Method for rapidly acquiring parameters of structural plane in drill hole

Abstract

The invention discloses a method for rapidly acquiring parameters of a structural plane in a drill hole, and belongs to the technical field of intelligent and rapid detection of surrounding rock cracks in geotechnical engineering and mining engineering. Establishing a structural plane pixel recognition model by adopting a neural network; recognizing structural plane pixels of the image in the anchoring hole by using a structural plane pixel recognition model, and reducing the number of the structural plane pixels by using a refinement algorithm; marking the structural plane skeleton pixels and other pixels of the image in the anchoring hole, segmenting the structural plane skeleton pixels in each column, and dividing the continuous structural plane skeleton pixels in each column into one section; obtaining a structural plane skeleton expression; and according to the structural plane skeleton expression, outputting structural plane parameters including a structural plane position, an inclination angle and an inclination direction. According to the method, the problems of low field surrounding rock fracture detection speed and poor precision can be solved, the intelligent rapid recognition efficiency of the structural surface can be effectively improved by utilizing the technology, and important reference is provided for safety construction of geotechnical engineering, mining engineering and the like.

Description

Technical Field

[0001] The invention relates to the technical field of intelligent detection of surrounding rock cracks, and in particular to a method for quickly acquiring structural surface parameters in a borehole. Background Art

[0002] Geotechnical engineering, bridge and tunnel engineering, etc. have high requirements for geological exploration, especially focusing on the distribution characteristics of cracks, which have an important impact on safe construction. Borehole imaging technology has been widely used in the field of surrounding rock detection. The characteristics of the rock formation on the hole wall can be clearly seen through the anchor hole image. However, the analysis of rock formation characteristics mainly relies on the experience of professionals, which is greatly affected by subjective factors and may ignore some tiny cracks. In addition, the speed of artificial surrounding rock analysis is slow. There is an urgent need for an intelligent surrounding rock identification system that can intelligently and quickly analyze the surrounding rock characteristics to ensure safe construction.

[0003] Prior art CN113298140A discloses a method for quantitative evaluation and classification of the quality of rock mass under mining based on borehole images, which relates to the field of mining engineering and solid material measurement technology. The steps of this method include: drilling holes in the surrounding rock of the tunnel before mining, and observing the borehole images of the surrounding rock structure surface; dividing the filling closed structure surface, filling cracked structure surface, transverse cracked structure surface, longitudinal cracked structure surface and transverse crushing area; classifying and counting the structure surfaces on the borehole images, and determining the comprehensive width of the structure surfaces in each evaluation area; after the borehole is affected by mining, the borehole images of the surrounding rock structure surfaces are observed again, and the structure surface categories are divided, and classified and counted to determine the comprehensive width of the structure surfaces in each evaluation area; the maximum value of the comprehensive width of the structure surface is compared, the quality of the mining rock mass is classified, and measures are taken to reinforce the surrounding rock of each classified rock mass. This method can only provide a rough classification, and the accuracy is low. Summary of the invention

[0004] In view of the shortcomings of the prior art, a method for quickly acquiring parameters of a structural surface in a borehole is provided, which has simple steps, is easy to use, and can quickly acquire the characteristics of the structural surface in the borehole.

[0005] In order to achieve the above technical objectives, the present invention discloses a method for quickly acquiring parameters of structural surfaces in a borehole, and the specific steps are as follows:

[0006] S100, collect a large number of on-site anchor hole images, use neural network to establish a structural surface pixel recognition model and complete the training;

[0007] S200, inputting the image inside the anchor hole into the trained structural surface pixel recognition model, using the structural surface pixel recognition model to identify the structural surface pixels of the image inside the anchor hole, using a thinning algorithm to reduce the number of structural surface pixels and maintain the skeleton of the original structural surface pixels, thereby obtaining an anchor hole image containing only structural surface skeleton pixels and other pixels;

[0008] S300, using a binomial value method to mark the structural surface skeleton pixels and other pixels of the image in the anchor hole, and segmenting each column of structural surface skeleton pixels by judging whether each column of structural surface skeleton pixels in the anchor hole image is continuous, thereby dividing the continuous structural surface skeleton pixels in each column into one segment;

[0009] Specifically, the coordinates of each column of structural surface skeleton pixels in the anchor hole image are stored in a matrix, and each column of structural surface skeleton pixels is judged one by one. If the row numbers corresponding to the structural surface skeleton pixels in the anchor hole image are continuous, they are not split. If they are not continuous, they are split at the discontinuity. The structural surface skeleton pixels of each segment after the split are stored in different matrices, and the row numbers corresponding to the structural surface skeleton pixels in each matrix are continuous.

