Stratum fracture identification method, device, equipment, medium and program product

Abstract

The embodiment of the invention provides a formation fracture identification method and device, equipment, a medium and a program product. The method comprises the following steps: acquiring an imaging logging image corresponding to a target stratum; obtaining a plurality of corresponding input sub-images based on the imaging logging image; generating output sub-images corresponding to the multiple input sub-images based on the multiple input sub-images corresponding to the imaging logging image and the trained to convergent stratum fracture recognition model so as to complete fracture recognition; the output sub-image is generated after the stratum fracture recognition model performs local feature and global feature extraction on the input sub-image; according to the label image, the target stratum comprises deep and ultra-deep complex tight sandstone oil and gas reservoir stratums, so that the output sub-images of the local features and the global features can be fused in the face of complex cracks and severe conditions, and the cracks can be identified more accurately.

Description

Technical Field

[0001] This application relates to the technical field of oil and gas reservoir development, and particularly to a method, device, equipment, medium and program product for identifying formation fractures. Background Art

[0002] In recent years, the global oil and gas reservoir exploration and development field has entered a period of rapid development. The formations of deep and ultra-deep tight sandstone oil and gas reservoirs have attracted much attention in the energy field, indicating that they have good exploration potential. In the formation, the occurrence of fractures will also lead to the destruction of the integrity of the caprock and cause oil and gas loss. Therefore, studying the identification of fractures in the formation and their development characteristics can provide reliable geological and engineering bases for the exploration and development of oil and gas in tight sandstone oil and gas reservoir formations.

[0003] In the prior art, an oil-based drilling fluid is added into a formation wellbore. The oil-based drilling fluid fills the formation fractures. A resistivity difference is formed between the oil-based drilling fluid and the wellbore skeleton, and the formation fractures are identified through resistivity difference imaging.

[0004] However, the oil-based drilling fluid actually weakens the conductivity of the fluid in the wellbore, resulting in poor conductivity and increased resistivity of the oil-based drilling fluid in the wellbore. Furthermore, the resistivity difference between the oil-based drilling fluid in the formation fractures and the wellbore skeleton becomes smaller, so that the micro-resistivity imaging effect is poor, the ability to identify formation fractures is reduced, and the formation fractures cannot be accurately identified. Summary of the Invention

[0005] Embodiments of this application provide a method, device, equipment, medium and program product for identifying formation fractures, so as to achieve the effect of accurately identifying formation fractures.

[0006] In a first aspect, embodiments of this application provide a method for identifying formation fractures, including:

[0007] Obtain an imaging logging image corresponding to a target formation; the imaging logging image is at least one; the target formation includes deep and ultra-deep complex tight sandstone oil and gas reservoir formations; the imaging logging image includes a micro-resistivity imaging logging map under an oil-based drilling fluid;

[0008] For each imaging logging image, obtain a corresponding plurality of input sub-images based on the imaging logging image;

[0009] Based on multiple input sub-images corresponding to the imaging logging images and the stratum fracture recognition model that has been trained to convergence, generate output sub-images corresponding to the multiple input sub-images to complete fracture recognition; the output sub-images include the fracture conditions corresponding to the input sub-images; the output sub-images are generated after the stratum fracture recognition model extracts local features and global features from the input sub-images; the stratum fracture recognition model is generated by training with a training set; the training set includes at least one training image and its corresponding label image; the label image includes the fractures existing in the training image; the label image is obtained by performing label marking on the training image using a preset marking software and then performing binary processing.

[0010] For each imaging logging image, splice the multiple output sub-images corresponding to the imaging logging image to obtain the stratum fracture image corresponding to the imaging logging image.

[0011] Optionally, obtaining the imaging logging image corresponding to the target stratum includes:

[0012] Obtain the electrical imaging data corresponding to the target stratum;

[0013] Based on the electrical imaging data, obtain the imaging logging image.

[0014] Optionally, obtaining multiple input sub-images based on the imaging logging image includes:

[0015] Fill the blank area to obtain the filled image;

[0016] Cut the filled image to obtain multiple input sub-images.

[0017] Optionally, cutting the filled image to obtain multiple input sub-images includes:

[0018] Keep the cutting width consistent with the image shooting width, and perform longitudinal overlapping cutting on the filled image within the image shooting height according to a preset step size and a preset cutting height to obtain multiple input sub-images; the overlapping height is the difference between the preset cutting height and the preset step size.

