Method and system for automatic construction of geotechnical engineering investigation three-dimensional geological model based on machine learning and BIM

Abstract

The application provides a kind of geotechnical engineering investigation three-dimensional geological model automatic construction method and system based on machine learning and BIM, it is related to geotechnical engineering investigation technical field, wherein, the method comprises: after obtaining borehole investigation data, color correction is carried out on original core image based on color chart image, corrected core image is input into convolutional neural network model to identify rock type and demarcate first rock layer boundary line;Key area of resistivity gradient change is used to replace fuzzy area to obtain second rock layer boundary line;According to the inclination correction space position and fill the wave velocity value to generate target borehole columnar section;According to the degree of stratigraphic relief, the calculation weight of spatial interpolation algorithm is adaptively adjusted, and three-dimensional stratum data is generated;Three-dimensional stratum data is imported into BIM software interface to automatically assemble three-dimensional geological model.The application improves the degree of automation and spatial accuracy of geological model construction.

Description

Technical Field

[0001] This application relates to the field of geotechnical engineering investigation technology, and in particular to an automated method and system for constructing three-dimensional geological models for geotechnical engineering investigation based on machine learning and BIM. Background Technology

[0002] Geotechnical engineering investigation is a fundamental part of construction engineering. Its core task is to obtain information on the geological structure through drilling, sampling, and on-site testing, so as to provide a basis for engineering design and construction. Traditional geotechnical investigation results are mainly expressed through paper borehole columnar diagrams and two-dimensional profile diagrams. The information is scattered and difficult to integrate. When facing complex engineering projects such as high-rise buildings and deep foundation pits, the limitations of two-dimensional drawings in terms of spatial expression become more and more prominent. Geologists are prone to misunderstandings during the interpretation process, which in turn affects the accuracy of engineering decisions.

[0003] In recent years, with the penetration of digital technology into the engineering field, integrating drilling data with 3D visualization technology has become an important direction for improving exploration efficiency. Several existing technologies have emerged that combine Building Information Modeling (BIM) with geotechnical exploration data. For example, drilling column data can be manually annotated and then imported into modeling software to generate 3D geological bodies, or geographic information platforms can be used to spatially display exploration data. However, the above solutions rely on manual experience to interpret core photos during the stratum identification stage. The accuracy of determining stratigraphic boundaries fluctuates greatly due to the combined influence of the surveyor's subjective judgment and on-site lighting conditions. Furthermore, the entire identification and annotation process is time-consuming, which limits efficiency when facing large-scale drilling projects.

[0004] In the three-dimensional geological modeling stage, existing solutions typically use spatial interpolation algorithms with fixed parameters to extrapolate the strata between boreholes. When the strata between adjacent boreholes change drastically, the fixed-parameter interpolation algorithm is prone to producing overly smooth extrapolation results, leading to a deviation between the generated three-dimensional geological model and the actual geological conditions. In engineering applications, this may result in the omission of unfavorable geological conditions. Summary of the Invention

[0005] This application provides an automated method and system for constructing three-dimensional geological models for geotechnical engineering exploration based on machine learning and BIM, in order to solve the technical problems in the prior art where the reliance on human experience in rock strata identification leads to large fluctuations in accuracy, and the use of fixed parameters in spatial interpolation algorithms leads to modeling deviations in undulating areas of strata.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides an automated method for constructing three-dimensional geological models for geotechnical engineering investigation based on machine learning and BIM, comprising: Acquire borehole exploration data collected on-site, including original core images, as well as borehole inclination, resistivity, and wave velocity; Based on the original core image and color chart image, a color-accurate corrected core image is obtained. The corrected core image is then input into a pre-trained convolutional neural network model to identify the rock type and analyze the variation characteristics of the image texture to delineate the first rock layer boundary. Find the key areas where the resistivity change of the borehole meets the preset conditions, and use the key areas to replace the blurred areas in the core image to obtain the second rock layer boundary. Based on the borehole inclination, the spatial position of the second stratum boundary is corrected, a basic borehole columnar section is generated, and the wave velocity value corresponding to each borehole depth is filled into the basic borehole columnar section to generate the target borehole columnar section. The degree of undulation of the same strata between multiple boreholes is calculated. The calculation weight of the spatial interpolation algorithm is adjusted according to the degree of undulation. The spatial interpolation algorithm with adjusted calculation weight is used to calculate the columnar section of multiple target boreholes to generate three-dimensional stratigraphic data for the entire site. The spatial coordinates and corresponding wave velocity values ​​of the three-dimensional stratigraphic data are extracted and imported into the BIM software interface to generate three-dimensional solid components, which are then automatically assembled into a three-dimensional geological model.

[0007] Optionally, key regions where the resistivity change in the borehole meets preset conditions are identified, and these key regions are used to replace blurred regions in the corrected core image to obtain the second stratum boundary line, including: Extract the depth intervals between adjacent depth boundary points in the first rock stratum boundary line, which are determined by the contrast of adjacent image slices of the corrected core image being less than a preset contrast threshold, and mark the depth intervals as blurred regions. Calculate the resistivity gradient of the borehole along the drilling depth direction; Extract the physical depth coordinates where the resistivity gradient value is greater than a preset gradient threshold, and mark the physical depth coordinates as key areas; By replacing the depth coordinates of the fuzzy areas with the physical depth coordinates of the key areas, the second rock stratum boundary line is obtained.

[0008] Optionally, based on the borehole inclination, the spatial position of the second stratum boundary is corrected to generate a basic borehole columnar section. The wave velocity values ​​corresponding to each borehole depth are then filled into the basic borehole columnar section to generate a target borehole columnar section, including: Based on the borehole inclination, the horizontal and vertical offsets at each sampling depth are calculated. The horizontal and vertical offsets are then used to perform a three-dimensional coordinate transformation on the second rock stratum boundary line to obtain a spatial boundary curve that reflects the actual borehole trajectory. Using geophysical exploration components to transmit and receive acoustic signals inside the borehole, the propagation time difference of the acoustic signals at different strata depths is extracted, and the wave velocity value at the corresponding depth is calculated based on the propagation time difference. Based on the spatial boundary curve, the depth range within the borehole is segmented and labeled to generate a basic borehole columnar diagram that shows the distribution characteristics of stratigraphic lithology. The wave velocity values ​​are mapped to the corresponding depth positions within the segments of the basic borehole columnar section, and the physical property parameters are filled to generate the target borehole columnar section.

[0009] Optionally, the degree of undulation of the same strata among multiple boreholes is calculated, the calculation weights of the spatial interpolation algorithm are adjusted according to the degree of undulation, and the spatial interpolation algorithm with adjusted calculation weights is used to calculate the columnar section of multiple target boreholes to generate three-dimensional stratigraphic data for the entire site, including: Obtain the coordinates of the top plate of rock strata of the same rock category in the columnar diagrams of multiple target boreholes; Calculate the elevation difference of the top plate coordinates of adjacent boreholes and the horizontal spacing between boreholes; The degree of undulation is obtained by calculating the relative ratio of the elevation difference to the horizontal spacing of the boreholes; Construct an inverse proportional function between the degree of fluctuation and the preset interpolation search radius to generate interpolation weight adjustment coefficients; The distance attenuation factor of the spatial interpolation algorithm is corrected by using the interpolation weight adjustment coefficient, resulting in a spatial interpolation algorithm with adjusted calculation weights. A spatial interpolation algorithm with adjusted computational weights is used to perform three-dimensional spatial interpolation on the wave velocity values ​​and coordinates of multiple target borehole columnar sections to generate three-dimensional formation data.

[0010] Optionally, based on the original core image and the colorimetric chart image, a color-accurate corrected core image is obtained, including: Locate the center point of the color patch in the color chart image and extract the standard reference color of the center point of the color patch; By comparing the standard reference color with the preset actual target color, the color difference compensation amount for different color channels is calculated. For each color channel, a polynomial fitting is performed with the actual target color value of each color block on the color chart in that color channel as the independent variable and the color difference compensation amount of the corresponding color block in that color channel as the dependent variable to construct a color mapping table for that color channel. The color mapping table is a polynomial function that takes the channel value of any pixel in that color channel as input and outputs the color difference compensation amount of that pixel in that color channel. For each pixel in the original core image, substitute the red, green, and blue channel values ​​of the pixel into the color mapping table of the corresponding color channel to calculate the color difference compensation amount of the pixel in each color channel. Then, superimpose the color difference compensation amount of each color channel onto the original channel value of the corresponding color channel to obtain the corrected core image.

[0011] Optionally, the corrected core image is input into a pre-trained convolutional neural network model to identify the rock type and analyze the variation characteristics of image texture to delineate the first rock stratum boundary, including: The core image is divided into continuously arranged image slices according to the preset depth step size.

[0012] Visual features of each image slice are extracted using a convolutional neural network model, and category matching is performed to determine the rock type of each image slice.

[0013] By comparing the rock types of two adjacent image slices, within the interval where the rock type changes, the color histogram and texture gray-level co-occurrence matrix of the two adjacent image slices within the interval are extracted respectively. The distance between the two color histograms and the contrast difference between the two texture gray-level co-occurrence matrices are calculated respectively to obtain the color histogram distance and texture contrast difference. The color histogram distance and texture contrast difference are weighted and summed to obtain the jump gradient. When the jump gradient is greater than the preset jump gradient threshold, the boundary depth position of the two adjacent image slices is determined as the depth limit point.

