructing a feature pyramid for the high-definition images, progressively extracting features from the high-definition images using the feature pyramid to obtain multi-scale features, and performing high-level semantic reconstruction on the high-definition images based on the multi-scale features to obtain a preliminary reconstructed image; Extracting multi-scale features of the preliminary reconstructed image, and jointly encoding the multi-scale features to obtain potential coding features, performing wavelet frequency domain feature encoding on the preliminary reconstructed image to obtain wavelet coding features, performing coding fusion on the potential coding features and the wavelet coding features to obtain joint coding features, performing progressive image decoding on the joint coding features to obtain a sub-band image, and performing inverse encoding on the sub-band image to obtain a target reconstructed image; Identifying an image structure of the target reconstructed image, deploying multiple agents in the target reconstructed image based on the image structure to obtain deployed agents, identifying regional information of the target reconstructed image using the deployed agents, and performing regional segmentation on the target reconstructed image based on the regional information to obtain a segmented image; The segmented image is binarized to obtain a binary image, a simplicial complex of the binary image is constructed, and a homology group of the simplicial complex is identified. Based on the homology group, a persistence graph of the binary image is constructed, and topological features of the persistence graph are identified. Defects of the geological exploration borehole are identified using the topological features, and the quality of the geological exploration borehole is evaluated based on the defect identification results.

2. The method for detecting the quality of geological exploration drilling based on image technology according to claim 1, characterized in that: The step of progressively extracting features from the high-definition image using the feature pyramid to obtain multi-scale features includes: The feature pyramid is layered up and down to obtain a layered feature pyramid; Performing hierarchical feature detection operator deployment on the hierarchical feature pyramid to obtain a first deployment operator and a second deployment operator; Using the first deployment operator to perform Gaussian difference processing on the corresponding image of the lower layer of the hierarchical feature pyramid to obtain a differential image; Identifying image extreme points of the differential image and calculating the gradient histogram of the image extreme points to obtain lower-layer features; Using the second deployment operator to perform unit division on the upper layer corresponding image of the hierarchical feature pyramid to obtain a unit image; Calculating the gradient histogram of the unit image to obtain upper-layer features; The lower layer features and the upper layer features are subjected to feature fusion, and the fused features are subjected to feature screening to obtain multi-scale features.

3. The method for detecting the quality of geological exploration drilling based on image technology according to claim 1, characterized in that: The performing high-level semantic reconstruction on the high-definition image based on the multi-scale features to obtain a preliminary reconstructed image includes: Performing high-dimensional mapping on the multi-scale features to obtain high-dimensional multi-scale features; Performing self-attention encoding on the high-dimensional multi-scale features to obtain encoded multi-scale features; Constructing a semantic graph encoding multi-scale features; Performing relationship mining on the encoded multi-scale features in the semantic graph to construct a relationship model of the encoded multi-scale features; Based on the relationship model, image reconstruction is performed on the encoded multi-scale features to obtain a preliminary reconstructed image.

4. The method for detecting the quality of geological exploration drilling holes based on image technology according to claim 1, wherein: The performing wavelet frequency domain feature coding on the preliminary reconstructed image to obtain wavelet coding features includes: performing detail enhancement processing on the preliminary reconstructed image to obtain a preliminary processed image; identifying image features of the preliminarily processed image to configure a wavelet basis function of the preliminarily processed image; Performing multi-level sub-band decomposition on the preliminary processed image using the wavelet basis function to obtain multi-level image sub-bands; Calculating statistical features, energy features, and texture features of each sub-band of the multi-level image to obtain multi-level quantized features; Adaptively quantize the multi-level image subbands to obtain quantized features Zerotree coding is performed on the quantized features to obtain wavelet coding features.

