A method for longitudinal crack identification and automatic parameter extraction based on borehole imaging

Abstract

The present invention discloses a method for longitudinal crack identification and automatic parameter extraction based on borehole imaging, which belongs to the technical field of rock formation feature detection in underground engineering. First, a longitudinal crack identification sample database for borehole imaging is established; then a longitudinal crack pixel recognition model is established using a deep learning algorithm; longitudinal cracks are automatically detected based on a Hough transform algorithm; finally, longitudinal crack parameters are automatically calculated, including the inclination angle, position, continuity coefficient and length of the longitudinal cracks. It effectively solves the problem that the existing borehole imaging method still mainly relies on human eye observation for crack feature analysis, which is greatly affected by subjective influences, has slow speed and low precision. It will promote the transformation of borehole imaging technology from manual qualitative analysis to intelligent quantitative analysis, and provide geological guarantee for underground engineering activities.

Description

Technical Field

[0001] The invention relates to the technical field of geological exploration, and in particular to a method for longitudinal crack identification and automatic parameter extraction based on borehole imaging. Background Art

[0002] The crack characteristics have an important impact on the stability of the rock mass, especially the longitudinal cracks in the tunnel roof, which can easily lead to the instability of the tunnel. The borehole imaging method is commonly used underground to observe and analyze the distribution of cracks in the roof rock layer. The borehole imager can take pictures and videos of the hole wall. The current borehole imager can automatically export digital core columnar diagrams. By observing the borehole columnar diagram, the characteristics of lithology, cracks, etc. can be analyzed, thus providing important geological guarantees for underground engineering activities.

[0003] At present, the borehole imager lacks an intelligent identification system for rock fractures, and people still mainly rely on observation to analyze fracture characteristics. When the sample size of the core column chart to be analyzed is large, this is not only time-consuming and laborious, but also requires analysis by personnel in the field of geology. The observation method can easily overlook some important fracture information. In the prior art, application number: 202110056319.6, a method for automatic extraction of fracture parameters of borehole images based on deep learning, uses the trained fracture image recognition model to detect the fracture image to be detected, and uses the curve detection technology of the PR curve or the ROC curve to process the detection results, and extracts the relevant pixel information of the fracture in the fracture image to be detected. This method has a large amount of calculation and relies entirely on neural networks for recognition. It has high requirements for the clarity of the imaging image. The method for automatic extraction of parameters for multiple fractures in the borehole image is not explained. The recognition system has high computing power requirements and low recognition efficiency. It is mainly aimed at sinusoidal characteristic fracture detection;

[0004] Therefore, a new method for longitudinal crack identification and automatic parameter extraction based on borehole imaging is urgently needed to solve the above problems. Summary of the invention

[0005] In view of the shortcomings of the above-mentioned technologies, a method for longitudinal crack identification and automatic parameter extraction based on borehole imaging is provided, which can obtain the distribution characteristics of longitudinal cracks in rock strata quickly and accurately, and can effectively avoid tunnel instability accidents caused by misjudgment due to subjective influence.

[0006] The present invention provides a method for longitudinal crack identification and automatic parameter extraction based on borehole imaging, and the specific steps are as follows:

[0007] a. Use a 3D borehole imager to collect a large number of video images inside the borehole, and finally form a longitudinal crack recognition sample database for borehole imaging.

[0008] b Using the longitudinal crack recognition sample database as the training set and test set, a longitudinal crack pixel recognition model was established using a deep learning algorithm;

[0009] c. Automatically detect longitudinal cracks based on the longitudinal crack pixel recognition model and Hough transform algorithm, and propose a parameter calculation model for longitudinal cracks;

[0010] d Calculation and output of longitudinal crack parameters.

[0011] Furthermore, the step of establishing a longitudinal crack identification sample database for borehole imaging includes:

[0012] a1 Obtain a large number of anchor hole images;

[0013] a2 divides the anchor hole image pixels into background pixels, interference pixels and longitudinal crack pixels;

[0014] a3 labels the anchor hole image pixels;

[0015] Among them, the anchor hole image is a digital core columnar expansion image; the background pixels are complete rock pixels; the interference pixels are pixels similar to longitudinal cracks: such as sinusoidal crack pixels, broken area pixels, and rock layer interface pixels; the longitudinal crack pixels are crack pixels with an angle greater than 45° with the horizontal plane.

