A method for automatic identification of rock fractures based on convolutional neural network

Abstract

The invention discloses a rock crack automatic identification method based on a convolutional neural network, relates to the technical field of rock crack prediction, and comprises the following steps: collecting a large number of borehole wall rock images; using an image annotation tool to annotate joints and cracks in the borehole wall rock images, to obtain joint and crack morphological images, and a joint and crack morphological data set composed of the joint and crack morphological images; based on a deep learning model, segmenting the joint and crack morphological images, and constructing a crack identification model; the rock crack automatic identification method based on a convolutional neural network, by introducing a deep neural network, can quickly and accurately detect and identify the joints and cracks existing in the borehole wall rock images, so that the identification and confirmation of the cracks are not affected by secondary factors, the field data collection speed is fast, the entry threshold of the work is lowered, and the efficiency of completing the rock, joint and crack identification is improved.

Description

Technical Field

[0001] The invention relates to the technical field of rock crack prediction, and in particular to a rock crack automatic identification method based on a convolutional neural network. Background Art

[0002] As a kind of structural surface, joints and fissures have complex development direction, length, opening, scale and degree of interlacing, and their spatial distribution is highly random, anisotropic and hidden. The specific analysis and research of rock joint and fissure characteristics is one of the important indicators for engineering rock mass classification and rock mass quality evaluation.

[0003] At present, in the field of geotechnical engineering investigation, the identification and occurrence measurement of rock joints and fissures during drilling and coring are mainly carried out by manually searching and identifying the joints and fissures of the core, and using traditional measuring tools to measure the center axis angle of the fissure and record it manually. This method relies heavily on technical experience, has slow identification and measurement speed, low occurrence acquisition efficiency, and wastes a lot of human resources. In addition, during the drilling process, due to the influence of the technical level of the drilling operators and the drilling process, the mechanical crushing of the core, the release of the original stress of the formation, and human subjective factors, the identification of rock joints and fissures is difficult, and the development morphological characteristics and the occurrence measurement errors are large.

[0004] The use of computer artificial intelligence based on convolutional neural networks to replace a large amount of repetitive manual work, shorten the time to complete the work, and improve the accuracy of recognition, so as to achieve the goal of reducing costs and increasing efficiency, is of great significance to the field of engineering survey. Literature search found that the industry's research on artificial intelligence technology involved in this technology is in the pure theoretical stage, and its application in practical engineering is still vacant, and it is urgently needed for relevant personnel to conduct research and application. Summary of the invention

[0005] The purpose of the present invention is to provide a method for automatic identification of rock fractures based on convolutional neural network to solve the above-mentioned deficiencies in the prior art.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a method for automatically identifying rock fractures based on a convolutional neural network, comprising the following steps:

[0007] Collect a large number of borehole wall rock images;

[0008] An image annotation tool is used to annotate the joints and fissures in the rock image of the borehole wall to obtain a joint and fissure morphology image and a joint and fissure morphology data set composed of the joint and fissure morphology images; wherein the image annotation tool can be Labelme software.

[0009] Based on the deep learning model, the joint and fissure morphology images are segmented and processed to build a fissure recognition model;

[0010] Automatically identify the joints and fissures in the rock image of the borehole wall based on the fissure identification model, and automatically cut the joint and fissure parts to obtain the corresponding joint and fissure morphology images, and the obtained joint and fissure images are used as basic data for subsequent work;

[0011] Normalizing the rock image of the borehole wall, extracting the fracture skeleton, and obtaining the fracture skeleton curve coordinate point data of the fracture skeleton;

[0012] The fracture line skeleton curve is depicted and processed using the sine curve morphology to reconstruct the rock structure surface;

[0013] The dip and inclination of joints and fissures are calculated by using trigonometric functions and partial functions to obtain the occurrence of joints and fissures.

[0014] Furthermore, when using an image annotation tool to annotate the joints and fissures in the rock image of the borehole wall, a normal distribution model is embedded so that the joint and fissure morphology data set satisfies the normal distribution, which is beneficial to improving the speed of learning and model convergence when training a deep learning model.

[0015] Furthermore, the construction of the crack identification model specifically includes the following steps:

[0016] Select and determine a deep learning model, that is, select a suitable deep learning model so that the accuracy of the crack recognition model trained based on the model is high. The present application does not limit the specific deep learning model. Preferably, the present application adopts the Maks R-CNN algorithm model, that is, the instance segmentation model;

[0017] The joint and fissure morphology data set is divided into a training set and a test set;

[0018] Using the training set to train the deep learning model to obtain various parameters of the deep learning model;

[0019] Using the test set to test the accuracy of the trained deep learning model;

[0020] If the accuracy of the test set meets the set accuracy, the parameters are determined to obtain the crack recognition model. Otherwise, the parameters are adjusted and retested until the accuracy of the test set meets the set accuracy.