[0019] Optionally, based on multiple input sub-images corresponding to the imaging logging image and the stratum fracture recognition model that has been trained to convergence, generate output sub-images corresponding to the multiple input sub-images to complete fracture recognition, including:

[0020] Batch input the multiple input sub-images corresponding to the imaging logging image into the stratum fracture recognition model;

[0021] Use the stratum fracture recognition model to generate low-resolution feature maps corresponding to the multiple input sub-images;

[0022] The formation fracture identification model is used to generate output sub-images corresponding to multiple input sub-images based on each low-resolution feature map to complete fracture identification; the output sub-images are binary images.

[0023] Optionally, the formation fracture identification model is used to generate low-resolution feature maps corresponding to multiple input sub-images, including:

[0024] The first convolutional layer in the formation fracture identification model is used to extract local features in each input sub-image respectively to obtain local feature maps corresponding to the input sub-images; the local features include edge and corner features in the input sub-images; the first convolutional layer is located in the encoder;

[0025] The pooling layer in the formation fracture identification model is used to perform downsampling on each local feature map respectively to capture abstract features and generate corresponding low-resolution feature maps.

[0026] Optionally, the formation fracture identification model is used to generate output sub-images corresponding to multiple input sub-images based on each low-resolution feature map to complete fracture identification, including:

[0027] The upsampling layer in the formation fracture identification model is used to perform upsampling on the low-resolution feature map to identify the upsampled feature map; the resolution of the upsampled feature map is the same as that of the corresponding input sub-image;

[0028] The second convolutional layer in the formation fracture identification model is used to extract global features in each input sub-image hierarchically in multiple layers; the second convolutional layer is located in the decoder; the global features include appearance and shape features of the input sub-images; the second convolutional layer includes at least one;

[0029] Fuse each global feature into the corresponding upsampled feature map to generate a fused feature map corresponding to each input sub-image; the fused feature map includes local features and global features;

[0030] Determine each fused feature map as the corresponding output sub-image to complete fracture identification.

[0031] In a second aspect, an embodiment of the present application provides a formation fracture identification device, including:

[0032] An acquisition module, configured to acquire an imaging logging image corresponding to a target formation; the imaging logging image is at least one; the target formation includes deep and ultra-deep complex tight sandstone oil and gas reservoir formations; the imaging logging image includes a micro-resistivity imaging logging map under oil-based drilling fluid;

[0033] An obtaining module, configured to, for each imaging logging image, obtain corresponding multiple input sub-images based on the imaging logging image;

[0034] An identification module, configured to generate output sub-images corresponding to multiple input sub-images based on multiple input sub-images corresponding to an imaging logging image and a formation fracture identification model that has been trained to convergence, so as to complete fracture identification; the output sub-images include the fracture conditions corresponding to the input sub-images; the output sub-images are generated after the formation fracture identification model extracts local features and global features of the input sub-images; the formation fracture identification model is generated by training with a training set; the training set includes at least one training image and its corresponding label image; the label image includes the fractures existing in the training image; the label image is obtained by performing label marking on the training image using a preset marking software and performing binarization processing.

[0035] A stitching module, configured to stitch multiple output sub-images corresponding to an imaging logging image for each imaging logging image, so as to obtain a formation fracture image corresponding to the imaging logging image.

[0036] In a third aspect, an embodiment of the present application provides an electronic device, including: a memory, a processor;

[0037] The memory stores computer-executable instructions;

[0038] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementation manners of the first aspect.

[0039] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored, and when the computer-executable instructions are executed by a processor, they are used to implement the above first aspect and / or various possible implementation manners of the first aspect.

[0040] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the above first aspect and / or various possible implementation manners of the first aspect.