[0016] Furthermore, the deep learning algorithm includes any one of FCN, U-Net, SegNet, and DeepLabv3+.

[0017] Furthermore, the step of automatically detecting longitudinal cracks based on the Hough transform algorithm includes:

[0018] b1 replaces the pixels of the longitudinal cracks with the number "1" among all the identified pixels in each anchor hole image, and replaces the other identified pixels with "0", so as to convert the anchor hole image into a matrix containing only the numbers "0" and "1";

[0019] b2 Based on the matrix obtained above, the Hough transform algorithm package in MATLAB is called, and the threshold is set to 1 to identify the longitudinal cracks formed by the number "1" in the matrix. The longitudinal cracks may be continuous or segmented. If "1" forms a continuous crack, a line segment and its two endpoint coordinates are obtained. If the crack formed by "1" is a segmented crack, the line segment and two endpoint coordinates of each segmented crack are obtained.

[0020] b3 When there are only two endpoint coordinates, if the identified fissure is judged to be a continuous fissure, the coordinates of the two end points of the longitudinal fissure are stored;

[0021] b4 When 2n endpoint coordinates are identified, the identified fissure is judged to be n segmented fissures. In order to avoid misjudgment, the inclination of each segment of the fissure is calculated respectively, and adjacent segmented fissures with an inclination difference of less than 5° are identified as the same longitudinal fissure, otherwise they are regarded as different longitudinal fissures, and the endpoint coordinates of each longitudinal fissure are stored;

[0022] b5 Delete the identified longitudinal cracks by resetting the pixel positions previously marked with the number "1" to "0";

[0023] b6 continues to detect whether there is still a number "1" in the matrix. If so, repeat steps b2 to b5, otherwise end.

[0024] Furthermore, pixels marked with the number "1" are filled with black, and pixels marked with the number "0" are filled with white. The obtained coordinates of the endpoint of the longitudinal crack are used as the base point to search whether there are black pixels around the endpoint coordinate position. If so, the black pixel is used as the base point to continue searching for the black pixels around it, and finally all the pixel points of the longitudinal crack are obtained, thereby obtaining the image information of the longitudinal crack; then all the pixels of the longitudinal crack are replaced with white, and other cracks are continued to be searched through the pixel filling color.

[0025] Furthermore, the parameters for calculating the longitudinal cracks include: the inclination angle, the starting point or end point position, the continuity coefficient and the length of the longitudinal cracks.

[0026] Furthermore, the calculation process of the pixel coordinates of the starting point or end point of the longitudinal crack is:

[0027]

[0028]

[0029] Where L1 and L2 represent the starting and ending positions of the longitudinal cracks, respectively; H represents the depth of the anchor hole; M represents the total number of longitudinal pixels in the anchor hole image; y a and b They represent the vertical coordinates of the starting and ending points of the longitudinal crack in the image.

[0030] Furthermore, the calculation process of the continuity coefficient of the longitudinal crack is:

[0031]

[0032] Where λ represents the longitudinal crack continuity coefficient; k i represents the length of the i-th longitudinal crack; K represents the total length of the longitudinal crack.

[0033] Furthermore, the calculation process of the inclination angle of the longitudinal crack is:

[0034]

[0035] Where, β represents the longitudinal crack inclination; D represents the anchor hole diameter; x a and x b The horizontal coordinates represent the starting and ending points of the longitudinal cracks.

[0036] Further, the length calculation process of the longitudinal crack is:

[0037]

[0038] Where L is the straight-line distance between the endpoints of the longitudinal fracture in the three-dimensional core column diagram.

[0039] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0040] 1. The present invention proposes a method for calculating the position, inclination, continuity coefficient and length of longitudinal cracks in anchor hole images, which lays a foundation for automatic identification of cracks in drilling imaging and automatic output of parameters.

[0041] 2. The present invention proposes to divide the pixels of the anchor hole image into background pixels, interference pixels and longitudinal crack pixels, and uses the semantic segmentation model of the deep learning algorithm to identify the longitudinal crack pixels. Compared with the traditional edge detection algorithm, the longitudinal crack pixel recognition method of the present invention has higher accuracy.