5. The method for detecting the quality of geological exploration drilling based on image technology according to claim 1, characterized in that: The using the deployed intelligent agent to identify the region information of the target reconstructed image includes: Constructing a collaboration mechanism for the deployment agents and formulating a communication protocol for the deployment agents to obtain a communication agent; Using the communication agent to perform multimodal perception on the target reconstructed image to obtain multimodal data; querying a graph structure of the target reconstructed image, and performing regional modeling on the graph structure using the multimodal data to obtain a regional graph model; The communication agent is used to perform integrated learning on the region graph model to obtain region information of the target reconstructed image.

6. The method for detecting the quality of geological exploration drilling based on image technology according to claim 1, characterized in that: The performing region segmentation on the target reconstructed image based on the region information to obtain a segmented image includes: Segmenting the target reconstructed image into superpixels based on the region information; Querying pixel features of the superpixel; Calculating edge weights of the target reconstructed image based on the pixel features; Based on the edge weights, construct the region labels of the target reconstructed image Performing a preliminary segmentation on the target reconstructed image based on the region labels to obtain a preliminary segmentation image; inputting the preliminary segmentation image and the target reconstructed image into a preconfigured graph learning model to identify an optimized segmentation boundary of the target reconstructed image; Segmentation optimization is performed on the preliminary segmented image based on the optimized segmentation boundary to obtain a segmented image.

7. The method for detecting the quality of geological exploration drilling based on image technology according to claim 6, characterized in that: The calculating the edge weight of the target reconstructed image based on the pixel features includes: Querying the color feature, texture feature, and spatial distance in the pixel feature to obtain a query feature; Quantifying the query feature to obtain a quantitative feature; Based on the quantitative features, edge weights of the target reconstructed image are calculated.

8. The method for detecting the quality of geological exploration drilling based on image technology according to claim 1, characterized in that: The identifying the topological features of the persistent graph includes: Gridding the persistence graph to obtain a grid graph; Identifying homology class points in the grid graph and calculating the persistence interval of the homology class points; plotting a histogram of the persistence interval; Identifying interval features of the persistence graph based on the histogram, and counting point densities of homology points in different grid intervals in the grid graph; A topological feature of the persistence graph is identified based on the interval feature and the point density.

9. The method for detecting the quality of geological exploration drilling based on image technology according to claim 1, characterized in that: The step of evaluating the quality of the geological exploration borehole based on the defect identification result includes: Determining the defect category of the geological exploration borehole based on the defect identification result; Analyzing the importance coefficients of different categories in the defect category; Calculating a quality assessment value of the geological exploration borehole based on the defect category and the importance coefficient; Based on the quality assessment value, the quality of the geological exploration borehole is determined.

10. A geological exploration drilling quality detection system based on image technology, characterized in that: The system comprises: a preliminary image reconstruction module for collecting high-definition images of geological exploration boreholes, constructing a feature pyramid for the high-definition images, performing progressive feature extraction on the high-definition images using the feature pyramid to obtain multi-scale features, and performing high-level semantic reconstruction on the high-definition images based on the multi-scale features to obtain a preliminary reconstructed image; an image secondary reconstruction module, configured to extract multi-scale features of the preliminary reconstructed image, perform joint encoding on the multi-scale features to obtain potential coding features, perform wavelet frequency domain feature encoding on the preliminary reconstructed image to obtain wavelet coding features, perform coding fusion on the potential coding features and the wavelet coding features to obtain joint coding features, perform progressive image decoding on the joint coding features to obtain a sub-band image, and perform inverse encoding on the sub-band image to obtain a target reconstructed image; an image segmentation module, configured to identify an image structure of the target reconstructed image, deploy multiple agents in the target reconstructed image based on the image structure to obtain deployed agents, use the deployed agents to identify regional information of the target reconstructed image, and perform regional segmentation on the target reconstructed image based on the regional information to obtain a segmented image; A quality assessment module is configured to binarize the segmented image to obtain a binary image, construct a simplicial complex of the binary image, identify a homology group of the simplicial complex, construct a persistence graph of the binary image based on the homology group, identify topological features of the persistence graph, perform defect identification on the geological exploration borehole using the topological features, and evaluate the quality of the geological exploration borehole based on the defect identification results.

Images