[0010] S400, judging whether two columns of structural surface skeleton pixels are adjacent one by one, and merging the two columns of structural surface skeleton pixels if the two columns of structural surface skeleton pixels are adjacent, until the merged structural surface skeleton pixels are not adjacent to the next column of structural surface skeleton pixels, or the structural surface skeleton pixels intersect, and the extraction of a structural surface skeleton pixel segment is completed; the coordinates corresponding to each structural surface skeleton pixel in all structural surface skeleton pixel segments are extracted and stored in different matrices, that is, the structural surface skeleton pixel coordinates of each structural surface skeleton pixel segment correspond to a matrix, which is recorded as a structural surface pixel segment coordinate matrix;

[0011] S500, fitting the coordinates of all structural surface skeleton pixel segments using a sine function model, the larger the fitting coefficient, the greater the probability that these structural surface skeleton pixel segments constitute the structural surface skeleton, and finally determining which structural surface pixel segment coordinate matrices correspond to coordinates that can constitute the structural surface skeleton, and the obtained sine function model expression is the structural surface skeleton expression;

[0012] S600, outputting structural surface parameters in the borehole according to the structural surface skeleton expression.

[0013] Furthermore, a large number of anchor hole images are collected on site, a neural network is used to establish a structural surface pixel recognition model, a large number of anchor hole images are obtained through a drilling imager, the structural surfaces in the anchor holes are annotated with structural surface pixels and a database is formed, and then a neural network is used for training to obtain a structural surface pixel recognition model.

[0014] Furthermore, the binomial method is used to annotate the structural surface skeleton pixels and other pixels in the image of the anchor hole. By judging whether each column of structural surface skeleton pixels is continuous, each column of structural surface skeleton pixels is segmented. The specific process of dividing the continuous structural surface skeleton pixels in each column into one segment is as follows:

[0015] a. The skeleton pixels of the structural surface are represented by the value "0", and the other pixels are represented by "1";

[0016] b. Obtain the position coordinates of the "0" value pixels, and store these pixel coordinates in a two-dimensional matrix R, where the first column coordinates of the two-dimensional matrix R correspond to the row number of the "0" value pixel coordinates in the anchor hole image, and the second column coordinates of the two-dimensional matrix R correspond to the column number of the "0" value pixel coordinates in the anchor hole image;

[0017] c. If there are multiple "0" values ​​in the same column of the anchor hole image in the two-dimensional matrix R, resulting in multiple repeated column numbers in the anchor hole image in the second column of the matrix R, then copy the second column of the two-dimensional matrix R to the matrix R 1 In the case of matrix R 1 When there are repeated values ​​in , only one value is retained, using the matrix R 1 Store the column number of the structural surface skeleton pixel in the anchor hole image to obtain the matrix R 1 The number of elements in n;

[0018] d. Using the matrix R with n elements 1 , create an empty cell B with n rows and 1 column 0 Then, the pixel coordinates of the structural surface skeleton with the same value in the second column of the two-dimensional matrix R are copied to the cell B 0 In the 0 It contains multiple matrices, each of which stores a column of structural surface skeleton pixel coordinates of the anchor hole image;

[0019] e. According to whether the vertical coordinates of the structural surface skeleton pixels in each column of the anchor hole image are continuous, the cell B 0 The matrix in is split, and each column of structural surface skeleton pixels with continuous ordinates is divided into a matrix. If there is only one structural surface skeleton pixel matrix with continuous ordinates in the i-th column of the matrix, then the matrix is ​​represented as B 0 {i}(1); if the i-th column is divided into m structural surface skeleton pixel matrices with continuous ordinates, then these matrices are represented as B 0 {i}(1), B 0 {i}(2), ... B 0 {i}(m); thereby storing the continuous structural surface skeleton pixel coordinate information in each column separately in a matrix.

[0020] Furthermore, the specific process of judging whether two columns of structural surface skeleton pixels are adjacent is to compare the two matrices using formula (1):

[0021]

[0022] In the formula, x 1 and x 2 Respectively represent the horizontal coordinates corresponding to the crack pixel positions in the two matrices; a and bThey represent the ordinates corresponding to the crack pixel positions in the two matrices respectively; the abscissas of the coordinates stored in each matrix are the same, but the ordinate values ​​are different; K represents the minimum distance between the two matrices;

[0023] When the minimum distance between two matrices The two matrices to be compared are judged to be adjacent: when the minimum distance between the two matrices The two matrices being compared are judged to be non-adjacent: if the structural surface skeleton pixels intersect, there must be a matrix whose minimum distance to the two or more matrices in the next column is less than or equal to Or there are two or more matrices in this column whose minimum distance to the same matrix in the next column is less than or equal to

[0024] Furthermore, the coordinates of the skeleton pixel segments of each structural surface are fitted through the sine function model, and finally the skeleton pixels of the structural surface are determined. The specific process is as follows:

[0025] a. Search the structural surface pixel segment coordinate matrix that contains the most structural surface skeleton pixels from the structural surface pixel segment coordinate matrix, and record it as matrix C 1 ;

[0026] b. Traverse all the structural surface skeleton pixel segment coordinate matrices. When the horizontal coordinate in the structural surface skeleton pixel segment coordinate matrix is ​​equal to the matrix C 1 When there are no repeated values ​​in the horizontal axis, the sine function model is used: The structural surface skeleton pixel segment coordinate matrix and matrix C 1 The coordinate points in are fitted; in the sine function model, A represents the amplitude of the structural surface curve, represents the initial phase of the structural surface curve, k represents the constant term of the structural surface curve, P 1 represents the lateral resolution of the anchor hole image;

[0027] c. Select the structural surface skeleton pixel segment coordinate matrix with the highest fitting coefficient greater than 0.9, and add all the coordinates in it to the matrix C. 1 ;

[0028] d. Repeat the traversal of all the structural surface skeleton pixel segment coordinate matrices, and store all the pixel coordinates that can form the structural surface skeleton in C 1 In the case of the horizontal coordinates in the structural surface skeleton pixel segment coordinate matrix and the matrix C 1 When there is no duplicate value in the horizontal coordinate, the coordinate matrix of the skeleton pixel segment of the structure surface is combined with the matrix C 1 The coordinate points in the matrix are fitted, and the structural surface skeleton pixel segment coordinate matrix with the highest fitting coefficient greater than 0.9 is selected, and all the coordinates in it are added to the matrix C 1 , until the matrix C 1Until the number of pixel coordinates of the structural surface skeleton in no longer increases;

[0029] e. Statistics of the current matrix C 1 In which columns of the anchor hole image are the corresponding structural surface skeleton pixels distributed? 1 The corresponding structural surface skeleton pixels account for more than 80% of the total number of columns in the anchor hole image, then C 1 The structural surface skeleton pixels corresponding to the matrix are regarded as a structural surface;

[0030] f. Since there may be multiple structural surfaces in the anchor hole image, repeat steps a to e to search for other structural surfaces. Delete the matrix C before each repetition of steps a to e. 1 The corresponding structural surface skeleton pixel segment coordinate matrix is ​​re-arranged into matrix C 1 Assignment;

[0031] g. Through steps a to f, determine whether there is still a structural surface skeleton pixel segment coordinate matrix. If not, end; if so, determine whether there are coordinates in the structural surface pixel segment coordinate matrix located at the boundary of the anchor hole. If so, determine whether there is a structural surface skeleton pixel segment coordinate matrix at the boundary of adjacent anchor hole images and the matrix also has coordinates located at the boundary of the anchor hole. Through steps b to e, determine whether the structural surface skeleton pixels corresponding to the structural surface skeleton pixel segment coordinate matrices at the boundaries of two adjacent anchor holes can constitute a structural surface.

[0032] Furthermore, according to the structural surface skeleton expression obtained by fitting, the output structural surface parameters include: structural surface inclination α, structural surface inclination angle β, and structural surface position L, and the formula is as follows:

[0033]

[0034] Where P 2 represents the lateral resolution and longitudinal resolution of the anchor hole image; D represents the anchor hole diameter; h represents the actual depth of the anchor hole; A represents the amplitude of the structural surface curve, represents the initial phase of the structural surface curve, and k represents the constant term of the structural surface curve.

[0035] A computer device comprising: a processor, a memory and a network interface;

[0036] The processor is connected to the memory and the network interface, wherein the network interface is used to provide a network communication function, the memory is used to store program codes, and the processor is used to call the program codes to execute a method for quickly acquiring parameters of structural surfaces within a borehole.

[0037] A computer-readable storage medium stores a computer program, which is suitable for being loaded by a processor and executing a method for quickly acquiring parameters of a structural surface in a borehole.

[0038] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0039] 1. The present invention proposes a new method for rapid extraction of structural surface parameters, which takes into account various complex situations such as the structural surface being affected by the rock layer interface, the structural surface being discontinuous, the structural surfaces intersecting with each other, and the structural surface being at the boundary of the anchor hole image. The recognition efficiency of this method is higher.

[0040] 2. The present invention proposes a method for determining whether two columns of structural surface skeleton pixels are adjacent, realizes the extraction of continuous structural surface skeleton pixels and structural surface skeleton pixel segments in each column, and lays a foundation for the subsequent determination of structural surface pixels.