[0042] 3. The present invention directly uses the Hough transform algorithm to perform straight line detection on the identified longitudinal crack pixels. This method has strong anti-interference ability and can effectively identify longitudinal cracks when the longitudinal cracks intersect with the broken zone or are affected by the rock layer interface.

[0043] 4. The present invention proposes to represent the longitudinal crack pixels with the number "1" and other pixels with the number "0", and convert the anchor hole image into a digital matrix. On this basis, a method for identifying multiple longitudinal cracks using the Hough transform method is proposed.

[0044] 5. The present invention has a high degree of automation and can automatically output longitudinal crack parameters, thus promoting the transformation of the original drilling imaging method from manual qualitative analysis to intelligent quantitative analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The longitudinal crack identification and parameter automatic extraction method of the present invention is described below with reference to the accompanying drawings. In the accompanying drawings:

[0046] Figure 1 Schematic diagram of the flow of a method for longitudinal crack identification and automatic parameter extraction based on borehole imaging in this embodiment of the present invention.

[0047] Figure 2This is an image of an anchor hole of a section of 0 to 0.5 m in an embodiment of the present invention.

[0048] Figure 3 This is a schematic diagram of longitudinal crack pixels identified based on a machine learning model according to an embodiment of the present invention.

[0049] Figure 4 This is a schematic diagram of the output results of longitudinal crack identification according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The specific embodiments of the present invention are described below with reference to the accompanying drawings and examples. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0051] like Figure 1 As shown, a method for longitudinal crack identification and automatic parameter extraction based on borehole imaging includes the following steps:

[0052] S100, constructing a sample database for longitudinal crack identification of borehole imaging.

[0053] In order to build a rich sample database, a three-dimensional borehole imager was used to carry out borehole imaging in Shanxi Laomupo Mine, Jingu Mine, Youzhong Mine, Anhui Xieqiao Mine, Guizhou Panzhihua Mine and Wangjiazhai Mine. The rock strata detected included sandstone, mudstone, sandy mudstone and limestone. The obtained anchor hole images were output according to 1350×300 pixels, and a total of 454 anchor hole images were obtained.

[0054] The anchor hole image pixels are divided into background pixels, interference pixels and longitudinal crack pixels, and annotated using the annotation tool provided by MATLAB to finally form a sample database. In order to ensure the accuracy of pixel annotation, this process was completed by three scholars in this field.

[0055] S200. Use deep learning algorithm to establish a longitudinal crack pixel recognition model.

[0056] 70% of the sample database is randomly divided into the training set, and the remaining 30% is divided into the test set. The deep learning algorithm includes any one of FCN, U-Net, SegNet, and DeepLabv3+. DeepLabv3+ is used to train the training set, and the learning rate is set to 5×10 -4 , the number of training steps is 30. Finally, the longitudinal crack recognition model is obtained, and the test set data shows that the recognition rate of the model for longitudinal crack pixels is 0.82.

[0057] S300, automatic detection of longitudinal cracks based on Hough transform algorithm.

[0058] The anchor hole image is input into MATLAB, and the longitudinal crack recognition model obtained through the above training is called by programming. This model can quickly identify various types of pixels in the input anchor hole image. Figure 2 As input anchor hole image.

[0059] Based on the recognition results of various pixels in the anchor hole image by the above model, the longitudinal crack pixels are set to the number "1" (representing black), and the other pixels are set to the number "0" (representing white), such as Figure 3 The following is the corresponding longitudinal crack pixel recognition result. The black pixels in the figure are longitudinal crack pixels, and the white pixels are other pixels.

[0060] Call MATLAB's Hough transform algorithm to perform straight line detection, and the result is as follows Figure 4 The gray straight line in the black pixel on the left, where the right side is the output longitudinal crack parameters, L1 and L2 are the starting and ending positions of the longitudinal crack respectively, L is the crack length, λ is the longitudinal crack continuity coefficient, β is the longitudinal crack inclination; the gray straight line in the black pixel on the left is a schematic diagram of the results obtained by the Hough transform algorithm. The longitudinal crack has two sections, and the difference between the inclination of the first section and the second section is less than 5°, so it is judged that the two sections belong to the same crack, and the coordinates of the endpoints of the longitudinal crack are stored.