[0041] A method, device, equipment, medium and program product for identifying formation fractures provided by an embodiment of the present application. In the present application, first, a plurality of corresponding input sub-images are obtained for the acquired imaging logging images, and then, based on the plurality of input sub-images and a formation fracture identification model that has been trained to convergence, the corresponding output sub-images are identified. The output sub-images include corresponding fractures. Since the output sub-images in the present application are generated after the formation fracture identification model extracts local features and global features from the corresponding input sub-images, the output sub-images fuse local features and global features, thereby making the output sub-images more in line with the actual situation and thus more accurate. Therefore, the fractures in the output sub-images are more accurate, reducing the situations of missed identification and misidentification. In addition, the formation fracture identification model in the present application is trained and generated based on at least one training image and its corresponding label image. Since the formation fracture identification model can extract local features and global features, when training the model, the local features and global features of the training images are also considered; the training set in the present application includes at least one training image and its corresponding label image, and the label image therein is obtained by performing label marking on the training image using a preset marking software and binaryzation processing, so the label image is more accurate, thereby making the formation fracture identification model more optimized. Therefore, the formation fracture identification model is an improved model and can obtain the output sub-images more accurately; the target formation in the present application includes deep and ultra-deep complex tight sandstone oil and gas reservoir formations. Therefore, for complex formations, in the present application, imaging logging images can still be used to identify the fractures included therein, and since local features and global features can be identified, the fractures therein can be identified more accurately; in addition, the imaging logging images in the present application can include micro-resistivity imaging logging maps under oil-based drilling fluids. Therefore, for under oil-based drilling fluids, the present application can also accurately identify the fractures in the target formation. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The accompanying drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present application and, together with the specification, are used to explain the principles of the present application.

[0043] Figure 1 It is a schematic diagram of the scenario for formation fracture identification provided by the present application;

[0044] Figure 2 It is a schematic flowchart of a method for identifying formation fractures provided in Embodiment 1;

[0045] Figure 3 It is a schematic flowchart of a method for identifying formation fractures provided in Embodiment 3;

[0046] Figure 4 It is a schematic flowchart of a method for identifying formation fractures provided in Embodiment 6;

[0047] Figure 5 A schematic diagram of overlapping cutting provided for Example 7;

[0048] Figure 6 A complete schematic diagram of formation fracture identification provided for Example 7;

[0049] Figure 7 A schematic diagram of the structure of a formation fracture identification device provided for Example 8;

[0050] Figure 8 A schematic diagram of the structure of the electronic device provided by the present application.

[0051] Through the above-mentioned drawings, specific embodiments of the present application have been shown, and there will be more detailed descriptions hereinafter. These drawings and the written descriptions are not intended to limit the scope of the concept of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. Detailed Description of the Embodiments

[0052] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0053] In the prior art, an oil-based drilling fluid is added into a formation wellbore. The oil-based drilling fluid fills into formation fractures. A resistivity difference is formed between the oil-based drilling fluid and the wellbore skeleton. Imaging is performed through the resistivity difference to realize the identification of formation fractures. However, the oil-based drilling fluid actually weakens the conductivity of the fluid in the wellbore, resulting in poor conductivity of the oil-based drilling fluid in the wellbore. Furthermore, the resistivity difference between the oil-based drilling fluid in the formation fractures and the wellbore skeleton is smaller, so that the micro-resistivity imaging effect is poor. Due to the small resistivity difference, fractures are misidentified, or even the fractures cannot be identified, thus reducing the formation fracture identification ability. To solve the defects of the prior art, the inventor of this solution has conducted creative research and designed a new solution. This solution provides a method for identifying formation fractures. To solve the problem of low accuracy in identifying formation fractures in the prior art, after obtaining an imaging logging image, multiple corresponding input sub-images are obtained based on the above imaging logging image. Further, output sub-images corresponding to the multiple input sub-images are identified based on a formation fracture identification model that has been trained to convergence. In this solution, the output sub-images are generated by the formation fracture identification model through extracting local features and global features of the input sub-images, including fractures corresponding to the input sub-images. Since local features and global features are fused in this solution, the output sub-images are more in line with the actual situation, and the fractures included therein are more accurate, thus enabling accurate identification of fractures. Moreover, since local features are extracted in this solution, the edges, corners, etc. of the fractures can be accurately characterized. In addition, the formation fracture identification model in this application is trained based on at least one training image and its corresponding label image. Since the formation fracture identification model can extract global features and local features, when training the model, global features and local features of the training images are also extracted, thus learning to extract features in multiple aspects. In addition, the training set in this application includes at least one training image and its corresponding label image. The above label image is obtained by labeling the training image with a preset labeling software and performing binarization processing. The label image includes the fractures existing in the training image. Since the label image is obtained by using the preset labeling software, the label image is also more accurate, so that the formation fracture identification model is more optimized. It is an improved fracture identification model. Therefore, by using the above formation fracture identification model, the output sub-images can be accurately identified, making the identification of formation fractures more accurate. In addition, in this solution, the target formation includes deep and ultra-deep complex tight sandstone oil and gas reservoir formations. Therefore, for complex formations, this application can also accurately identify them through the formation fracture identification model.