[0041] 3. The present invention proposes a new method for rapid extraction of structural surface parameters, which can accurately obtain structural surface pixels and then obtain structural surface parameters. Compared with the traditional Hough transform method and sine function fitting method, the structural surface parameter extraction speed of this method is significantly improved, and the structural surface recognition rate is significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The following describes the method for quickly acquiring structural surface parameters of the present invention with reference to the accompanying drawings. In the accompanying drawings:

[0043] Figure 1 The figure is a flow chart of a method for quickly acquiring parameters of structural surfaces in a borehole according to the present embodiment.

[0044] Figure 2 Schematic diagram of the structural surface pixel determination process; wherein, (a) is a schematic diagram of the structural surface pixel recognition result of the anchor hole image; (b) is a schematic diagram of the structural surface pixel skeleton extraction result; (c) is a schematic diagram of the structural surface skeleton pixel segment extraction result; and (d) is a schematic diagram of the structural surface skeleton pixel segment matching. DETAILED DESCRIPTION

[0045] The embodiments of the present invention will be further described below in conjunction with the accompanying drawings:

[0046] like Figure 1 As shown, a method for quickly acquiring parameters of structural surfaces in a borehole comprises the following steps:

[0047] S100. Collect a large number of on-site anchor hole images, use a neural network to establish a structural surface pixel recognition model and complete the training.

[0048] A large number of anchor hole images are collected on-site using a borehole imager to form a sample database, which should include various rock formations such as sandstone, mudstone, sandy mudstone, limestone, etc. The structural surface pixels in the sample database are labeled, and a neural network is used for sample training to obtain a structural surface pixel recognition model.

[0049] S200, input the image inside the anchor hole into the trained structural surface pixel recognition model, use the structural surface pixel recognition model to identify the structural surface pixels of the image inside the anchor hole, use a refinement algorithm to reduce the number of structural surface pixels and maintain the skeleton of the original structural surface pixels, so as to obtain an anchor hole image containing only structural surface skeleton pixels and other pixels.

[0050] The anchor hole image is input into the trained structural surface pixel recognition model to identify the structural surface pixels. Figure 2 (a) shows a schematic diagram of the identified structural surface pixels. The black pixels in the figure are structural surface pixels, and the two structural surfaces intersect each other. Using a thinning algorithm to process the structural surface pixels can reduce the number of structural surface pixels, but keep the skeleton of the original structural surface pixels. Figure 2 (b) shows the pixel skeleton of the structural surface obtained by the refinement algorithm. Reducing the number of structural surfaces is beneficial to reducing the amount of subsequent calculations and improving the efficiency of structural surface recognition.

[0051] S300, using the binomial method to mark the structural surface skeleton pixels and other pixels of the image in the anchor hole, and segmenting each column of structural surface skeleton pixels by judging whether each column of structural surface skeleton pixels is segmented continuously, thereby dividing the continuous structural surface skeleton pixels in each column into one segment.

[0052] Specifically, the coordinates of each column of structural surface skeleton pixels in the anchor hole image are stored in a matrix, and each column of structural surface skeleton pixels is judged one by one. If the row numbers corresponding to the structural surface skeleton pixels in the anchor hole image are continuous, they are not split. If they are not continuous, they are split at the discontinuous position. The structural surface skeleton pixels of each segment after the split are stored in different matrices, and the row numbers corresponding to the structural surface skeleton pixels in each matrix are continuous. The specific steps include:

[0053] a. The skeleton pixels of the structural surface are represented by the value "0", and the other pixels are represented by "1";

[0054] b. Obtain the position coordinates of the "0" value pixel, and store these pixel coordinates in a two-dimensional matrix R. The first column coordinates of the matrix R correspond to the row number of the "0" value pixel coordinates in the anchor hole image, and the second column coordinates of the matrix R correspond to the column number of the "0" value pixel coordinates in the anchor hole image; R is a two-dimensional matrix used to store coordinate values;

[0055] c. Because there are multiple "0" values ​​in the same column of the anchor hole image, the values ​​in the second column of matrix R will have multiple repeated column numbers in the anchor hole image. Copy the values ​​in the second column of matrix R to matrix R 1 In the case of R 1 When there are repeated values ​​in , only one value is retained, that is, the matrix stores the column number of the anchor hole image where the structural surface skeleton pixel is located;

[0056] d. Get the matrix R 1 The number of elements in is n, create an empty cell B with n rows and 1 column 0 ; Then, the pixel coordinates of the structural surface skeleton with the same value in the second column of matrix R are copied to cell B respectively 0 At this time, cell B 0 It contains multiple matrices, each of which stores the pixel coordinates of a certain column of the structural surface skeleton of the anchor hole image;

[0057] e. According to whether the vertical coordinates of the structural surface skeleton pixels in each column of the anchor hole image are continuous, the cell B 0 The matrix in is split, and each column of structural surface skeleton pixels with continuous ordinates is divided into a matrix. If there is only one structural surface skeleton pixel matrix with continuous ordinates in the i-th column, the matrix is ​​represented as B 0 {i}(1); if the i-th column is divided into m structural surface skeleton pixel matrices with continuous ordinates, then these matrices are represented as B 0 {i}(1), B 0 {i}(2), ... B 0 {i}(m); thereby storing the continuous structural surface skeleton pixel coordinate information in each column separately in a matrix.