[0021] Furthermore, the method also includes preprocessing the joint and fissure morphology image.

[0022] Furthermore, the preprocessing includes binarizing the joint and fissure morphology image to convert the color joint and fissure morphology image into a black and white image, wherein the white part represents the joint and fissure, and the black part represents the rock and the other irrelevant parts.

[0023] Furthermore, the borehole wall rock image is normalized to extract the fracture skeleton and obtain the fracture skeleton curve coordinate point data of the fracture skeleton, which specifically includes the following steps:

[0024] Normalizing the rock image of the borehole wall;

[0025] Pixel values ​​of joints and fissures in the normalized borehole wall rock image are extracted to obtain relative coordinates of joints and fissures in the borehole wall rock image, and the relative coordinates are the fracture skeleton curve coordinate point data.

[0026] Furthermore, the fracture line skeleton curve is characterized and processed using the sine curve morphology to reconstruct the rock structure surface, which specifically includes the following steps:

[0027] By refining the uneven joint and fracture curve in the rock image of the borehole wall into a uniform joint and fracture curve;

[0028] Based on the changing trend of the joint crack curve, a sine curve is used for fitting to obtain the fitting curve and fitting curve function.

[0029] The dip and inclination of the joints and fissures are calculated by using trigonometric functions and partial functions to obtain the occurrence of the joints and fissures, which specifically includes the following steps:

[0030] Based on trigonometric functions and partial functions, the slope of each point of the fitting curve is calculated;

[0031] The inclination angle of the joints and fissures is obtained based on the slope, and the inclination of the joints and fissures is obtained based on the positive and negative of the slope;

[0032] The coordinates of each point of the joint fissure are obtained based on the fitting curve function, and the depth of each point of the joint fissure and the depth of the joint fissure are obtained in combination with the depth of the origin of the fitting curve.

[0033] 1. Compared with the prior art, the present invention provides a method for automatic identification of rock fractures based on convolutional neural networks. By introducing deep neural networks, the joints and fractures existing in the rock images of the borehole wall can be quickly and accurately detected and identified, so that the identification and confirmation of the fractures are not affected by secondary factors, the speed of field data collection is fast, the entry threshold of this work is lowered, and the efficiency of completing the identification of rocks, joints and fractures is improved.

[0034] 2. Compared with the prior art, the present invention provides a method for automatic identification of rock fractures based on a convolutional neural network. By fitting the fracture curve with a sine curve, the inclination, dip and other characteristics of the fracture and the depth of the fracture can be quickly calculated. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0036] Figure 1 A schematic diagram of steps provided for an embodiment of the present invention;

[0037] Figure 2 This is an automatic identification of rock joints and fissures and a result interface diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0039] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0040] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined. In addition, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal connection of two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0041] Example embodiments will be described more fully below with reference to the accompanying drawings, but the example embodiments may be embodied in different forms and should not be construed as limited to the embodiments set forth herein. On the contrary, the purpose of providing these embodiments is to make the present disclosure thorough and complete and to enable those skilled in the art to fully understand the scope of the present disclosure.

[0042] In the absence of conflict, the various embodiments of the present disclosure and the various features therein may be combined with each other.

[0043] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0044] The terms used herein are only used to describe specific embodiments and are not intended to limit the present disclosure. As used herein, the singular forms "a", "an" and "the" are also intended to include the plural forms, unless the context clearly indicates otherwise. It will also be understood that when the terms "comprising" and / or "made of" are used in this specification, the presence of the features, wholes, steps, operations, elements and / or components is specified, but the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof is not excluded.

[0045] The embodiments described herein may be described with reference to plan views and / or cross-sectional views by means of idealized schematic diagrams of the present disclosure. Therefore, the example illustrations may be modified according to manufacturing techniques and / or tolerances. Therefore, the embodiments are not limited to the embodiments shown in the accompanying drawings, but include modifications of the configurations formed based on the manufacturing process. Therefore, the regions illustrated in the accompanying drawings have schematic properties, and the shapes of the regions shown in the figures illustrate the specific shapes of the regions of the elements, but are not intended to be limiting.

[0046] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present disclosure, and will not be interpreted as having an idealized or overly formal meaning unless explicitly defined as such herein.