[0014] For each pair of adjacent different rock types, multiple depth boundary points located at the junction of the adjacent different rock types are connected in sequence to generate the first stratum boundary line between the adjacent different rock types.

[0015] Optionally, the spatial coordinates and corresponding wave velocity values ​​of the 3D stratigraphic data are extracted and imported into the BIM software interface to generate 3D solid components, which are then automatically assembled into a 3D geological model, including: Analyze the stratigraphic contour boundaries of 3D stratigraphic data and extract the spatial coordinates of the stratigraphic contour boundaries; Based on spatial coordinates, spatial geometric meshing is performed to generate a closed three-dimensional stratigraphic shell; The corresponding wave velocity value is mapped as a physical property parameter to the three-dimensional stratum shell. The mapped three-dimensional stratum shell is instantiated through the interface of the BIM software to generate a three-dimensional solid component. The relative positional relationship of multiple three-dimensional solid components is determined based on spatial coordinates. The three-dimensional solid components are then spatially spliced ​​according to their relative positional relationship to automatically assemble a three-dimensional geological model.

[0016] Optionally, the specific process of obtaining the original core image includes: Fix the color chart with the standard colors onto the core tray to be photographed; Industrial cameras were used to capture raw field images, including rock cores and color charts. A laser rangefinder was used to obtain the actual shooting distance between the industrial camera and the core tray. The original site images are scaled according to the actual shooting distance to obtain the original core images.

[0017] Optionally, after the step of automatically assembling the three-dimensional geological model, the following steps are also included: Extract weak interlayers in the 3D geological model whose stratum thickness is less than the preset thickness value; The spatial foundation pit outline data of the proposed project is superimposed onto the three-dimensional geological model; Extract the spatial intersection coordinates between the spatial foundation pit outline data and the weak interlayer body; Based on the spatial intersection coordinates, perform early warning color rendering on the corresponding components in the three-dimensional geological model.

[0018] Secondly, this application provides an automated system for constructing three-dimensional geological models for geotechnical engineering investigation based on machine learning and BIM, including: The acquisition module is used to acquire borehole exploration data collected on site. The borehole exploration data includes original core images, as well as borehole inclination, resistivity and wave velocity. The input module is used to obtain a color-accurate corrected core image based on the original core image and the color chart image. The corrected core image is then input into a pre-trained convolutional neural network model to identify the rock type, analyze the variation characteristics of the image texture, and delineate the first rock layer boundary. The search module is used to find key areas where the resistivity change of the borehole meets the preset conditions, and use the key areas to replace and correct the blurred areas in the core image to obtain the second rock layer boundary line. The correction module is used to correct the spatial position of the second rock stratum boundary line according to the borehole inclination, generate a basic borehole columnar section, fill the basic borehole columnar section with the wave velocity value corresponding to each borehole depth, and generate the target borehole columnar section. The calculation module is used to calculate the degree of undulation of the same strata among multiple boreholes, adjust the calculation weight of the spatial interpolation algorithm according to the degree of undulation, and use the spatial interpolation algorithm with adjusted calculation weights to calculate the columnar section of multiple target boreholes to generate three-dimensional stratigraphic data of the entire site. The extraction module is used to extract the spatial coordinates and corresponding wave velocity values ​​of 3D stratigraphic data, and import them into the BIM software interface to generate 3D solid components, automatically assembling a 3D geological model.

[0019] This application provides an automated method for constructing a 3D geological model for geotechnical engineering investigation based on machine learning and BIM. The method includes: acquiring borehole exploration data collected on-site, including original core images, borehole inclination, resistivity, and wave velocity; obtaining a color-accurate corrected core image based on the original core image and a colorimetric chart image; inputting the corrected core image into a pre-trained convolutional neural network model to identify rock types and analyze image texture variation characteristics to delineate the first rock stratum boundary; identifying key areas where the borehole resistivity variation meets preset conditions, and using these key areas to replace blurry areas in the corrected core image. The region is identified by obtaining the second stratum boundary line. Based on the borehole inclination, the spatial position of the second stratum boundary line is corrected to generate a basic borehole columnar section. The wave velocity value corresponding to each borehole depth is then filled into the basic borehole columnar section to generate the target borehole columnar section. The undulation degree of the same strata between multiple boreholes is calculated. The calculation weight of the spatial interpolation algorithm is adjusted according to the undulation degree. The spatial interpolation algorithm with adjusted calculation weights is then used to calculate multiple target borehole columnar sections to generate three-dimensional stratigraphic data for the entire site. The spatial coordinates and corresponding wave velocity values ​​of the three-dimensional stratigraphic data are extracted and imported into the BIM software interface to generate three-dimensional solid components, which are then automatically assembled into a three-dimensional geological model.

[0020] The beneficial effects of this application are as follows: First, by automatically recognizing the corrected core images using a convolutional neural network model, the interference of human experience in determining the rock strata boundaries is reduced; second, by replacing and correcting blurred areas in image recognition using resistivity gradient data, the positioning accuracy of rock strata boundaries is further improved; third, by adaptively adjusting the calculation weights of the spatial interpolation algorithm according to the degree of strata undulation, the modeling effect of areas with severe strata undulation is improved; in addition, by automatically generating and assembling three-dimensional solid components by importing three-dimensional strata data into the BIM software interface, the automation level of three-dimensional geological model construction is improved. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the structure of a geotechnical investigation data acquisition device disclosed in an embodiment of this application.

[0023] Figure 2 This is an overall flowchart of an automated method for constructing a three-dimensional geological model for geotechnical engineering investigation based on machine learning and BIM, as disclosed in an embodiment of this application.

[0024] Figure 3 This is a schematic diagram of data interaction for a color correction process disclosed in an embodiment of this application.

[0025] Figure 4 This is a schematic diagram of the structure of a first rock stratum boundary line and an ambiguous region disclosed in an embodiment of this application.

[0026] Figure 5 This is a schematic diagram illustrating the relationship between resistivity gradient and image contrast in the borehole depth direction, as disclosed in an embodiment of this application.

[0027] Figures 6a-6c This is a schematic diagram of a fuzzy region nearest neighbor matching and replacement process and the generation result of the second rock stratum boundary line disclosed in an embodiment of this application.

[0028] Figure 7a This is a schematic diagram of a basic borehole columnar section disclosed in an embodiment of this application.

[0029] Figure 7b This is a schematic diagram of a target borehole columnar section disclosed in an embodiment of this application.

[0030] Figure 8 This is a flowchart illustrating an adaptive adjustment of spatial interpolation weights disclosed in an embodiment of this application.

[0031] Figure 9 This is a planar schematic diagram of a three-dimensional spatial interpolation process disclosed in an embodiment of this application.

[0032] Figure 10 This is a schematic diagram of a three-dimensional geological model and a foundation pit outline superimposed for early warning, as disclosed in an embodiment of this application. Detailed Implementation

[0033] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. The described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0034] See Figure 1 As shown, the geotechnical exploration data acquisition device involved in this application embodiment includes a core tray 101, an industrial camera 102, a laser rangefinder 103, a tilt measuring device 104, a geophysical logging device 105, and a geophysical exploration component 106; wherein, the core tray 101 is used to hold the core samples taken from the borehole, and the length of the core tray 101 matches the length of the core sample in a single drilling round.

[0035] The industrial camera 102 is mounted on a bracket directly above the core tray 101 and is used to collect original on-site images containing core samples and color charts. The resolution of the industrial camera 102 is no less than 12 megapixels. The laser rangefinder 103 is fixed on the same bracket as the industrial camera 102, and their optical axes are kept parallel. The laser rangefinder 103 is used to measure the actual shooting distance between the industrial camera 102 and the core tray 101 in real time. The ranging accuracy can be set according to actual needs. For example, this application uses an accuracy of ±1 mm.

[0036] The inclination measuring device 104 is installed inside the drilling tool and is used to record the azimuth and zenith angle of the borehole at different depths in real time during the drilling process, thereby obtaining the inclination of the borehole; the geophysical logging device 105 is deployed along the depth direction of the borehole and is used to continuously record the apparent resistivity value of the formation, thereby obtaining the resistivity of the borehole; the geophysical exploration component 106 includes an acoustic wave transmitting probe and an acoustic wave receiving probe, which are deployed at a fixed interval along the depth direction of the borehole and are used to transmit and receive acoustic wave signals to obtain the wave velocity of the borehole.

[0037] Based on the aforementioned data acquisition device, this application provides an automated method for constructing three-dimensional geological models for geotechnical engineering investigation based on machine learning and BIM. (See [link to relevant documentation]). Figure 2 As shown, the method includes the following steps: S1: Acquire borehole exploration data collected on-site, including original core images, as well as borehole inclination, resistivity, and wave velocity.

[0038] The specific process for obtaining the original core image and colorimetric card image is as follows: First, the colorimetric card is fixed on the core tray 101. The colorimetric card is a rigid plate printed with multiple standard color blocks. The standard color blocks refer to color blocks that are precisely mixed and solidified according to the International Commission on Illumination color space standard during manufacturing. Each standard color block has a known and stable color value. The core tray 101 is a long strip container used to hold the columnar core sample extracted from the borehole.

[0039] Then, the industrial camera 102 is used to acquire original field images containing core samples and color charts. The original field images refer to complete digital images that are directly captured by the industrial camera 102 at the exploration site without any color correction or cropping.