[0058] S400, judging whether two columns of structural surface skeleton pixels are adjacent one by one, and merging the two columns of structural surface skeleton pixels if they are adjacent, until the merged structural surface skeleton pixels are not adjacent to the next column of structural surface skeleton pixels, or the structural surface skeleton pixels intersect, and the extraction of a structural surface skeleton pixel segment is completed; the coordinates corresponding to each structural surface skeleton pixel in all structural surface skeleton pixel segments are extracted and stored in different matrices, that is, the structural surface skeleton pixel coordinates of each structural surface skeleton pixel segment correspond to a matrix, which is called the structural surface pixel segment coordinate matrix

[0059] Assume that the pixel coordinates of a structural face skeleton in a column are (x 1 ,y 1 ), if there is a structural face skeleton pixel in the next column with (x 1 ,y 1 ) positions are adjacent, there are three positions, divided into (x 1 +1,y 1)、(x 1 +1,y 1 +1) and (x 1 +1,y 1 -1), these three positions are respectively 1 ,y 1 ) is 1, and Therefore, when two columns of structure surface pixels are adjacent, the minimum distance between the two columns of structure surface pixels should be less than or equal to

[0060] The specific process of judging whether two columns of structural surface skeleton pixels are adjacent is to compare the two matrices using formula (1). It is considered adjacent: If the structural surface skeleton pixels intersect, there must be a matrix whose minimum distance to the next column of two or more matrices is less than or equal to Or there are two or more matrices in this column whose minimum distance to the same matrix in the next column is less than or equal to

[0061]

[0062] In the formula, x 1 and x 2 Respectively represent the horizontal coordinates corresponding to the crack pixel positions in the two matrices; a and b They represent the ordinates corresponding to the crack pixel positions in the two matrices respectively; the abscissas of the coordinates stored in each matrix are the same, but the ordinate values ​​are different; K is the minimum distance between the two matrices.

[0063] Structural surface skeleton pixel segment extraction means that if a segment of structural surface skeleton pixels is continuous and does not intersect with other segments of structural surface skeleton pixels, the coordinates of this segment of structural surface skeleton pixels are stored in a matrix. If a segment of continuous structural surface skeleton pixels intersects with other segments of structural surface skeleton pixels, the structural surface skeleton pixels are divided into multiple segments using the intersection point as the segmentation point, and the coordinates of each segment of structural surface skeleton pixels exist in different matrices. The implementation method of this idea is to determine whether two columns of structural surface skeleton pixels are adjacent column by column. If they are adjacent, they are merged into one structural surface skeleton pixel segment. If the structural surface skeleton pixels intersect or are not adjacent, the extraction of the previous structural surface skeleton pixel segment is completed, and the next structural surface skeleton pixel segment is extracted. Each structural surface skeleton pixel segment is stored in a matrix. Figure 2 As shown in (c), the two structural surface skeleton pixels are divided into 6 structural surface skeleton pixel segments, which are stored in the matrix A 1 , A 2 , A 3 , A 4 , A5 and A 6 middle.

[0064] S500. Use a sine function model to fit the coordinates of each structural surface skeleton pixel segment. The larger the fitting coefficient, the greater the probability that these structural surface skeleton pixel segments constitute the structural surface skeleton. Finally, it is determined which structural surface pixel segment coordinate matrixes correspond to coordinates that can constitute the structural surface skeleton. The resulting sine function model expression is the structural surface skeleton expression.

[0065] The specific process is as follows:

[0066] a. Using the sine function model: Search the structural surface pixel segment coordinate matrix containing the most structural surface skeleton pixels from the structural surface pixel segment coordinate matrix, and record it as matrix C 1 ;

[0067] b. Traverse all the structural surface skeleton pixel segment coordinate matrices. When the horizontal coordinate in the structural surface skeleton pixel segment coordinate matrix is ​​equal to the matrix C 1 When there are no repeated values ​​in the horizontal axis, the sine function model is used: The pixel segment coordinate matrix of the structure surface skeleton is combined with C 1 The coordinate points in the matrix are fitted; in the sine function model, A represents the amplitude of the structural surface curve, represents the initial phase of the structural surface curve, k represents the constant term of the structural surface curve, P 1 represents the lateral resolution of the anchor hole image;