[0047] See also Figure 1-Figure 2 , a rock fracture automatic identification method based on convolutional neural network, comprising the following steps:

[0048] S1. Collect a large number of borehole wall rock images;

[0049] S2. Use an image annotation tool to annotate the joints and fissures in the rock image of the borehole wall to obtain a joint and fissure morphology image and a joint and fissure morphology data set composed of the joint and fissure morphology image; wherein the image annotation tool may be Labelme software. Wherein, when the joints and fissures in the rock image of the borehole wall are annotated using an image annotation tool, a normal distribution model is embedded so that the joint and fissure morphology data set satisfies the normal distribution, which is beneficial to improve the speed of learning and model convergence when training a deep learning model. Embedding a normal distribution model is common knowledge under the prior art, and is directly applied without modification, so it is not specifically described in this technical solution, and will not cause trouble to the technical solution based on this field.

[0050] S3. Based on the deep learning model, the joint and fissure morphology image is segmented and processed to construct a fissure recognition model, which specifically includes the following steps:

[0051] a1. Select and determine a deep learning model, that is, select a suitable deep learning model so that the accuracy of the crack recognition model trained based on the model is high. The present application does not limit the specific deep learning model. Preferably, the present application adopts the Maks R-CNN algorithm model, that is, the instance segmentation model;

[0052] a2. Divide the joint and fissure morphology dataset into a training set and a test set;

[0053] a3. Use the training set to train the deep learning model. During the training, the rock image of the borehole wall is used as the input of the model, and the joint and fissure morphology image corresponding to the rock image of the borehole wall is used as the output of the model. The various parameters of the deep learning model are calculated, and the obtained parameters are substituted into the deep learning model for subsequent testing;

[0054] a4. Use the test set to test the accuracy of the trained deep learning model. During the test, the borehole wall rock image in the test set is input into the parameterized deep learning model to obtain the predicted joint and fissure morphology image;

[0055] a5. By comparing the predicted joint and fissure morphology image with the joint and fissure morphology image corresponding to the borehole wall rock image in the test set, it is determined whether the model prediction is accurate. If the accuracy of the test set meets the set accuracy, the parameters are determined to obtain the fracture recognition model. Otherwise, the parameters are adjusted and retested until the accuracy of the test meets the set accuracy. The set accuracy can be set according to actual needs.

[0056] S4. After the fracture recognition model is obtained, the newly obtained rock image of the borehole wall is processed using the fracture recognition model. The joints and fractures in the rock image of the borehole wall are automatically identified based on the fracture recognition model, and the joints and fractures are automatically cut to obtain the corresponding joint and fracture morphology images. The obtained joint and fracture images are used as basic data for subsequent work, avoiding manual recognition and segmentation, greatly improving efficiency and reducing the difficulty of work.

[0057] S5, normalizing the rock image of the borehole wall, extracting the fracture skeleton, and obtaining fracture skeleton curve coordinate point data of the fracture skeleton, specifically comprising the following steps:

[0058] b1. Normalizing the rock image of the borehole wall to reduce the pixel value of the rock image of the borehole wall to [0, 1]. After normalization, the features of different dimensions can be numerically comparable to a certain extent, which can greatly improve the accuracy of the classifier in the fracture recognition model;

[0059] b2. Preprocessing the joint and fissure morphology image includes binarizing the joint and fissure morphology image to convert the color joint and fissure morphology image into a black and white image, wherein the white part represents the joint and fissure, and the black part represents the rock and other irrelevant parts; extracting the pixel values ​​of the joint and fissure in the normalized borehole wall rock image to obtain the relative coordinates of the joint and fissure in the borehole wall rock image, and the relative coordinates are the fracture skeletonization curve coordinate point data.

[0060] S6. Using the sine curve morphology to depict and process the fracture line skeleton curve to reconstruct the rock structure surface, specifically including the following steps:

[0061] c1. The uneven joint and fracture curve in the rock image of the borehole wall is refined into a uniform joint and fracture curve. When the joint and fracture curve is refined, the refined joint and fracture curve is located on the midline of the original joint and fracture curve. That is, when the horizontal coordinates of the refined joint and fracture curve are the same as those of the original joint and fracture curve, the vertical coordinate of the refined joint and fracture curve is the average of the maximum vertical coordinate and the minimum vertical coordinate of the original joint and fracture curve.