[0054] In other related technologies, logging software is used for processing to obtain a fracture image. Although the fracture can be identified from the fracture image, the identified fracture is a one-dimensional fracture (including the fracture shape and length), which is not as intuitive as a two-dimensional one. The fractures in the output sub-images in this solution are two-dimensional (including the fracture shape, length, and width), so they are more intuitive and more accurate.

[0055] Figure 1 The figure is a schematic diagram of the scenario for formation fracture identification provided by this application. As Figure 1 shown, in the specific application scenario of this application, it includes an electronic device 101.

[0056] Among them, the electronic device 101 can be a computer or a server, and there is no limit here.

[0057] Among them, the electronic device 101 includes a formation fracture identification model 102 that has been trained to convergence.

[0058] In this scenario, specifically: the electronic device 101 acquires an imaging logging image corresponding to a target formation, and cuts the imaging logging image to obtain a corresponding plurality of input sub-images.

[0059] Furthermore, the plurality of input sub-images are input into the formation fracture identification model 102, and local features and global features are extracted therefrom to identify the output sub-images corresponding to the respective input sub-images.

[0060] Furthermore, the plurality of output sub-images are stitched together to obtain a formation fracture image corresponding to the imaging logging image.

[0061] Combined with the above scenario, it can be seen that in the prior art, the method of using resistivity difference to identify fractures cannot accurately obtain fracture patterns. In this solution, the output sub-images include local features and global features corresponding to the input sub-images. Therefore, starting from multiple aspects of features, the output sub-images are more accurate. Since local features are extracted, in the actual environment, prominent and minute features in the fractures of the target formation are extracted, and thus the true situation of the fractures in the target formation can be accurately characterized. Local features and global features can be accurately extracted from the color and patterns of the fractures. Therefore, this solution can accurately identify fractures. In addition, the formation fracture identification model in this solution is trained based on at least one training image and its corresponding label image. During training, global features and local features in the training image can also be extracted, so as to learn this function. Furthermore, the formation fracture identification model in this solution is an improved fracture identification model. Therefore, the formation fracture identification model in this solution can accurately identify fractures.

[0062] It should be noted that when the formation fracture identification model is used to extract features from the input sub-image in this application, local features and global features are extracted based on the feature information such as the color and texture of the fractures in the input sub-image, so that the output sub-image is more accurate.

[0063] It should be noted that the formation fracture identification method of this application can be applied to the deep and ultra-deep complex tight sandstone oil and gas reservoir formations under oil-based drilling fluids. The fractures in this application can be structural fractures and attitude fractures. Among them, the attitude fractures include nearly horizontal fractures, low-angle oblique joints, medium-angle oblique joints, and high-angle oblique joints. In this application, for different types of fractures, the formation fracture identification model can extract local features and global features among them, and can accurately identify the fractures. Therefore, for complex fractures, in this application, the formation fracture identification model can be used to extract local features and global features in multiple layers to obtain more accurate local features and global features; in this application, abstract features can also be captured to improve the accuracy.

[0064] The technical solutions of this application and how the technical solutions of this application solve the above technical problems will be described in detail below with specific embodiments. These several specific embodiments below can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of this application will be described below with reference to the drawings.

[0065] Embodiment 1

[0066] The execution subject of Embodiment 1 to Embodiment 7 of this application is a formation fracture identification device (hereinafter referred to as the identification device), and the formation fracture identification device is located in an electronic device.

[0067] Figure 2 A schematic flow chart of a formation fracture identification method provided for Embodiment 1. As Figure 2 shown, it specifically includes:

[0068] S201, obtaining an imaging logging image corresponding to the target formation; the imaging logging image is at least one; the target formation includes deep and ultra-deep complex tight sandstone oil and gas reservoir formations; the imaging logging image includes a micro-resistivity imaging logging map under oil-based drilling fluid.

[0069] It should be noted that before obtaining the imaging logging image corresponding to the target formation, the user can input basic data into the identification device, so that the identification device generates and obtains the imaging logging image based on the obtained basic data.