[0068] c. Select the structural surface skeleton pixel segment coordinate matrix with the highest fitting coefficient greater than 0.9, and add all the coordinates in it to the matrix C. 1 ;

[0069] d. In order to store all pixel coordinates that can form the structural surface skeleton in the matrix C 1 In the process, all the structural surface skeleton pixel segment coordinate matrices are repeatedly traversed. When the horizontal coordinate in the structural surface skeleton pixel segment coordinate matrix is ​​consistent with the matrix C 1 When there is no duplicate value in the horizontal coordinate, the coordinate matrix of the skeleton pixel segment of the structure surface is combined with the matrix C 1 The coordinate points in the matrix are fitted, and the structural surface skeleton pixel segment coordinate matrix with the highest fitting coefficient greater than 0.9 is selected, and all the coordinates in it are added to the matrix C 1 , until the matrix C 1 Until the number of pixel coordinates of the structural surface skeleton in no longer increases;

[0070] e. Statistics of the current matrix C 1 In which columns of the anchor hole image are the corresponding structural surface skeleton pixels distributed? 1The corresponding structural surface skeleton pixels account for more than 80% of the total number of columns in the anchor hole image, then the matrix C 1 The corresponding structural surface skeleton pixels are regarded as a structural surface;

[0071] Assume that there are 7 segments of structural surface skeleton pixels in an anchor hole image, and one of the structural surfaces is composed of 4 segments of structural surface skeleton pixels (x 1 , x 2 , x 3 and x 4 ), select the segment x with the largest number of structural surface skeleton pixels among the four segments through step a 1 , through step b, the remaining 6 structural surface pixel segments are combined with x 1 Fitting is performed and x is determined through step c 2 , x 1 , x 2 The matrix C 1 , repeat step d for the first time and determine x 3 , X 3 Add to matrix C 1 , repeat the traversal for the second time, and change x 4 Add to matrix C 1 , when repeated, the matrix C 1 The coordinates of no longer increase, and the matrix C 1 x in 1 , x 2 , x 3 and x 4 is a structural surface;

[0072] f. There may be multiple structural surfaces in the anchor hole image. Repeat steps a to e to search for other structural surfaces. Delete C before repeating steps a to e each time. 1 The structural surface skeleton pixel segment coordinate matrix corresponding to the matrix;

[0073] g. Through steps a to f, determine whether there is still a structural surface skeleton pixel segment coordinate matrix. If not, end; if so, determine whether there are coordinates in the structural surface pixel segment coordinate matrix located at the boundary of the anchor hole. If so, determine whether there is a structural surface skeleton pixel segment coordinate matrix at the boundary of adjacent anchor hole images and the matrix also has coordinates located at the boundary of the anchor hole. Use the principle of the above method to determine whether the structural surface skeleton pixels corresponding to the structural surface skeleton pixel segment coordinate matrices at the boundaries of two adjacent anchor holes can constitute a structural surface.

[0074] According to this process, the structural surface skeleton pixel segment coordinate matrix A 1 There are 4 matching schemes, such as Figure 2 As shown in (d), the matrix A 1 and A 4 , matrix A1 and A 3 , matrix A 1 and A 5 , matrix A 1 and A 6 By fitting the above matching scheme with the sine function model, it can be confirmed that the matrix A 1 and A 4 , matrix A 1 and A 5 The fitting effect is good. 1 , A 4 and A 5 By fitting, all pixels of the structural surface skeleton can be determined. The remaining matrices can be matched and fitted using a similar method to determine A 2 , A 3 and A 6 Belongs to the second structural surface.

[0075] S600. Output structural surface parameters according to the structural surface skeleton expression.

[0076] After obtaining the structural surface skeleton expression, the structural surface inclination, tilt and position can be calculated through the following equations according to the sine function parameter values, thereby quickly outputting the structural surface parameters.

[0077]

[0078]

[0079] Where P 2 represents the lateral resolution and longitudinal resolution of the anchor hole image; D is the anchor hole diameter; h is the actual depth of the anchor hole; α and β are the structural surface inclination and dip, respectively; L represents the structural surface position.