[0040] Then, the actual shooting distance between the industrial camera 102 and the core tray 101 is obtained by using the laser rangefinder 103. The actual shooting distance refers to the vertical spatial distance between the center of the lens of the industrial camera 102 and the upper surface of the core tray 101.

[0041] Finally, the original field images are scaled according to the actual shooting distance to obtain the original core images. Since different shooting distances will cause the pixel size of the core in the image to change, it is necessary to calculate the scaling factor based on the ratio of the actual shooting distance to the preset standard distance, and crop and scale the core area and color chart area in the original field images to a uniform pixel scale.

[0042] In practical engineering applications, assuming that the preset standard shooting distance of the industrial camera 102 is 500 mm, and the laser rangefinder 103 measures the actual shooting distance of a certain shooting as 520 mm, then the scaling factor K = 500 / 520 = 0.9615. Multiplying the width and height of the original scene image by K respectively will yield the image size corresponding to the standard distance.

[0043] Meanwhile, the inclination measuring device 104 records the azimuth and zenith angle at each depth in real time during drilling to obtain the borehole inclination. The inclination is determined by the triaxial accelerometer and triaxial magnetic flux sensor built into the inclination measuring device 104 simultaneously sampling at each preset depth interval. The zenith angle is obtained by measuring the deflection angle of the drill bit relative to the direction of gravity using the accelerometer, and the azimuth angle is obtained by measuring the rotation angle of the drill bit relative to the Earth's magnetic north pole using the magnetic flux sensor. The combination of the zenith angle and azimuth angle constitutes the inclination data at that depth. In practical engineering applications, assuming the inclination measuring device 104 samples a zenith angle of 2.5 degrees and an azimuth angle of 120 degrees at a depth of 5.0 meters, then the inclination at that depth... The tilt data is recorded as a combination of a zenith angle of 2.5 degrees and an azimuth angle of 120 degrees; the geophysical logging device 105 continuously records the apparent resistivity value of the formation along the borehole depth direction to obtain the resistivity of the borehole; the geophysical exploration component 106 emits pulsed sound waves in the borehole and receives echo signals to obtain the wave velocity of the borehole. The wave velocity is obtained by the sound wave emitting probe emitting pulsed sound waves into the formation, the pulsed sound waves propagate in the formation medium and generate echo signals, and the sound wave receiving probe captures the echo signals and records the propagation time difference T between the pulsed sound waves from the time of emission to the time the echo signals are received. The wave velocity value V is equal to the distance Ds between the sound wave emitting probe and the sound wave receiving probe divided by the propagation time difference T, i.e., V=Ds / T.

[0044] This step, through standardized data acquisition procedures and equipment calibration, ensures the quality and consistency of the data required for subsequent rock type identification and 3D modeling.

[0045] S2: Correct the original core image using a standard color chart to obtain a corrected core image. Input the corrected core image into a pre-trained convolutional neural network model to identify the rock type and analyze the variation characteristics of the image texture to delineate the first rock layer boundary.

[0046] In step S1, the original core image and color chart image have been acquired. Since the on-site lighting conditions are constantly changing due to factors such as weather, time of day and surrounding buildings blocking the light, the core images taken at different times will have deviations in hue, brightness and saturation. Therefore, it is necessary to use the known standard color blocks in the color chart image to perform color correction on the original core image in order to eliminate the interference of ambient light color cast on subsequent rock type identification.

[0047] See Figure 3 As shown, the data interaction process for color correction is as follows: First, the area containing the color chart is cropped from the original on-site image. The center point of each standard color block is located, and the standard reference color and the actual target color are extracted. Specifically, the color chart area is segmented using an image segmentation algorithm to identify the outline boundary of each standard color block. Then, the geometric center pixel of the outline area of ​​each standard color block is taken as the center point of the color block. The standard reference color refers to the theoretical color value of each standard color block under standard light source, which is pre-calibrated by the color chart manufacturer and is stored in the calibration database. Each standard color block has a unique number, and the number corresponds one-to-one with the standard reference color value. At the same time, the pixel value of the center point of the color block is actually captured by the industrial camera 102 under the current lighting conditions, and this pixel value is recorded as the actual target color of the standard color block.

[0048] Then, by comparing the standard reference color with the preset actual target color, the color difference compensation amount for different color channels is calculated. The actual target color refers to the pixel value actually captured by the industrial camera 102 under the current ambient lighting conditions for each standard color patch on the color chart; that is, the actual image color value of the same standard color patch due to ambient light color cast. Specifically, the difference between the standard reference color and the actual target color is calculated in the red, green, and blue channels to obtain the color difference compensation amount for each color channel. For example, if the standard reference color of a certain standard color patch has a value of 192 in the red channel, and the actual target color of that standard color patch captured by the industrial camera 102 under the current lighting conditions has a value of 185 in the red channel, then the color difference compensation amount for that standard color patch in the red channel is 192 - 185 = 7. Furthermore, when performing pixel correction on the original core image subsequently, a continuous mapping function is constructed for the color difference compensation amounts of all standard color patches through polynomial fitting, so that the corresponding color difference compensation amount can be calculated for any pixel value in the original core image using this mapping function.

[0049] Furthermore, for each color channel, a polynomial fitting is performed with the actual target color value of each color patch on the color chart in that color channel as the independent variable and the color difference compensation amount of the corresponding color patch in that color channel as the dependent variable to construct a color mapping table for that color channel. The color mapping table is a polynomial function that takes the channel value of any pixel in that color channel as input and outputs the color difference compensation amount of that pixel in that color channel. Since the influence of ambient light on different color regions is not nonlinear and uniform, if a simple linear offset is used for correction, the correction error in dark and light regions will increase significantly. Therefore, polynomial fitting is used to capture the nonlinear color shift pattern.

[0050] Finally, channel compensation is performed on the global pixels of the original core image according to the color mapping table to obtain the corrected core image. The specific process is as follows: traverse each pixel in the original core image, substitute the red channel value, green channel value, and blue channel value of each pixel into the polynomial fitting function of the color mapping table of the corresponding color channel, calculate the color difference compensation amount of the pixel in each color channel, and then add the color difference compensation amount of the corresponding color channel to the original value of each color channel of each pixel to obtain the corrected pixel value. After traversing all pixels, the corrected core image is obtained.

[0051] After color correction to obtain the corrected core image, the corrected core image needs to be input into a pre-trained convolutional neural network model for rock type identification. Since the determination of rock type involves multiple levels of visual features such as color distribution, texture direction and mineral particle morphology, the traditional threshold segmentation method is limited in accuracy when dealing with multi-class rock images. Therefore, a convolutional neural network model is used to achieve automated feature extraction and class matching.

[0052] The construction process of the convolutional neural network model is as follows: This application adopts a deep residual network structure as the backbone architecture of the convolutional neural network model. This network structure includes an input layer, several convolutional layers, pooling layers, batch normalization layers, residual connection modules, and fully connected classification layers. Among them, the input layer receives image slices of fixed size, and the convolutional layers perform spatial convolution operations on the input image through multiple sets of convolutional kernels to extract low-level visual features such as edges, textures, and color distributions. The pooling layers downsample the feature maps output by the convolutional layers to reduce the spatial resolution of the feature maps and reduce the amount of computation. The batch normalization layers normalize the mean and variance of each batch of feature maps to accelerate the convergence speed of network training. The residual connection modules make skip connections between shallow features and deep features to avoid the gradient vanishing problem when the number of network layers increases. The fully connected classification layer maps the high-dimensional feature vectors extracted by the previous layers to the probability distribution space of rock categories and outputs the confidence scores of each rock type.

[0053] The training sample set was labeled by professionals in the field of geotechnical engineering, including labeled images of common rock types such as granite, sandstone, mudstone, limestone, and shale, as well as Quaternary loose sediments. The total number of training samples is no less than 5,000. The training process uses the cross-entropy loss function and stochastic gradient descent optimizer, with 200 training rounds. The initial learning rate is set to 0.001 and decays using a cosine annealing strategy.

[0054] After the convolutional neural network model is trained, rock type identification and delineation of the first rock stratum boundary are performed according to the following steps: The first step is to divide the calibrated core image into continuously arranged image slices according to a preset depth step size. The preset depth step size is used to control the granularity of image segmentation and can be set according to actual needs. For example, this application uses a depth step size of 0.1 meters for segmentation, that is, each 0.1-meter depth range corresponds to one image slice. There are no overlapping areas between image slices to avoid the same depth range being repeatedly identified.

[0055] The second step involves extracting the visual features of each image slice using a convolutional neural network model and performing category matching to determine the rock type of each image slice. Since image feature extraction and classification using a convolutional neural network model is a common technique in the field of image recognition, this application will not elaborate on the specific convolution operation process inside the convolutional neural network model. In addition to the convolutional neural network model, other image classification models can also be used to achieve the same function. The basic process is as follows: after each image slice is input into the convolutional neural network model, it undergoes feature extraction and multi-scale feature fusion processing. The fully connected classification layer outputs the confidence score of the image slice belonging to each rock type, and the category with the highest confidence score is taken as the rock type label of the image slice.

[0056] The third step is to compare the rock types of two adjacent image slices, identify the gradient changes in texture and color in the interval where the rock type changes, and locate the depth boundary point of the rock property change by the gradient change.

[0057] Specifically, when the gradient jump is greater than the preset gradient jump threshold, it is confirmed that there is a valid rock layer boundary in the interval. The preset gradient jump threshold is used to filter out false jumps caused by image noise, and can be set according to actual needs. For example, this application uses a preset gradient jump threshold of 0.35. When the rock type label of two adjacent image slices changes, it indicates that there is a rock layer boundary near the intersection depth of the two image slices.