[0070] Among them, the basic data includes core data, test data, and imaging logging data corresponding to the target formation. Among them, the core data includes surface gamma data, sweep and longitudinal cut photos. The imaging logging data includes electrical imaging data.

[0071] Among them, the target formation includes deep and ultra-deep complex tight sandstone oil and gas reservoir formations, and can also be shallow formations (or conventional formations), which are not limited here.

[0072] Among them, the imaging logging image includes a micro-resistivity imaging logging map under oil-based drilling fluid. Among them, the micro-resistivity imaging logging map can be obtained or generated from electrical imaging data.

[0073] It should be noted that the imaging logging image includes the corresponding image shooting width and image shooting height. It should be noted that the image shooting width symbolizes the width range of the target formation; the image shooting height symbolizes the height range of the target formation. Exemplarily, assuming the image shooting width is 0 - 5m, then it corresponds to the range of 0 - 5m in the horizontal direction of the target formation. In the above horizontal direction, when facing the wellbore directly, the point with a height of 0m on the right wall of the wellbore is used as the center point O, and the right wall of the wellbore extends to the right as the horizontal direction (i.e., the X direction). The image shooting height is the vertical downward height of the formation, starting from the point O on the right wall of the wellbore and extending downward (i.e., longitudinal extension, the Z direction).

[0074] Among them, the image shooting width refers to the width of the target formation captured in the actual scene of the target formation.

[0075] Among them, the image shooting height refers to the height of the target formation captured in the actual scene of the target formation.

[0076] It should be noted that the image has a certain image width and image height. The image width is characterized by a certain proportion to the image shooting width. Similarly, the image height is characterized by a certain proportion to the image shooting height.

[0077] It should be noted that the imaging logging image can have any resolution, which is not limited here.

[0078] It should be noted that the imaging logging image can be at least one, and each corresponds to a different image shooting width. It should be noted that the image shooting height corresponding to each imaging logging image can be the same. Exemplarily, the image shooting width of the first imaging logging image is 0 - 5m, and the image shooting height is 0 - 2000m. The image shooting width of the second imaging logging image is 5 - 10m, and the image shooting height is 0 - 2000m. Among them, the starting height in 0 - 2000m of the image shooting height is 0m, and the ending height is 2000m.

[0079] It should be noted that in this application, the imaging logging image can also be the image corresponding to the cross-section at any angle in the 360° around in the Z direction.

[0080] S202. For each imaging logging image, obtain the corresponding multiple input sub-images based on the imaging logging image.

[0081] In one way, the imaging logging image can be evenly cut according to a preset cutting height, so as to obtain multiple input sub-images corresponding to the imaging logging image.

[0082] Exemplarily, assume that the image shooting height of the imaging logging image is 0 - 2000m, and the preset cutting height is 1m. By evenly cutting according to the preset cutting height, 2000 input sub-images can be obtained. Among them, the first input sub-image occupies the shooting height of 0 - 1m. Among them, the initial height of the image shooting height of the imaging logging image is 0m, and the termination height is 2000m.

[0083] Exemplarily, in this way, the input sub-image can be expressed as: , where refers to the th input sub-image, is the initial height of the image shooting height corresponding to the imaging logging image, refers to the shooting height occupied by the input sub-image corresponding to the imaging logging image, refers to the preset cutting height. According to the above example, the shooting height occupied by the first input sub-image is , that is, the shooting height occupied by the first input sub-image corresponding to the imaging logging image is 0 - 1m.

[0084] S203. Generate output sub-images corresponding to multiple input sub-images based on the multiple input sub-images corresponding to the imaging logging image and the stratum fracture recognition model that has been trained to convergence to complete fracture recognition; the output sub-image includes the fracture situation corresponding to the input sub-image; the output sub-image is generated after the stratum fracture recognition model extracts local features and global features from the input sub-image; the stratum fracture recognition model is generated by training with a training set; the training set includes at least one training image and its corresponding label image; the label image includes the fractures existing in the training image; the label image is obtained by performing label marking on the training image using a preset marking software and performing binary processing.

[0085] Among them, the local feature refers to the prominent and tiny features on the surface layer of fractures in the input sub-image, which can include the edge features and corner features of the fractures, and are not limited here.

[0086] Among them, the global feature refers to the features on the deep layer of the fracture, which are the appearance and shape of the fracture in the input sub-image.