Claims

1. A method for quickly acquiring parameters of structural surfaces in a borehole, characterized in that: The specific steps are as follows: S100, collect a large number of on-site anchor hole images, use neural network to establish a structural surface pixel recognition model and complete the training; S200, inputting the image inside the anchor hole into the trained structural surface pixel recognition model, using the structural surface pixel recognition model to identify the structural surface pixels of the image inside the anchor hole, using a thinning algorithm to reduce the number of structural surface pixels and maintain the skeleton of the original structural surface pixels, thereby obtaining an anchor hole image containing only structural surface skeleton pixels and other pixels; S300, using a binomial value method to mark the structural surface skeleton pixels and other pixels of the image in the anchor hole, and segmenting each column of structural surface skeleton pixels by judging whether each column of structural surface skeleton pixels in the anchor hole image is continuous, thereby dividing the continuous structural surface skeleton pixels in each column into one segment; Specifically, the coordinates of each column of structural surface skeleton pixels in the anchor hole image are stored in a matrix, and each column of structural surface skeleton pixels is judged one by one. If the row numbers corresponding to the structural surface skeleton pixels in the anchor hole image are continuous, they are not split. If they are not continuous, they are split at the discontinuity. The structural surface skeleton pixels of each segment after the split are stored in different matrices, and the row numbers corresponding to the structural surface skeleton pixels in each matrix are continuous. S400, judging whether two columns of structural surface skeleton pixels are adjacent one by one, and merging the two columns of structural surface skeleton pixels if the two columns of structural surface skeleton pixels are adjacent, until the merged structural surface skeleton pixels are not adjacent to the next column of structural surface skeleton pixels, or the structural surface skeleton pixels intersect, and the extraction of a structural surface skeleton pixel segment is completed; the coordinates corresponding to each structural surface skeleton pixel in all structural surface skeleton pixel segments are extracted and stored in different matrices, that is, the structural surface skeleton pixel coordinates of each structural surface skeleton pixel segment correspond to a matrix, which is recorded as a structural surface pixel segment coordinate matrix; S500, fitting the coordinates of all structural surface skeleton pixel segments using a sine function model, the larger the fitting coefficient, the greater the probability that these structural surface skeleton pixel segments constitute the structural surface skeleton, and finally determining which structural surface pixel segment coordinate matrices correspond to coordinates that can constitute the structural surface skeleton, and the obtained sine function model expression is the structural surface skeleton expression; S600, outputting structural surface parameters in the borehole according to the structural surface skeleton expression.

2. According to the method for quickly acquiring parameters of structural surfaces in boreholes according to claim 1, a large number of anchor hole images are collected on site, and a structural surface pixel recognition model is established using a neural network, characterized in that: A large number of anchor hole images are obtained through a borehole imager, and the structural surface pixels in the anchor holes are annotated to form a database. Then, a neural network is used for training to obtain a structural surface pixel recognition model.

3. A method for quickly acquiring parameters of structural surfaces in a borehole according to claim 1, characterized in that: The binomial method is used to annotate the structural surface skeleton pixels and other pixels in the image of the anchor hole. By judging whether each column of structural surface skeleton pixels is continuous, each column of structural surface skeleton pixels is segmented. The specific process of dividing the continuous structural surface skeleton pixels in each column into one segment is as follows: a. The skeleton pixels of the structural surface are represented by the value "0", and the other pixels are represented by "1"; b. Obtain the position coordinates of the "0" value pixels, and store these pixel coordinates in a two-dimensional matrix R, where the first column coordinates of the two-dimensional matrix R correspond to the row number of the "0" value pixel coordinates in the anchor hole image, and the second column coordinates of the two-dimensional matrix R correspond to the column number of the "0" value pixel coordinates in the anchor hole image; c. If there are multiple "0" values ​​in the two-dimensional matrix R located in the same column of the anchor hole image, resulting in multiple repeated column numbers in the anchor hole image in the values ​​of the second column of the matrix R, the values ​​of the second column of the two-dimensional matrix R are copied to the matrix R1. When there are repeated values ​​in the matrix R1, only one value is retained, and the matrix R1 is used to store the column number in the anchor hole image where the structural surface skeleton pixel is located, thereby obtaining the number of elements n in the matrix R1; d. Using the matrix R1 with n elements, create an empty cell B0 with n rows and 1 column, and then copy the structural surface skeleton pixel coordinates with the same value in the second column of the two-dimensional matrix R to the cell B0, so that the cell B0 contains multiple matrices, each of which stores a column of structural surface skeleton pixel coordinates of the anchor hole image; e. According to whether the vertical coordinates of the structural surface skeleton pixels in each column in the anchor hole image are continuous, the matrix in the cell B0 is split again, and the structural surface skeleton pixels with continuous vertical coordinates in each column are divided into matrices. If the i-th column in the matrix has only one structural surface skeleton pixel matrix with continuous vertical coordinates, the matrix is ​​represented as B0{i}(1); if the i-th column is divided into m structural surface skeleton pixel matrices with continuous vertical coordinates, these matrices are represented as B0{i}(1), B0{i}(2), ... B0{i}(m) respectively; thereby, the continuous structural surface skeleton pixel coordinate information in each column is stored separately in a matrix.