[0058] The specific process is as follows: For the pixels in the boundary area of ​​two adjacent image slices, the frequency of color values ​​is counted according to the three channels of red, green and blue. The frequency of each channel is discretized and normalized according to the number of preset intervals to obtain a color histogram representing the color distribution characteristics of the image slice. The sum of the absolute values ​​of the differences between the color histograms of two adjacent image slices according to the corresponding intervals is used as the color histogram distance. At the same time, for the grayscale images of two adjacent image slices in the boundary area, the co-occurrence frequency of pixel grayscale values ​​is counted according to the preset direction and preset step size to construct a texture grayscale co-occurrence matrix representing the texture spatial distribution characteristics of the image slice. The absolute value of the difference between the contrast operators of the corresponding elements of the two texture grayscale co-occurrence matrices is calculated as the texture contrast difference. Then, the weighted sum of the color histogram distance and the texture contrast difference is used as the jump gradient. The calculation formula of the jump gradient is shown in Equation (1).

[0059]

[0060] In formula (1): For abrupt gradients, The histogram distance between two adjacent image slices is the color histogram distance. This represents the difference in texture contrast. and The weighting coefficients and + =1, which can be set according to actual needs, for example, this application adopts =0.6 and =0.4.

[0061] In practical engineering applications, assuming an image slice at a depth of 5.0 meters is identified as sandstone and an image slice at a depth of 5.1 meters is identified as mudstone, the color histogram of the boundary region between the two image slices at 5.0 meters and 5.1 meters is extracted, and the histogram distance is calculated. =0.42; Extract the texture gray-level co-occurrence matrix and calculate the texture contrast difference. =0.28; Substitute into equation (1) =0.6×0.42+0.4×0.28=0.252+0.112=0.364; Since the gradient jump of 0.364 is greater than the preset gradient jump threshold of 0.35, it is confirmed that there is an effective rock stratum boundary in the boundary area; Since the two image slices correspond to a depth range of 5.0 meters to 5.1 meters respectively, the depth limit point is taken as the middle depth position of the boundary area between the two image slices, that is, (5.0+5.1) / 2=5.05 meters, and the depth of 5.05 meters is determined as the depth limit point.

[0062] The fourth step is to connect multiple depth boundary points located at the junction of each pair of adjacent different rock types to generate the first stratum boundary line between the adjacent different rock types.

[0063] The specific process is as follows: according to the boundary relationship between adjacent different rock types corresponding to each depth boundary point, multiple depth boundary points located at the same boundary are sequentially connected to form a first stratum boundary line representing the boundary position of the adjacent different rock types. Each pair of adjacent different rock types within the borehole corresponds to the generation of a first stratum boundary line. For example... Figure 4 As shown, Figure 4 The diagram illustrates the first stratum boundary and the structure of an ambiguous region in a borehole. With depth (m) as the vertical axis, each stratum segment is distinguished by a region with a different filling pattern, from top to bottom: fill, sandstone, mudstone, granite, and limestone. Multiple depth boundary points, located by the aforementioned gradient, are distributed between adjacent different rock types. Each depth boundary point is marked with a black dot. For each pair of adjacent different rock types, the adjacent depth boundary points located at the boundary of the adjacent different rock types are connected sequentially with line segments to form the first stratum boundary between the adjacent different rock types.

[0064] This step eliminates the interference of on-site lighting conditions through color correction and realizes automatic identification of rock types through a convolutional neural network model, reducing the influence of human experience on boundary determination.

[0065] S3: Within the preset range of the first stratum boundary, find the key area where the resistivity change of the borehole meets the preset conditions, and use the key area to replace the blurred area in the core image. Then, determine the second stratum boundary based on the adjusted core image.

[0066] In step S2, the first rock stratum boundary line has been obtained. Since the convolutional neural network model is based on the core surface image for recognition, when the color and texture differences between adjacent rock strata are small, the contrast between image slices is low, which leads to a large uncertainty in the model's recognition results in these areas (i.e., blurred areas).

[0067] To further identify these ambiguous areas, this application considers that differences in mineral composition and water content among different rock types can cause abrupt changes in resistivity. Therefore, resistivity data can be used to provide supplementary physical evidence for identifying ambiguous areas in core images, thereby improving the overall positioning accuracy of rock strata boundaries.

[0068] See Figure 5 As shown, Figure 5This diagram illustrates the correlation between borehole depth and image contrast and resistivity gradient. The dark black line represents the image contrast curve, maintaining high contrast (value 8) in the depth ranges of 0-5m and 15-20m, while significantly decreasing to low contrast (value 2) in the blurred region at depth 10m. The dark gray line represents the resistivity gradient curve, maintaining a low gradient (value 2) in the depth ranges of 0-5m and 15-20m, while exhibiting a distinct peak and reaching a high gradient (value 9) in the blurred region at depth 10m. The two curves show opposite trends in the blurred region, intuitively reflecting the spatial correspondence between image blur and resistivity anomalies within this depth range.

[0069] The specific implementation process for this step is as follows: First, the depth intervals between adjacent depth boundary points in the first rock stratum boundary line, determined by the contrast of adjacent image slices of the corrected core image being less than a preset contrast threshold, are extracted. These depth intervals are marked as blurred regions, that is, the blurred regions correspond to the line segments between adjacent depth boundary points on the first rock stratum boundary line where no new depth boundary points can be identified within the line segment. The preset contrast threshold can be set according to actual needs; for example, the preset contrast threshold used in this application is 0.15.

[0070] like Figure 4 As shown, for the line segment connecting adjacent depth boundary points in the first rock stratum boundary line, if the line segment corresponds to a low contrast depth segment in the corrected core image where the contrast of adjacent image slices is lower than a preset contrast threshold, i.e., no new depth boundary point can be identified within the line segment, the line segment is marked as a blurred area. The blurred area is enclosed in the figure with a dashed box, and the line segment within the blurred area is drawn as a dashed line segment to distinguish it from the solid line segment boundary line formed by adjacent depth boundary points that meet the contrast requirements.

[0071] The contrast ratio is calculated by taking the absolute value of the difference between the mean gray values ​​of two adjacent image slices and dividing it by the sum of their mean gray values. The calculation formula is shown in equation (2).

[0072] (2) In equation (2): C is the contrast ratio. and These are the average gray values ​​of two adjacent image slices; when C is less than 0.15, this depth range is marked as a blurred region.

[0073] Then, calculate the resistivity gradient of the borehole along the drilling depth direction; the calculation formula is shown in equation (3).

[0074]

[0075] In formula (3): The resistivity gradient value, and Depth and The resistivity value at that point, The depth spacing between two adjacent sampling points.

[0076] Then, the resistivity gradient value is extracted. Physical depth coordinates greater than a preset gradient threshold are marked as key areas; the preset gradient threshold can be set according to actual needs, for example, this application uses twice the average site resistivity gradient as the preset gradient threshold.

[0077] Finally, the physical depth coordinates of the key region are used to replace the depth coordinates of the fuzzy region to obtain the second stratum boundary. It should be noted that, in order to avoid the resistivity abrupt change locations that are too far from the first stratum boundary being misjudged as the depth limit points of the stratum boundary, this step limits the search range of the key region to within 0.5 meters above and below each depth limit point of the first stratum boundary, and only within this preset range are the locations where the resistivity gradient meets the preset conditions. Since the fuzzy region and the key region come from two independent data sources, image recognition and resistivity logging, respectively, there is a non-strict correspondence between the two in terms of depth range. Therefore, the nearest neighbor matching strategy is adopted during the replacement. That is, for each fuzzy region, the physical depth coordinates of the nearest key region are found within the preset search range of the fuzzy region, and the original boundary depth of the fuzzy region is replaced with the physical depth coordinates of the key region.

[0078] See Figures 6a-6c As shown, Figures 6a-6c This demonstrates the nearest neighbor matching and replacement process for the fuzzy region and the generation results of the second stratum boundary. Figure 6a , Figure 6b , Figure 6c It is formed by sequential connection. Figure 6a The first stratum boundary obtained from image recognition is marked by multiple black dots at the junction of sandstone and mudstone, indicating depth boundary points located by abrupt gradients. Adjacent depth boundary points are connected sequentially by line segments to form the first stratum boundary. Among them, no new depth boundary point was identified between the two adjacent depth boundary points at depths of 8.2 meters and 8.6 meters, and the contrast of adjacent image slices in the corrected core image corresponding to this line segment was lower than the preset contrast threshold. Therefore, this line segment was selected with a dashed box and marked as a blurred area. The line segments within the blurred area are shown as dashed lines, and the midpoint of this line segment (depth 8.4 meters) is the original boundary depth of the blurred area. Figure 6bThe curve shows the distribution of the resistivity gradient along the depth direction. Most sections of the curve lie within a preset gradient threshold (vertical dashed line in the figure), with a peak appearing only at a depth of 8.35 meters that exceeds the preset threshold. The depth coordinates of this peak are marked as the critical region. The two arrows from left to right in the figure represent the actions of "finding the critical region of resistivity gradient" and "nearest neighbor matching and replacing depth coordinates," respectively. Figure 6c To complete the second stratum boundary line obtained after nearest neighbor matching and replacement, the physical depth coordinate of the key area, 8.35 meters, is used as a new depth boundary point (shown as a dot with an outer ring in the figure) and inserted into the fuzzy area. Together with the original depth boundary point, they are connected in sequence to form the second stratum boundary line. The original fuzzy area is eliminated, and the boundary depth is accurately corrected from 8.4 meters to 8.35 meters.