[0087] Among them, the fracture situation corresponding to the input sub-image includes the situation of existing fractures and the situation of non-existing fractures. If it is the situation of existing fractures, the output sub-image describes the fractures corresponding to the corresponding input sub-image; if it is the situation of non-existing fractures, the output sub-image is a blank image.

[0088] Among them, the cracks in the output sub-images are two-dimensional cracks, including the shape, length, and width of the cracks.

[0089] In one way, for each imaging logging image, the corresponding multiple input sub-images are respectively input into the formation fracture recognition model to respectively obtain the output sub-images corresponding to the input sub-images.

[0090] In one way, for each imaging logging image, the corresponding multiple input sub-images are batch input into the formation fracture recognition model, so as to batch obtain the output sub-images corresponding to the input sub-images.

[0091] In this step, the formation fracture recognition model is trained and generated based on at least one training image and its corresponding label image. Among them, the label image is a binary image, that is, a monochromatic image, which includes the fractures corresponding to the training image. The fractures in the label image in this application are also two-dimensional fractures.

[0092] Among them, the trained formation fracture recognition model converged can be trained based on a medical image segmentation model (U-net neural network).

[0093] It should be noted that the label image is obtained by performing label marking using a preset marking software. Specifically: the fractures in the training image are marked with labels using the preset marking software to outline the appearance, shape, length, and width of the fractures, and then binaryzation processing is performed to obtain the label image.

[0094] It should be noted that the resolution of the training image in this step is the same as that of the input sub-image.

[0095] It should be noted that the training image in this step is cut after filling the training imaging image. The cutting method in the training stage is the same as that in the application stage. For details, refer to step S202 or Embodiment 3, which will not be elaborated here.

[0096] S204, for each imaging logging image, splice the multiple output sub-images corresponding to the imaging logging image to obtain the formation fracture image corresponding to the imaging logging image.

[0097] In one way, the shooting height occupied by the output sub-image is the same as that of the corresponding input sub-image. For the multiple output sub-images corresponding to each imaging logging image, splice them in the order from top to bottom according to the shooting height corresponding to each output sub-image, so as to obtain the formation fracture image corresponding to the imaging logging image.

[0098] It should be noted that the image capture height of the formation fracture image is the same as that of the imaging logging image. Similarly, the image capture widths of the two are also the same.

[0099] In one way, in this embodiment, the formation fracture images corresponding to the imaging logging images are stitched together. At the same time, the fracture types included therein can be identified based on the formation fracture images. Among them, the fracture types include filled fractures, semi-open fractures, and open fractures. Specifically, corresponding fracture type images can be generated based on the formation fracture images, and the types corresponding to each fracture are marked in the fracture type images.

[0100] This embodiment provides a method for identifying formation fractures. In this embodiment, first, a plurality of corresponding input sub-images are obtained for the acquired imaging logging images. Then, based on the plurality of input sub-images and the formation fracture recognition model that has been trained to convergence, the corresponding output sub-images are identified. The output sub-images include corresponding fractures. Since the output sub-images in this embodiment are generated after the formation fracture recognition model extracts local features and global features from the corresponding input sub-images, the output sub-images fuse local features and global features, thereby making the output sub-images more accurate. Therefore, the fractures in the output sub-images are more accurate, reducing the situations of missed recognition and misrecognition. In addition, the formation fracture recognition model in this embodiment is trained and generated based on at least one training image and its corresponding label image. Since the formation fracture recognition model can extract local features and global features, when training the model, the local features and global features of the training images are also considered, thereby making the formation fracture recognition model more optimized. Therefore, the formation fracture recognition model is an improved model and can obtain more accurate output sub-images. In addition, the target formation in this embodiment can be the deep and ultra-deep tight sandstone oil and gas reservoir formation under oil-based drilling fluid. Therefore, in complex formations and in the face of complex fractures, this embodiment can use the formation fracture recognition model to identify the input sub-images. Even if the fracture conditions represented in the input sub-images are complex during the recognition process, the local features and global features can still be extracted and fused, thereby obtaining a more accurate output sub-image and more accurately representing the fracture conditions in the output sub-image. In this embodiment, the label image is obtained by label-marking the training image with a preset marking software and performing binary processing. Therefore, the fractures included in the training image are outlined in the label image, so the label image is accurate, and thus the formation fracture recognition model trained based on the training set is more accurate.