[0061] Using the obtained coordinates of the endpoint of the longitudinal crack as the base point, search for the number "1" around the coordinate position of the endpoint. If there is a number "1", continue to search for the number "1" around it with the number "1" as the base point, and finally obtain all the pixel points of the longitudinal crack, and replace all the numbers "1" at the pixel points of these longitudinal cracks with "0". Figure 2 The vertical crack pixels (black pixels) in are reset to white pixels. Figure 2 The middle longitudinal crack is close to the left side of the borehole image, and it is easy to ignore it by human eye observation; Figure 2 There are many horizontal straight lines in the image, which are the stitching lines of the borehole imaging. They are caused by the high speed of borehole imaging, which brings challenges to the identification of longitudinal cracks.

[0062] The search matrix is ​​searched to see if there is a number "1". If not, it is found that there is only one longitudinal crack in the anchor hole image.

[0063] S400, automatic calculation and output of longitudinal crack parameters.

[0064] Based on the coordinates of the endpoints of the longitudinal crack identified above, the position, continuity coefficient, length and inclination of the longitudinal crack are calculated.

[0065] The location of the longitudinal crack is calculated as:

[0066]

[0067]

[0068] Where L1 and L2 represent the starting and ending positions of the longitudinal cracks, respectively; H represents the depth of the anchor hole; M represents the total number of longitudinal pixels in the anchor hole image; y a and b They represent the ordinates of the starting and ending points of the longitudinal cracks respectively.

[0069] The calculation method of the continuity coefficient of the longitudinal crack is:

[0070]

[0071] Where λ represents the longitudinal crack continuity coefficient; k i represents the length of the i-th longitudinal crack; K represents the total length of the longitudinal crack. i and K can be calculated based on the coordinates of the longitudinal crack endpoints obtained after Hough transform detection.

[0072] The calculation method of the longitudinal crack inclination is:

[0073]

[0074] Where, β represents the longitudinal crack inclination; D represents the anchor hole diameter; x a and x b The horizontal coordinates of the starting and ending points of the longitudinal fractures are shown in Figure 1. β refers to the angle between the line connecting the end points of the longitudinal fractures in the 3D digital core column diagram and the projection line of the line on the horizontal plane.

[0075] The calculation method of the longitudinal crack length is:

[0076]

[0077] It should be noted that L is not the difference between L2 and L1. Here L is the straight-line distance between the end points of the longitudinal fracture in the three-dimensional core column diagram, that is, the true fracture length.

[0078] According to the above longitudinal crack parameter calculation formula, the longitudinal crack is located between 180.81 mm and 417.05 mm, with an inclination of -89.94°, a length of 236.24 mm, and a continuity coefficient of 0.51. Figure 4 Shown are the results of longitudinal crack recognition and automatic parameter extraction of the anchor hole image.

[0079] Although the detailed steps of the method of the present invention are described in detail above, those skilled in the art may combine, split and change the order of the above steps without departing from the basic principles of the present invention. The modified technical scheme does not change the basic concept of the present invention and therefore falls within the scope of protection of the present invention.

[0080] So far, the technical solutions of the present invention have been described in conjunction with the accompanying drawings, but it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A method for longitudinal crack identification and automatic parameter extraction based on borehole imaging, characterized in that: The specific steps are as follows: a. Use a 3D borehole imager to collect a large number of video images in the borehole, and finally form a longitudinal crack recognition sample database for borehole imaging; b Using the longitudinal crack recognition sample database as the training set and test set, a longitudinal crack pixel recognition model was established using a deep learning algorithm; c. Automatically detect longitudinal cracks based on the longitudinal crack pixel recognition model and Hough transform algorithm, and propose a parameter calculation model for longitudinal cracks; The specific steps include: b1 replaces the pixels of the longitudinal cracks with the number "1" among all the identified pixels in each anchor hole image, and replaces the other identified pixels with "0", so as to convert the anchor hole image into a matrix containing only the numbers "0" and "1"; b2 Based on the matrix obtained above, the Hough transform algorithm in MATLAB is used, and the threshold is set to 1 to identify the longitudinal cracks formed by the number "1" in the matrix. The longitudinal cracks may be continuous or segmented. If the cracks formed by "1" are continuous, a line segment and its two endpoint coordinates are obtained. If the cracks formed by "1" are segmented cracks, the line segments and two endpoint coordinates of each segmented crack are obtained. b3 When there are only two endpoint coordinates, if the identified fissure is judged to be a continuous fissure, the coordinates of the two end points of the longitudinal fissure are stored; b4 When 2n endpoint coordinates are identified and the identified fissure is judged to be n segmented fissures, in order to avoid misjudgment, the inclination of each segment of the fissure is calculated respectively, and the adjacent segmented fissures with an inclination difference of less than 5° are regarded as the same longitudinal fissure, otherwise they are regarded as different longitudinal fissures, and the endpoint coordinates of each longitudinal fissure are stored; b5 Delete the identified longitudinal cracks by resetting the pixel positions previously marked with the number "1" to "0"; b6 continues to detect whether there is still a number "1" in the matrix. If so, repeat steps b2 to b5, otherwise end; d Calculation and output of longitudinal crack parameters.