[0079] In practical engineering applications, suppose a borehole is marked as an ambiguous region between a depth of 8.2 meters and 8.6 meters. Within this range, resistivity logging data shows a resistivity gradient value at a depth of 8.35 meters. If the value reaches 3.2 times the average value of the site, exceeding the preset gradient threshold, the boundary depth of the blurred area will be corrected from 8.4 meters given by the original image recognition to 8.35 meters.

[0080] This step supplements and corrects blurred areas in image recognition using resistivity gradient data, leveraging the complementarity of different data sources to further improve the positioning accuracy of rock strata boundaries in areas with minimal visual differences.

[0081] S4: Based on the borehole inclination, correct the spatial position of the second stratum boundary line, generate a basic borehole columnar section, and fill the basic borehole columnar section with the wave velocity value corresponding to each borehole depth to generate the target borehole columnar section.

[0082] In step S3, the second stratum boundary line after resistivity gradient correction has been obtained. Since the second stratum boundary line is calibrated based on the borehole depth, the actual borehole will bend and deviate during the drilling process due to the difference in stratum hardness and the deflection of the drill bit. Therefore, it is necessary to use the inclination to convert the second stratum boundary line based on the borehole depth into real three-dimensional spatial coordinates to reflect the actual trajectory inside the hole.

[0083] The specific implementation process for this step is as follows: First, based on the borehole inclination, the horizontal and vertical offsets at each sampling depth are calculated. Using these offsets, a three-dimensional coordinate transformation is performed on the second stratum boundary line to obtain a spatial boundary curve reflecting the actual borehole trajectory. Specifically, the inclinometer 104 records the azimuth and zenith angle at each sampling depth. The borehole depth increment between two adjacent sampling depths is used as the micro-segment length, and the three-dimensional coordinate increment of each micro-segment is calculated using the following formula:

[0084]

[0085]

[0086] in, This represents the horizontal offset in the east-west direction. This represents the horizontal offset in the north-south direction. This is the vertical offset. The length of the micro-segment represents the borehole depth increment between adjacent sampling depths. Zenith angle, It is the azimuth angle.

[0087] By accumulating the 3D coordinate increments of each micro-segment segment starting from the borehole opening, the 3D spatial coordinates at each sampling depth within the borehole can be obtained. Specifically, let the 3D spatial coordinates of the borehole opening be (X0, Y0, Z0), and the coordinate increment of the first micro-segment be (dx1, dy1, dz1). Then, the coordinates of the end of the first micro-segment are (X0+dx1, Y0+dy1, Z0-dz1), where the Z coordinate is subtracted because the depth direction is vertically downward; the second micro-segment... If the coordinate increment is (dx2, dy2, dz2), then the coordinates of the end of the second micro-segment are (X0+dx1+dx2, Y0+dy1+dy2, Z0-dz1-dz2), and so on. The three-dimensional spatial coordinates of the end of the nth micro-segment, i.e., at depth dn, are Xn=X0+dx1+dx2+...+dxn, Yn=Y0+dy1+dy2+...+dyn, and Zn=Z0-(dz1+dz2+...+dzn).

[0088] The specific accumulation process is as follows: Let the three-dimensional spatial coordinates of the borehole be (X0, Y0, Z0), then the three-dimensional spatial coordinates at the nth sampling depth are Xn=X0+dx1+dx2+...+dxn, Yn=Y0+dy1+dy2+...+dyn, Zn=Z0-(dz1+dz2+...+dzn), where the minus sign indicates that the depth direction is vertically downward; after obtaining the three-dimensional spatial coordinates of each sampling depth, find the corresponding sampling depth interval for each depth boundary point in the second rock stratum boundary line, and calculate the three-dimensional spatial coordinates of the depth boundary point by linear interpolation. For each pair of adjacent different rock types, connect the three-dimensional spatial coordinates of the depth boundary points located at the junction of the adjacent different rock types in sequence to form the spatial boundary curve between the adjacent different rock types.

[0089] In practical engineering applications, assume the coordinates of a borehole opening ( , , The zenith angle recorded by the inclinometer 104 within a micro-segment at depths of 3.0 to 3.5 meters (1000 meters, 2000 meters, 50 meters) is given by the inclinometer 104. , If the length of the micro-segment is L = 0.5 meters, then , , Assuming the accumulated coordinates from the orifice to a depth of 3.0 meters are (1000.12 meters, 2000.08 meters, 47.05 meters), then the three-dimensional spatial coordinates at a depth of 3.5 meters are X = 1000.12 + 0.0185 = 1000.1385 meters, Y = 2000.08 + 0.0185 = 2000.0985 meters, and Z = 47.05 - 0.4993 = 46.5507 meters.

[0090] Then, the geophysical exploration component 106 is used to transmit and receive acoustic signals in the borehole, extract the propagation time difference of the acoustic signals at different stratum depths, and calculate the wave velocity value at the corresponding depth based on the propagation time difference; the acoustic wave transmitting probe transmits pulsed acoustic waves to the stratum, and the acoustic waves are received by the acoustic wave receiving probe after propagating in the stratum. The calculation formula for the wave velocity value is shown in equation (4).

[0091]

[0092] In equation (4): V is the wave velocity value. The distance between the acoustic wave transmitting probe and the acoustic wave receiving probe is T, and the time difference of acoustic wave propagation is T. When the on-site equipment conditions support the synchronous acquisition of wave velocity during drilling, the synchronously acquired wave velocity data can be used directly. In this step, the independent measurement of wave velocity by the geophysical component 106 is an alternative solution when the conditions for synchronous acquisition are not available.

[0093] Furthermore, based on the spatial boundary curve, the depth range within the borehole is segmented and labeled to generate a basic borehole columnar section displaying the lithological distribution characteristics of the formation; see also Figure 7a As shown, the basic borehole columnar section is used to display the lithological distribution characteristics of strata within a single borehole. It uses depth (m) as the vertical axis (range 0 to 15) and stratum type as the horizontal axis, visually presenting the vertical distribution and thickness of each soil layer through columnar strips. The thickness of the fill layer is 2.5m (depth 0.0~2.5m), the sandstone layer is 2.6m (depth 2.5~5.1m), the mudstone layer is 3.25m (depth 5.1~8.35m), the granite layer is 3.65m (depth 8.35~12.0m), and the limestone layer is 3.0m (depth 12.0~15.0m). The basic borehole columnar section is a columnar data structure with depth as the vertical axis and stratum segments as the content. Each stratum segment includes start and end depths and rock type labels.

[0094] Finally, the wave velocity values ​​are mapped to the corresponding depth positions within the segments of the basic borehole columnar section, and the physical property parameters are filled to generate the target borehole columnar section.

[0095] The specific mapping process is as follows: Wave velocity values ​​measured by the geophysical component 106 at each depth are aligned and matched with the start and end depths of each stratigraphic segment in the basic borehole columnar section. For each stratigraphic segment, all wave velocity values ​​falling within the start and end depth range of that segment are extracted, and their arithmetic mean is taken. This arithmetic mean is written into the basic borehole columnar section as the wave velocity attribute value of that stratigraphic segment. After writing the wave velocity attribute values ​​for all stratigraphic segments, the basic borehole columnar section is transformed into the target borehole columnar section. The target borehole columnar section is an extended columnar data structure with added wave velocity attribute values ​​based on the basic borehole columnar section. Each stratigraphic segment not only contains a rock type label but also a wave velocity attribute value characterizing the elastic wave propagation characteristics of that stratigraphic segment. For example, see... Figure 7b As shown, the depth range of the fill layer is 0.0 m to 2.5 m. The geophysical exploration component 106 measured wave velocities of 1150 m / s, 1200 m / s, 1250 m / s, and 1200 m / s at depths of 0.5 m, 1.0 m, 1.5 m, and 2.0 m within this range, respectively. Therefore, the wave velocity attribute value of the fill layer is (1150 + 1200 + 1250 + 1200) / 4 = 1200 m / s; the sandstone layer... The depth range is 2.5 meters to 5.1 meters. Wave velocities measured at depths of 3.0 meters, 3.5 meters, 4.0 meters, 4.5 meters, and 5.0 meters were 2700 m / s, 2800 m / s, 2850 m / s, 2800 m / s, and 2850 m / s, respectively. Therefore, the wave velocity attribute value of the sandstone layer is (2700+2800+2850+2800+2850) / 5 = 2800 m / s; the depth range of the mudstone layer... The circumference ranges from 5.1 meters to 8.35 meters, with measured wave velocities of 2050 m / s, 2100 m / s, 2150 m / s, and 2100 m / s. Therefore, the wave velocity attribute value of the mudstone layer is (2050 + 2100 + 2150 + 2100) / 4 = 2100 m / s. The granite layer has a depth range of 8.35 meters to 12.0 meters, with measured wave velocities of 4400 m / s, 4500 m / s, and 4600 m / s. The wave velocity of the granite layer is 4500 m / s, and the measured wave velocity is 3750 m / s, 3800 m / s, and 3850 m / s. Therefore, the wave velocity of the limestone layer is (4400+4500+4600+4500) / 4=4500 m / s. The depth range of the limestone layer is 12.0 m to 15.0 m, and the measured wave velocity values ​​are 3750 m / s, 3800 m / s, and 3850 m / s. Therefore, the wave velocity of the limestone layer is (3750+3800+3850) / 3=3800 m / s.