[0101] Embodiment 2

[0102] This embodiment is a further refinement of any of the above embodiments and is an optional way to obtain the imaging logging image corresponding to the target formation. Specifically, it includes:

[0103] Step 1:

[0104] Obtain the electrical image data corresponding to the target formation.

[0105] Among them, the above electrical image data may include the resistivity detected by eight electrodes around the wellbore for 360°.

[0106] In one way, the imaging logging image can also be obtained by using the noise data in the acoustic imaging data. Among them, the acoustic imaging data is based on the noise data collected by the sound detector. In one way, the imaging logging image is generated and obtained from the noise data of the fluid in the acoustic imaging data.

[0107] Step 2: Obtain the imaging logging image based on the electrical image data.

[0108] In one way, the resistivity of the fractures and the wellbore is obtained from the electrical image data, and thus the micro-resistivity imaging logging map is outlined based on each resistivity. The above micro-resistivity imaging logging map is the imaging logging image, and then the imaging logging image is obtained.

[0109] This embodiment provides a method for identifying formation fractures. In this embodiment, the imaging logging image can be accurately generated and obtained based on the electrical image data or the acoustic imaging data.

[0110] Embodiment 3

[0111] This embodiment is a further refinement of any of the above embodiments. This embodiment is an optional way to obtain multiple input sub-images based on the imaging logging image.

[0112] Figure 3 A schematic flow chart of a method for identifying formation fractures provided for Embodiment 3 is shown in Figure 3 As shown, it specifically includes:

[0113] It should be noted that in the imaging logging image, due to the harsh test environment, there are some areas that are not tested, which are blank areas.

[0114] S301, Fill the blank areas to obtain the filled image.

[0115] It should be noted that filling the blank areas can ensure the integrity of the image.

[0116] It should be noted that the imaging logging image is a three-channel image, and it is filled according to the color near the blank areas to obtain the filled image.

[0117] Specifically, an interpolation mechanism is used to process the blank areas to complete the filling.

[0118] S302, Cut the filled image to obtain multiple input sub-images.

[0119] In one way, the filled image can be evenly cut or overlapped cut.

[0120] Among them, the overlapping cutting method is shown in detail in Embodiment VII and will not be elaborated here.

[0121] This embodiment provides a method for identifying formation fractures. In this embodiment, in order to ensure integrity, the blank areas in the imaging logging image are filled, so that a more complete filled image for describing the target formation can be obtained.

[0122] Embodiment IV

[0123] This embodiment is a further refinement of any of the above embodiments. This embodiment is an optional way to complete fracture identification by generating output sub-images corresponding to multiple input sub-images based on the multiple input sub-images corresponding to the imaging logging image and the formation fracture identification model that has been trained to convergence.

[0124] Step 1, batch input multiple input sub-images corresponding to the imaging logging image into the formation fracture identification model.

[0125] In one way, all multiple input sub-images are batch input into the formation fracture identification model.

[0126] In another way, the multiple input sub-images are divided into batches to obtain multiple batches of input sub-images. Each batch of input sub-images includes at least two input sub-images. Then, the input sub-images included in the corresponding batch are batch input into the formation fracture identification model according to the batch.

[0127] Step 2, use the formation fracture identification model to generate low-resolution feature maps corresponding to multiple input sub-images.

[0128] In one way, the encoder in the formation fracture identification model is used to batch extract local features and downsample multiple input sub-images to generate low-resolution feature maps.

[0129] Among them, the low-resolution feature map refers to a feature map with a resolution lower than that of the input sub-image. Among them, the low-resolution feature map includes local features. Among them, the local features are also extracted based on the color and texture of the fractures in the input sub-image.

[0130] Step 3, use the formation fracture identification model to generate output sub-images corresponding to multiple input sub-images based on each low-resolution feature map to complete fracture identification; the output sub-image is a binary image.

[0131] In one way, the decoder in the formation fracture recognition model is used to upsample each low-resolution feature map in batches and extract global features, so as to generate corresponding output sub-images, and the corresponding fractures are included in the output sub-images, thus completing fracture recognition.

[0132] This embodiment provides a formation fracture recognition method. In this embodiment, multiple input sub-images can be input into the formation fracture recognition model in batches, and then batch processing can be realized, so the processing speed can be improved.