2. The method for longitudinal crack identification and automatic parameter extraction based on borehole imaging according to claim 1 is characterized in that: The steps of building a sample database for longitudinal crack identification of borehole imaging include: a1 Obtain a large number of anchor hole images; a2 divides the anchor hole image pixels into background pixels, interference pixels and longitudinal crack pixels; a3 labels the anchor hole image pixels; Among them, the anchor hole image is a digital core columnar expansion image; the background pixels are complete rock pixels; the interference pixels are pixels similar to longitudinal cracks: such as sinusoidal crack pixels, broken area pixels, and rock layer interface pixels; the longitudinal crack pixels are crack pixels with an angle greater than 45° with the horizontal plane.

3. The method for longitudinal crack identification and automatic parameter extraction based on borehole imaging according to claim 1 is characterized in that: The deep learning algorithm includes any one of FCN, U-Net, SegNet, and DeepLabv3+.

4. The method for longitudinal crack identification and automatic parameter extraction based on borehole imaging according to claim 1 is characterized in that: Fill the pixels marked with the number "1" with black, fill the pixels marked with the number "0" with white, and use the obtained coordinates of the endpoint of the longitudinal crack as the base point to search whether there are black pixels around the endpoint coordinate position. If there are black pixels, continue to search for the black pixels around it with the black pixel as the base point, and finally obtain all the pixel points of the longitudinal crack, thereby obtaining the image information of the longitudinal crack; then replace all the pixel points of the longitudinal crack with white, and continue to search for other cracks through pixel filling color.

5. The method for longitudinal crack identification and automatic parameter extraction based on borehole imaging according to claim 1, characterized in that: The parameters for calculating the longitudinal cracks include: the inclination angle, the starting point or end point position, the continuity coefficient and the length of the longitudinal cracks.

6. The method for longitudinal crack identification and automatic parameter extraction based on borehole imaging according to claim 5, characterized in that: The calculation process of the pixel coordinates of the starting point or end point of the longitudinal crack is: Where L1 and L2 represent the starting and ending positions of the longitudinal cracks, respectively; H represents the depth of the anchor hole; M represents the total number of longitudinal pixels in the anchor hole image; y a and b They represent the vertical coordinates of the starting and ending points of the longitudinal crack in the image.

7. The method for longitudinal crack identification and automatic parameter extraction based on borehole imaging according to claim 6, characterized in that: The calculation process of the continuity coefficient of the longitudinal crack is: Where λ represents the longitudinal crack continuity coefficient; k i represents the length of the i-th longitudinal crack; K represents the total length of the longitudinal crack.

8. The method for longitudinal crack identification and automatic parameter extraction based on borehole imaging according to claim 5, characterized in that: The calculation process of the inclination angle of the longitudinal crack is: Where, β represents the longitudinal crack inclination; D represents the anchor hole diameter; x a and x b The horizontal coordinates represent the starting and ending points of the longitudinal cracks.

9. The method for longitudinal crack identification and automatic parameter extraction based on borehole imaging according to claim 5, characterized in that: The length calculation process of the longitudinal crack is: Where L is the straight-line distance between the endpoints of the longitudinal fracture in the three-dimensional core column diagram.

Images