[0096] This step converts the second stratum boundary line from borehole depth coordinates to true three-dimensional spatial coordinates through tilt correction, and adds physical and mechanical property information to the borehole columnar section by filling in wave velocity values, so that the subsequent three-dimensional stratigraphic model has dual geometric and physical attributes.

[0097] S5: Calculate the degree of undulation of the same strata among multiple boreholes, adjust the calculation weight of the spatial interpolation algorithm according to the degree of undulation, and use the spatial interpolation algorithm with adjusted calculation weights to calculate the columnar section of multiple target boreholes to generate three-dimensional strata data for the entire site.

[0098] In step S4, target borehole columnar sections have been obtained for multiple boreholes. Each target borehole columnar section only records the stratigraphic information at the location of a single borehole. However, the engineering design requires understanding the stratigraphic distribution across the entire site. Therefore, it is necessary to extrapolate the stratigraphic distribution between multiple boreholes using a spatial interpolation algorithm. However, traditional spatial interpolation algorithms use fixed distance attenuation parameters, which can easily produce overly smoothed extrapolation results in areas with severe stratigraphic undulations. Therefore, it is necessary to adaptively adjust the computational weights of the interpolation algorithm according to the actual degree of stratigraphic undulation.

[0099] See Figure 8 As shown, the specific implementation process of this step is as follows: First, obtain the coordinates of the top plate of the strata of the same rock type in the columnar diagrams of multiple target boreholes. The coordinates of the top plate of the strata refer to the position coordinates of the top surface of the strata of a specific rock type in each borehole in three-dimensional space. Specifically, the method is to extract the shallowest depth position corresponding to the same rock type label from the columnar diagrams of each target borehole, and then convert the depth position into three-dimensional space coordinates according to the mapping relationship between depth and three-dimensional space coordinates established in step S4, which is the coordinates of the top plate of the strata.

[0100] For example, in the target borehole columnar section of borehole A, the sandstone strata are distributed starting from a depth of 2.5 meters. The borehole opening coordinates of borehole A are (1000 meters, 2000 meters, 50 meters). According to the segment-by-segment cumulative calculation in step S4, the cumulative horizontal offset values ​​of each micro-segment from the borehole opening to a depth of 2.5 meters are dx_sum = 0.02 meters and dy_sum = 0.01 meters, and the cumulative vertical offset value is dz_sum = 2.50 meters. Therefore, the three-dimensional spatial coordinates at a depth of 2.5 meters are X = 1000 + 0.02 = 1000.02 meters, Y = 2000 + 0.01 = 2000.01 meters, and Z = 50 - 2.50 = 47.50 meters. Thus, the coordinates of the top plate of the sandstone strata in borehole A are (1000.02 meters, 2000.01 meters, 47.50 meters).

[0101] Then, calculate the elevation difference of the top plate coordinates of adjacent boreholes and the horizontal spacing between boreholes; where the elevation difference is the absolute value of the difference in elevation between the top plates of the same rock type in two adjacent boreholes, and the horizontal spacing between boreholes is the Euclidean distance between the two boreholes on the horizontal plane.

[0102] Then, the relative ratio of the elevation difference to the horizontal spacing of the boreholes is calculated to obtain the degree of undulation; the calculation formula is as follows: Where S represents the degree of fluctuation, This represents the elevation difference. The horizontal spacing between boreholes is S; when S is large, it indicates that the strata in this area have undergone a large elevation change within a short distance.

[0103] Next, an inverse proportional function between the fluctuation level and the preset interpolation search radius is constructed to generate the interpolation weight adjustment coefficient; wherein, the preset interpolation search radius... This is used to control the search range of the spatial interpolation algorithm in the horizontal direction, and can be set according to actual needs, such as the one used in this application. It is 1.5 times the average spacing of boreholes in the site; the inverse proportional function is shown in equation (5).

[0104]

[0105] In equation (5): This is the interpolation weight adjustment coefficient. To preset the interpolation search radius, For the degree of fluctuation, The proportionality constant has the dimension of length (meters). The undulation degree S is the dimensionless ratio of the elevation difference to the horizontal spacing of the boreholes. k multiplied by S has the dimension of length (meters). This is compared with the preset interpolation search radius. The dimensions are consistent, and it is used to measure the degree of dimensionless fluctuations. Convert to preset interpolation search radius The same length units ensure that the denominator of equation (5) contains the same length units. and The dimensions are consistent; the value of k can be set according to actual needs, for example, this application uses k=10 meters.

[0106] Then, the interpolation weighting coefficients are adjusted. The distance decay factor of the spatial interpolation algorithm is modified to obtain the spatial interpolation algorithm with adjusted calculation weights.

[0107] The spatial interpolation algorithm used in this application is the inverse distance weighted method. Other algorithms, such as Kriging interpolation, can be selected according to actual needs. Taking the inverse distance weighted method as an example, its original calculated weights are: ;in This represents the weights calculated based on distance in the inverse distance weighting method, where d is the horizontal distance between the interpolation point and the known borehole, and p is the distance attenuation factor, typically p=2; the corrected distance attenuation factor is: .when When it is small, the corrected distance attenuation factor Increasing the distance causes the contribution weight of boreholes that are farther away to the interpolation result to decay more quickly.

[0108] In practical engineering applications, assuming the horizontal distance between boreholes A and B on site is 30 meters, and the elevation difference between the top of the granite strata in the two boreholes is 4.5 meters, then S = 4.5 / 30 = 0.15; assuming... =45 meters, k=10, then substituting into equation (5) yields =45 / (45+10×0.15)=45 / 46.5=0.968, after correction =2 / 0.968=2.066.

[0109] If the horizontal distance between boreholes C and D is 30 meters, but their elevation difference is 12 meters, then S = 12 / 30 = 0.4. =45 / (45+10×0.4)=45 / 49=0.918, after correction =2 / 0.918=2.178; it can be seen that the greater the fluctuation, the greater the distance to the attenuation factor. The larger the value, the more the interpolation result depends on the data from nearby boreholes.

[0110] Finally, a spatial interpolation algorithm with adjusted computational weights is used to perform three-dimensional spatial interpolation on the wave velocity values ​​and coordinates of multiple target borehole columnar sections to generate three-dimensional formation data.

[0111] The specific process involves establishing a three-dimensional spatial grid across the entire site. The horizontal resolution of the grid can be set according to actual needs; for example, this application uses a horizontal grid spacing of 1 meter. For each node in the grid, an interpolation search is performed with that node as the center, within a preset interpolation search radius. All known boreholes within the range are searched, and the contribution weight of each known borehole to the node is calculated using the modified inverse distance weighting method. The wave velocity values ​​of each borehole at the corresponding depth are weighted and summed to obtain the interpolated wave velocity value at the node. At the same time, based on the rock type label of each borehole at the corresponding depth, the rock type label of the borehole with the largest contribution weight is used as the stratigraphic type label of the node.

[0112] In practical engineering applications, see Figure 9 As shown, assuming the node to be interpolated, P, is 15 meters from borehole A and 25 meters from borehole B, the corrected distance attenuation factor is... =2.066, then the weight of borehole A is... Weight of borehole B After normalization , If the wave velocity at this depth is 3200 m / s for borehole A and 2800 m / s for borehole B, then the interpolated wave velocity at node P is 0.742×3200+0.258×2800=2374.4+722.4=3096.8 m / s.

[0113] This step improves the modeling effect in areas with severe geological undulations by adaptively adjusting the distance attenuation factor of the interpolation algorithm according to the degree of geological undulation, and avoids the over-smoothing problem of fixed parameter interpolation algorithms.

[0114] S6: Extract the spatial coordinates and corresponding wave velocity values ​​of the 3D stratigraphic data, import them into the BIM software interface to generate 3D solid components, and automatically assemble a 3D geological model.

[0115] In step S5, three-dimensional stratigraphic data containing stratigraphic information of the entire site's spatial grid nodes has been generated. However, the three-dimensional stratigraphic data is only a discrete set of spatial points at the data level and has not yet formed a three-dimensional geometric entity that can be directly used in engineering design. Therefore, it is necessary to import the three-dimensional stratigraphic data into BIM software to convert it into three-dimensional solid components and assemble them into a complete three-dimensional geological model. The three-dimensional geological model is a three-dimensional geometric and physical property integrated model covering the entire exploration site, which is spliced ​​together by multiple three-dimensional solid components according to the spatial positional relationship of the stratigraphic layers.

[0116] The specific implementation process for this step is as follows: First, the stratigraphic contour boundary of the 3D stratigraphic data is analyzed, and the spatial coordinates of the stratigraphic contour boundary are extracted. Specifically, for each rock type in the 3D stratigraphic data, all nodes belonging to that rock type in the spatial grid are traversed, and the nodes located at the distribution edge of that rock type are extracted as boundary nodes. The 3D spatial coordinate sequence of the boundary nodes constitutes the stratigraphic contour boundary.

[0117] Then, spatial geometric meshing is performed based on spatial coordinates to generate a closed three-dimensional stratigraphic shell. The spatial geometric meshing uses the Delaunay triangulation algorithm, taking the spatial coordinate points of the stratigraphic contour boundary as the input nodes for triangulation, generating a closed surface composed of triangular facets. This closed surface is the three-dimensional stratigraphic shell. During the meshing process, non-manifold boundaries need to be checked and eliminated. Specifically, edges with a shared number of triangular facets not equal to 2 are detected, and adjacent triangular facets of edges with a shared number of not equal to 2 are re-meshed or deleted to ensure the topological closure of the three-dimensional stratigraphic shell.