[0062] c2. Based on the changing trend of the joint crack curve, a sine curve is used for fitting to obtain a fitting curve and a fitting curve function;

[0063] S7. Calculate the inclination and dip angle of joints and fissures using trigonometric functions and partial functions to obtain the occurrence of joints and fissures, including:

[0064] d1. Based on trigonometric functions and partial functions, the slope of each point of the fitting curve is calculated. This is common knowledge in the prior art and is directly applied without modification. Therefore, it is not described in detail in this technical solution, and it will not cause trouble to this technical solution based on the art;

[0065] d2. Based on the slope, the inclination of the joints and fissures is obtained. Based on the positive and negative values ​​of the slope, the inclination of the joints and fissures is obtained. Similarly, if the slope is known, the angle between the tangent line corresponding to the slope and the horizontal line can be calculated through trigonometric functions, thereby obtaining the inclination and inclination of the joints and fissures.

[0066] d3. Based on the fitting curve function, the coordinates of each point of the joints and fissures are obtained, and the depth of each point of the joints and fissures is obtained by combining the depth of the origin of the fitting curve, so as to know the depth of the joints and fissures. The depth of the joints and fissures point = the depth of the origin of the fitting curve + the ordinate of the joints and fissures point.

[0067] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for automatic identification of rock fractures based on convolutional neural network, characterized in that: The following steps are involved: Collect a large number of borehole wall rock images; Annotate the joints and fissures in the rock image of the borehole wall using an image annotation tool to obtain a joint and fissure morphology image and a joint and fissure morphology data set composed of the joint and fissure morphology images; segment the joint and fissure morphology images based on a deep learning model to construct a fissure recognition model; Automatically identify the joints and cracks in the rock image of the borehole wall based on the crack recognition model, and automatically cut the joint and crack parts to obtain the corresponding joint and crack morphology image; Normalizing the rock image of the borehole wall, extracting the fracture skeleton, and obtaining the fracture skeleton curve coordinate point data of the fracture skeleton; The fracture line skeleton curve is depicted and processed using the sine curve morphology to reconstruct the rock structure surface; The dip and inclination of joints and fissures are calculated by trigonometric and partial functions to obtain the occurrence of joints and fissures. The method further comprises preprocessing the joint and fissure morphology image; The preprocessing includes binarizing the joint and fissure morphology image to convert the color joint and fissure morphology image into a black and white image; Normalizing the rock image of the borehole wall, extracting the fracture skeleton, and obtaining the fracture skeleton curve coordinate point data of the fracture skeleton, specifically includes the following steps: Normalizing the borehole wall rock image; Extracting pixel values ​​of joints and fissures in the normalized rock image of the borehole wall, and obtaining relative coordinates of the joints and fissures in the rock image of the borehole wall, wherein the relative coordinates are the coordinate point data of the skeletonized curve of the fissures; The fracture line skeleton curve is depicted and processed using the sine curve morphology to reconstruct the rock structure surface, which specifically includes the following steps: By refining the uneven joint and fracture curve in the rock image of the borehole wall into a uniform joint and fracture curve; Based on the variation trend of the joint crack curve, a sine curve is used for fitting to obtain a fitting curve and a fitting curve function; The dip and inclination of the joints and fissures are calculated by using trigonometric functions and partial functions to obtain the occurrence of the joints and fissures, which specifically includes the following steps: Based on trigonometric functions and partial functions, the slope of each point of the fitting curve is calculated; The inclination angle of the joints and fissures is obtained based on the slope, and the inclination of the joints and fissures is obtained based on the positive and negative of the slope; Based on the fitting curve function, the coordinates of each point of the joint fissure are obtained. Combined with the depth of the origin of the fitting curve, the depth of each point of the joint fissure and the depth of the joint fissure are obtained. The depth of the joint fissure point = the depth of the origin of the fitting curve + the ordinate of the joint fissure point.

2. The method for automatic identification of rock fractures based on convolutional neural network according to claim 1, characterized in that: When the joints and fissures in the rock image of the borehole wall are annotated using an image annotation tool, a normal distribution model is embedded so that the joint and fissure morphology data set satisfies the normal distribution.

3. The rock fracture automatic identification method based on convolutional neural network according to claim 1, characterized in that: The construction of the crack identification model specifically includes the following steps: Select and finalize a deep learning model; The joint and fissure morphology data set is divided into a training set and a test set; Using the training set to train the deep learning model to obtain various parameters of the deep learning model; Using the test set to test the accuracy of the trained deep learning model; If the accuracy of the test set meets the set accuracy, the parameters are determined to obtain the crack recognition model. Otherwise, the parameters are adjusted and retested until the accuracy of the test set meets the set accuracy.

Images