4. A method for quickly acquiring parameters of structural surfaces in a borehole according to claim 1, characterized in that: The specific process of judging whether two columns of structural surface skeleton pixels are adjacent is to compare the two matrices using formula (1): Where x1 and x2 represent the horizontal coordinates of the crack pixel positions in the two matrices respectively; y a and b They represent the ordinates corresponding to the crack pixel positions in the two matrices respectively; the abscissas of the coordinates stored in each matrix are the same, but the ordinate values ​​are different; K represents the minimum distance between the two matrices; When the minimum distance between two matrices Then the two matrices to be compared are judged to be adjacent: when the minimum distance between the two matrices The two matrices being compared are judged to be non-adjacent: if the structural surface skeleton pixels intersect, there must be a matrix whose minimum distance to the two or more matrices in the next column is less than or equal to Or there are two or more matrices in this column whose minimum distance to the same matrix in the next column is less than or equal to 5. A method for quickly acquiring parameters of structural surfaces in a borehole according to claim 1, characterized in that: The coordinates of the skeleton pixel segments of each structural surface are fitted through the sine function model, and finally the skeleton pixels of the structural surface are determined. The specific process is as follows: a. Search the structural surface pixel segment coordinate matrix containing the most structural surface skeleton pixels from the structural surface pixel segment coordinate matrix, and record it as matrix C1; b. Traverse all the structural surface skeleton pixel segment coordinate matrices. When the horizontal coordinates in the structural surface skeleton pixel segment coordinate matrix have no repeated values ​​with the horizontal coordinates in the matrix C1, use the sine function model: The coordinate matrix of the structural surface skeleton pixel segment is fitted with the coordinate points in the matrix C1; in the sine function model, A represents the amplitude of the structural surface curve, φ represents the initial phase of the structural surface curve, k represents the constant term of the structural surface curve, and P1 represents the lateral resolution of the anchor hole image; c. Select the structural surface skeleton pixel segment coordinate matrix with the highest fitting coefficient greater than 0.9, and add all the coordinates therein to the matrix C1; d. Repeatedly traverse all the structural surface skeleton pixel segment coordinate matrices, and store all the pixel coordinates that can constitute the structural surface skeleton in C1. When the horizontal coordinates in the structural surface skeleton pixel segment coordinate matrix and the horizontal coordinates in the matrix C1 have no repeated values, fit the structural surface skeleton pixel segment coordinate matrix with the coordinate points in the matrix C1, select the structural surface skeleton pixel segment coordinate matrix with the highest fitting coefficient greater than 0.9, and add all the coordinates therein to the matrix C1 until the number of structural surface skeleton pixel coordinates in the matrix C1 no longer increases; e. Count the columns in which the structural surface skeleton pixels corresponding to the current matrix C1 are distributed in the anchor hole image. When the structural surface skeleton pixels corresponding to the matrix C1 account for more than 80% of the total number of columns in the anchor hole image, the structural surface skeleton pixels corresponding to the C1 matrix are regarded as one structural surface; f. Since there may be multiple structural planes in the anchor hole image, repeat steps a to e to search for other structural planes. Before each repetition of steps a to e, delete the structural plane skeleton pixel segment coordinate matrix corresponding to the matrix C1 and reassign the matrix C1; g. Through steps a to f, determine whether there is still a structural surface skeleton pixel segment coordinate matrix. If not, end; if so, determine whether there are coordinates in the structural surface pixel segment coordinate matrix located at the boundary of the anchor hole. If so, determine whether there is a structural surface skeleton pixel segment coordinate matrix at the boundary of adjacent anchor hole images and the matrix also has coordinates located at the boundary of the anchor hole. Through steps b to e, determine whether the structural surface skeleton pixels corresponding to the structural surface skeleton pixel segment coordinate matrices at the boundaries of two adjacent anchor holes can constitute a structural surface.

6. A method for quickly acquiring parameters of structural surfaces in a borehole according to claim 5, characterized in that: According to the fitted structural surface skeleton expression, the output structural surface parameters include: structural surface inclination α, structural surface inclination angle β, and structural surface position L. The formula is as follows: Where P2 represents the lateral resolution and longitudinal resolution of the anchor hole image; D represents the anchor hole diameter; h represents the actual depth of the anchor hole; A represents the amplitude of the structural surface curve, represents the initial phase of the structural surface curve, and k represents the constant term of the structural surface curve.

7. A computer device, characterized in that: include: Processor, memory, and network interface; The processor is connected to the memory and the network interface, wherein the network interface is used to provide network communication functions, the memory is used to store program codes, and the processor is used to call the program codes to execute the method for quickly acquiring parameters of structural surfaces in a borehole as described in any one of claims 1-6.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which is suitable for being loaded by a processor and executing the method for quickly acquiring parameters of structural surfaces in a borehole as described in any one of claims 1-6.

Images