[0118] Then, the corresponding wave velocity value is mapped as a physical attribute parameter to the 3D geological shell. The mapped 3D geological shell is instantiated into a 3D solid component through the BIM software interface. Specifically, each triangular facet of the 3D geological shell is bound to the interpolated wave velocity value at its spatial location, i.e., each triangular facet is associated with a wave velocity attribute value. Then, the 3D geological shell with the bound wave velocity attribute value is imported into the BIM software through the BIM software's industrial basic standard data exchange interface. The BIM software converts each 3D geological shell with a wave velocity attribute value into a BIM component instance. The industrial basic standard data exchange interface is a common open data standard interface in the construction industry, which supports the unified encapsulation and transmission of geometric and attribute information.

[0119] Finally, the relative positional relationship of multiple three-dimensional solid components is determined based on spatial coordinates. The three-dimensional solid components are then spatially spliced ​​according to their relative positional relationship to automatically assemble a three-dimensional geological model. Specifically, the three-dimensional solid components are spatially spliced ​​according to the vertical stacking relationship and horizontal extension range of the strata. During splicing, spatial coordinates are used as the alignment reference to ensure that the interface of adjacent three-dimensional solid components remains geometrically continuous. After splicing, a three-dimensional geological model covering the entire exploration site is formed.

[0120] This step automatically generates and assembles 3D solid components by importing 3D stratigraphic data into the BIM software interface, thus achieving automated conversion from discrete exploration data to a 3D geological model.

[0121] After the automatic assembly of the three-dimensional geological model is completed, this application also includes the following early warning function processing steps: First, weak interlayers with a thickness less than a preset thickness value are extracted from the three-dimensional geological model. These weak interlayers refer to thin, low-strength strata existing between relatively hard rock layers. Their physical and mechanical properties are significantly weaker than the adjacent main rock layers, making them prone to forming unfavorable sliding surfaces during engineering excavation. The preset thickness value can be set according to actual needs; for example, this application uses a preset thickness value of 2 meters. Wave velocity ranges are used as auxiliary judgment conditions. When the thickness of a certain stratum segment is less than 2 meters and the wave velocity attribute value within that segment is less than 60% of the average wave velocity of the site, that stratum segment is marked as a weak interlayer.

[0122] Then, the spatial foundation pit outline data of the proposed project is superimposed onto the three-dimensional geological model. The spatial foundation pit outline data is the coordinate sequence of the design boundary line of the foundation pit of the proposed project in three-dimensional space. When superimposing, the coordinates of the measurement control points are used as the alignment reference to unify the spatial foundation pit outline data and the three-dimensional geological model under the same spatial reference coordinate. The tolerance of coordinate alignment can be set according to actual needs. For example, this application adopts a coordinate tolerance of ±5 mm.

[0123] Then, the spatial intersection coordinates of the spatial foundation pit outline data and the weak interlayer are extracted. Specifically, the geometric intersection between the three-dimensional curved surface of the foundation pit excavation range and the three-dimensional stratum shell of the weak interlayer is calculated. The spatial intersection coordinates describe the position and range of the intersection between the weak interlayer and the foundation pit excavation surface.

[0124] Finally, warning color rendering is performed on the corresponding components in the 3D geological model based on the spatial intersection coordinates; see [link / reference]. Figure 10 As shown, the early warning effect after overlaying the 3D geological model with the pit outline is presented in cross-sectional view. The surface line is located at the top of the figure, and below the surface, from top to bottom, the layers are: fill layer, sandstone layer, first weak interlayer, granite layer, second weak interlayer, and limestone layer. The pit outline is marked with a thick black line, and the rectangular area enclosed by the two side walls of the pit and the bottom surface of the pit represents the excavation range of the pit. The first weak interlayer is marked with a dark fill, and the intersection with the side wall of the pit is marked as a high-risk area. The second weak interlayer is marked with a light fill, and the intersection with the bottom surface of the pit is marked as a medium-risk area.

[0125] The color mapping rules for early warning color rendering are as follows: the area where the weak interlayer intersects with the sidewall of the foundation pit is rendered in dark color, indicating a high-risk area; the area where the weak interlayer intersects with the bottom of the foundation pit is rendered in light color, indicating a medium-risk area; after rendering, engineers can intuitively view the unfavorable geological areas that need to be focused on in the 3D geological model in the BIM software.

[0126] This step, through the identification of weak interlayers and the overlay analysis of the foundation pit outline, enables automatic early warning of potentially unfavorable geological conditions during engineering excavation, providing intuitive risk alerts for engineering design and construction decisions.

[0127] This application also provides an automated construction system for three-dimensional geological models for geotechnical engineering investigation based on machine learning and BIM. The system includes: The acquisition module is used to acquire borehole exploration data collected on-site, including original core images, as well as borehole inclination, resistivity, and wave velocity.

[0128] The input module is used to obtain a color-accurate corrected core image based on the original core image and the color chart image. The corrected core image is then input into a pre-trained convolutional neural network model to identify the rock type, analyze the variation characteristics of the image texture, and delineate the first rock layer boundary.

[0129] The search module is used to find key areas where the resistivity change of the borehole meets preset conditions, and to use the key areas to replace and correct the blurred areas in the core image to obtain the second rock layer boundary.

[0130] The correction module is used to correct the spatial position of the second stratum boundary line according to the borehole inclination, generate a basic borehole columnar section, fill the basic borehole columnar section with the wave velocity value corresponding to each borehole depth, and generate the target borehole columnar section.

[0131] The calculation module is used to calculate the degree of undulation of the same strata among multiple boreholes, adjust the calculation weight of the spatial interpolation algorithm according to the degree of undulation, and use the spatial interpolation algorithm with adjusted calculation weights to calculate the columnar section of multiple target boreholes to generate three-dimensional stratigraphic data for the entire site.

[0132] The extraction module is used to extract the spatial coordinates and corresponding wave velocity values ​​of 3D stratigraphic data, and import them into the BIM software interface to generate 3D solid components, automatically assembling a 3D geological model.

[0133] The automated construction system for three-dimensional geological models for geotechnical engineering exploration based on machine learning and BIM in this application embodiment is used to implement the aforementioned automated construction method for three-dimensional geological models for geotechnical engineering exploration based on machine learning and BIM. Therefore, the specific implementation of the automated construction system for three-dimensional geological models for geotechnical engineering exploration based on machine learning and BIM can be found in the embodiment section of the automated construction method for three-dimensional geological models for geotechnical engineering exploration based on machine learning and BIM mentioned above. The specific implementation can be referred to the description of the corresponding embodiments, and will not be repeated here.

[0134] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the above-described automated construction method for three-dimensional geological models for geotechnical engineering exploration based on machine learning and BIM.

[0135] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described automated construction methods for three-dimensional geological models for geotechnical engineering investigation based on machine learning and BIM.

[0136] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0137] The embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the automated construction method for three-dimensional geological models for geotechnical engineering investigation based on machine learning and BIM.

[0138] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0139] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of related data must comply with relevant laws, regulations and standards, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0140] The above provides a detailed description of the automated construction method and system for three-dimensional geological models in geotechnical engineering investigation based on machine learning and BIM, as provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for automatically constructing three-dimensional geological models for geotechnical engineering investigation based on machine learning and BIM, characterized in that, include: Acquire borehole exploration data collected on-site, including original core images, as well as borehole inclination, resistivity, and wave velocity; Based on the original core image and color chart image, a color-accurate corrected core image is obtained. The corrected core image is then input into a pre-trained convolutional neural network model to identify the rock type and analyze the variation characteristics of the image texture to delineate the first rock layer boundary. Find the key areas where the resistivity change of the borehole meets the preset conditions, and use the key areas to replace the blurred areas in the corrected core image to obtain the second rock stratum boundary line; Based on the inclination of the borehole, the spatial position of the second rock stratum boundary line is corrected, a basic borehole columnar section is generated, and the wave velocity value corresponding to each borehole depth is filled into the basic borehole columnar section to generate a target borehole columnar section. The degree of undulation of the same stratum between multiple boreholes is calculated, the calculation weight of the spatial interpolation algorithm is adjusted according to the degree of undulation, and the spatial interpolation algorithm with adjusted calculation weight is used to calculate the columnar section of multiple target boreholes to generate three-dimensional stratigraphic data of the entire site. The spatial coordinates and corresponding wave velocity values ​​of the three-dimensional stratigraphic data are extracted and imported into the BIM software interface to generate three-dimensional solid components, which are then automatically assembled into a three-dimensional geological model. The process of identifying key regions where the resistivity change in the borehole meets preset conditions, and using these key regions to replace blurred areas in the corrected core image to obtain the second stratum boundary line, includes: Extract the depth interval between adjacent depth boundary points in the first rock stratum boundary line, which is determined by the contrast of adjacent image slices of the corrected rock core image being less than a preset contrast threshold, and mark the depth interval as a blurred region. Calculate the resistivity gradient of the borehole along the drilling depth direction; Extract the physical depth coordinates where the resistivity gradient value is greater than a preset gradient threshold, and mark the physical depth coordinates as key areas; The second rock stratum boundary line is obtained by replacing the depth coordinates of the fuzzy region with the physical depth coordinates of the key region. The process of calculating the undulation of the same strata among multiple boreholes, adjusting the calculation weights of the spatial interpolation algorithm based on the undulation, and using the adjusted spatial interpolation algorithm to calculate the columnar sections of the multiple target boreholes to generate three-dimensional stratigraphic data for the entire site includes: Obtain the coordinates of the top plate of rock strata of the same rock category in multiple target borehole columnar diagrams; Calculate the elevation difference of the top plate coordinates of the rock strata between adjacent boreholes and the horizontal spacing between the boreholes; The degree of undulation is obtained by calculating the relative ratio between the elevation difference and the horizontal spacing of the boreholes; Construct an inverse proportional function between the fluctuation level and the preset interpolation search radius to generate interpolation weight adjustment coefficients; The distance attenuation factor of the spatial interpolation algorithm is corrected using the interpolation weight adjustment coefficient to obtain the spatial interpolation algorithm with adjusted calculation weights; A spatial interpolation algorithm with adjusted computational weights is used to perform three-dimensional spatial interpolation on the wave velocity values ​​and coordinates of multiple target borehole columnar sections to generate three-dimensional formation data.

2. The method according to claim 1, characterized in that, The step of correcting the spatial position of the second rock stratum boundary line based on the borehole inclination, generating a basic borehole columnar section, and filling the basic borehole columnar section with wave velocity values ​​corresponding to each borehole depth to generate a target borehole columnar section includes: Based on the borehole inclination, the horizontal and vertical offsets at each sampling depth are calculated. The horizontal and vertical offsets are then used to perform a three-dimensional coordinate transformation on the second rock stratum boundary line to obtain a spatial boundary curve that reflects the actual borehole trajectory. Using geophysical exploration components to transmit and receive acoustic signals within the borehole, the propagation time difference of the acoustic signals at different strata depths is extracted, and the wave velocity value at the corresponding depth is calculated based on the propagation time difference. Based on the spatial boundary curve, the depth range within the borehole is segmented and labeled to generate a basic borehole columnar section that displays the lithological distribution characteristics of the strata. The wave velocity values ​​are mapped to the corresponding depth positions within the segmented intervals of the basic borehole columnar section, and physical property parameters are filled to generate the target borehole columnar section.

3. The method according to claim 1, characterized in that, The process of obtaining a color-accurate corrected core image based on the original core image and the colorimetric chart image includes: Locate the center point of the color patch in the color chart image, and extract the standard reference color of the center point of the color patch; By comparing the standard reference color with the preset actual target color, the color difference compensation amount for different color channels is calculated; For each color channel, a polynomial fitting is performed with the actual target color value of each color block on the color chart in the color channel as the independent variable and the color difference compensation amount of the corresponding color block in the color channel as the dependent variable to construct a color mapping table for the color channel. The color mapping table is a polynomial function that takes the channel value of any pixel in the color channel as input and outputs the color difference compensation amount of the pixel in the color channel. By iterating through each pixel in the original core image, the red, green, and blue channel values ​​of the pixel are substituted into the color mapping table of the corresponding color channel to calculate the color difference compensation amount of the pixel in each color channel. The color difference compensation amount of each color channel is then superimposed on the original channel value of the corresponding color channel to obtain the corrected core image.

4. The method according to claim 1, characterized in that, The step of inputting the corrected core image into a pre-trained convolutional neural network model to identify the rock type, analyze the variation characteristics of image texture, and delineate the first rock stratum boundary includes: The corrected core image is divided into continuously arranged image slices according to a preset depth step size; Visual features of each image slice are extracted using a convolutional neural network model, and category matching is performed to determine the rock type of each image slice. By comparing the rock types of two adjacent image slices, within the interval where the rock type changes, the color histogram and texture gray-level co-occurrence matrix of the two adjacent image slices are extracted within the interval, and the distance between the two color histograms and the contrast difference between the two texture gray-level co-occurrence matrices are calculated to obtain the color histogram distance and texture contrast difference. The color histogram distance and the texture contrast difference are weighted and summed to obtain the jump gradient. When the jump gradient is greater than a preset jump gradient threshold, the boundary depth position of the two adjacent image slices is determined as the depth limit point. For each pair of adjacent different rock types, multiple depth boundary points located at the junction of the adjacent different rock types are connected in sequence to generate the first rock stratum boundary line between adjacent different rock types.

5. The method according to claim 1, characterized in that, The process of extracting the spatial coordinates and corresponding wave velocity values ​​of the three-dimensional stratigraphic data, importing them into the BIM software interface to generate three-dimensional solid components, and automatically assembling a three-dimensional geological model includes: The stratigraphic contour boundary of the three-dimensional stratigraphic data is analyzed, and the spatial coordinates of the stratigraphic contour boundary are extracted; Based on the spatial coordinates, a spatial geometric mesh is constructed to generate a closed three-dimensional stratigraphic shell; The corresponding wave velocity value is mapped as a physical property parameter to the three-dimensional geological shell. The mapped three-dimensional geological shell is instantiated through the interface of the BIM software to generate a three-dimensional solid component. The relative positional relationship of multiple three-dimensional solid components is determined based on the spatial coordinates, and spatial splicing is performed on the multiple three-dimensional solid components according to the relative positional relationship to automatically assemble a three-dimensional geological model.

6. The method according to claim 1, characterized in that, The specific process of obtaining raw core images includes: Fix the color chart with the standard colors onto the core tray to be photographed; Industrial cameras were used to capture raw field images, including rock cores and color charts. The actual shooting distance between the industrial camera and the core tray is obtained using a laser rangefinder. The original scene images are scaled according to the actual shooting distance to obtain the original core images.

7. The method according to claim 1, characterized in that, After the step of automatically assembling a three-dimensional geological model, the following steps are also included: Extract the weak interlayers in the three-dimensional geological model whose stratum thickness is less than the preset thickness value; The spatial foundation pit outline data of the proposed project is superimposed onto the three-dimensional geological model; Extract the spatial intersection coordinates between the spatial foundation pit outline data and the weak interlayer body; Based on the spatial intersection coordinates, the corresponding components in the three-dimensional geological model are subjected to early warning color rendering.

8. An automated system for constructing three-dimensional geological models for geotechnical engineering investigation based on machine learning and BIM, characterized in that, include: The acquisition module is used to acquire borehole exploration data collected on site, including original core images, as well as borehole inclination, resistivity and wave velocity. The input module is used to obtain a color-accurate corrected core image based on the original core image and the color chart image, input the corrected core image into a pre-trained convolutional neural network model to identify the rock type, analyze the variation characteristics of the image texture, and delineate the first rock layer boundary line. The search module is used to find key areas where the resistivity change of the borehole meets preset conditions, and use the key areas to replace the blurred areas in the corrected core image to obtain the second stratum boundary line. The correction module is used to correct the spatial position of the second rock stratum boundary line according to the inclination of the borehole, generate a basic borehole columnar section, and fill the wave velocity value corresponding to each borehole depth in the basic borehole columnar section to generate a target borehole columnar section. The calculation module is used to calculate the degree of undulation of the same stratum among multiple boreholes, adjust the calculation weight of the spatial interpolation algorithm according to the degree of undulation, and use the spatial interpolation algorithm with adjusted calculation weight to calculate the columnar diagrams of multiple target boreholes to generate three-dimensional stratum data of the entire site. The extraction module is used to extract the spatial coordinates and corresponding wave velocity values ​​of the three-dimensional stratigraphic data, and import them into the BIM software interface to generate three-dimensional solid components, and automatically assemble a three-dimensional geological model. The process of identifying key regions where the resistivity change in the borehole meets preset conditions, and using these key regions to replace blurred areas in the corrected core image to obtain the second stratum boundary line, includes: Extract the depth interval between adjacent depth boundary points in the first rock stratum boundary line, which is determined by the contrast of adjacent image slices of the corrected rock core image being less than a preset contrast threshold, and mark the depth interval as a blurred region. Calculate the resistivity gradient of the borehole along the drilling depth direction; Extract the physical depth coordinates where the resistivity gradient value is greater than a preset gradient threshold, and mark the physical depth coordinates as key areas; The second rock stratum boundary line is obtained by replacing the depth coordinates of the fuzzy region with the physical depth coordinates of the key region. The process of calculating the undulation of the same strata among multiple boreholes, adjusting the calculation weights of the spatial interpolation algorithm based on the undulation, and using the adjusted spatial interpolation algorithm to calculate the columnar sections of the multiple target boreholes to generate three-dimensional stratigraphic data for the entire site includes: Obtain the coordinates of the top plate of rock strata of the same rock category in multiple target borehole columnar diagrams; Calculate the elevation difference of the top plate coordinates of the rock strata between adjacent boreholes and the horizontal spacing between the boreholes; The degree of undulation is obtained by calculating the relative ratio between the elevation difference and the horizontal spacing of the boreholes; Construct an inverse proportional function between the fluctuation level and the preset interpolation search radius to generate interpolation weight adjustment coefficients; The distance attenuation factor of the spatial interpolation algorithm is corrected using the interpolation weight adjustment coefficient to obtain the spatial interpolation algorithm with adjusted calculation weights; A spatial interpolation algorithm with adjusted computational weights is used to perform three-dimensional spatial interpolation on the wave velocity values ​​and coordinates of multiple target borehole columnar sections to generate three-dimensional formation data.

Images