[0133] Embodiment Five

[0134] This embodiment is a further refinement of any of the above embodiments. This embodiment is an optional way to generate low-resolution feature maps corresponding to multiple input sub-images by using the formation fracture recognition model, and specifically includes:

[0135] Step 1: Use the first convolutional layer in the formation fracture recognition model to extract local features in each input sub-image respectively to obtain local feature maps corresponding to each input sub-image; the local features include edge and corner features in the input sub-image; the first convolutional layer is located in the encoder.

[0136] It should be noted that the formation fracture recognition model includes an encoder and a decoder. Among them, the encoder and the decoder are symmetric. The encoder includes a first convolutional layer and a pooling layer; the decoder includes an upsampling layer and a second convolutional layer. Among them, the first convolutional layer can include multiple ones.

[0137] Specifically, for any input sub-image, the first convolutional layer extracts local features from the input sub-image in order from top to bottom to realize the extraction of local characteristic information of the color and texture of the fracture in the input sub-image, and obtains the local feature map. In this embodiment, the extraction of local features in order from top to bottom realizes multi-layer local feature extraction, making the local features more accurate and more in line with the actual situation.

[0138] It should be noted that when generating the local feature map, binarization processing can be performed, and the obtained local feature map is a binary image.

[0139] Among them, the local feature map is an image generated after local feature extraction of the input sub-image.

[0140] It should be noted that local features are prominent and tiny features, which are a kind of shallow features.

[0141] Step 2: Use the pooling layer in the formation fracture recognition model to downsample each local feature map respectively to capture abstract features and generate corresponding low-resolution feature maps.

[0142] In one way, for any input sub-image, the pooling layer downsamples the local feature map according to a preset resolution, preserves key features during downsampling, and captures abstract features, thereby generating a low-resolution feature map. Among them, the preset resolution is smaller than the resolution corresponding to the input sub-image.

[0143] This embodiment provides a method for identifying formation fractures. In this embodiment, the first convolutional layer extracts local features in the input sub-image respectively. The local features include edge and corner features, so they are prominent and tiny features, which are shallow features. Therefore, fractures in the target formation can be identified more accurately. In addition, when downsampling, abstract features are captured to generate corresponding low-resolution feature maps. As a result, the low-resolution feature maps combine features from multiple aspects, so they are more accurate.

[0144] Embodiment Six

[0145] This embodiment is a further refinement of any of the above embodiments. This embodiment is an optional way to use the formation fracture recognition model to generate output sub-images corresponding to multiple input sub-images based on each low-resolution feature map to complete fracture recognition.

[0146] Figure 4 It is a schematic flowchart of a method for identifying formation fractures provided for Embodiment Six. As Figure 4 shown, it specifically includes:

[0147] S401, use the upsampling layer in the formation fracture recognition model to upsample the low-resolution feature map to identify the upsampled feature map; the resolution of the upsampled feature map is the same as the resolution of the corresponding input sub-image.

[0148] It should be noted that the purpose of upsampling is to restore the resolution of the low-resolution feature map to make its resolution the same as the resolution of the corresponding input sub-image.

[0149] Among them, the upsampled feature map is a feature map obtained by restoring the resolution through upsampling.

[0150] S402, use the second convolutional layer in the formation fracture recognition model to extract global features in each input sub-image in a hierarchical order; the second convolutional layer is located in the decoder; the global features include the appearance and shape features of the input sub-image; the second convolutional layer includes at least one.

[0151] For any input sub-image, the second convolutional layer extracts the global features of fractures in the input sub-image to extract the global characteristic information of the color and texture of fractures in the input sub-image and obtain the global features.

[0152] It should be noted that in this embodiment, the second convolutional layer may include at least one. When extracting global features from the input sub-images, the second convolutional layers in each level can sequentially extract global features according to the hierarchical order of the second convolutional layer, so as to obtain more complete and practical global features, that is, multi-layer extraction of global features is realized.

[0153] S403, fuse each global feature into the corresponding upsampled feature map to generate a fused feature map corresponding to each input sub-image; the fused feature map includes local features and global features.

[0154] Among them, the fused feature map combines local features and also combines global features.

[0155] S404, determine each fused feature map as the corresponding output sub-image to complete crack recognition.

[0156] It should be noted that after obtaining the output sub-image, it is output from the formation crack recognition model.

[0157] This embodiment provides a method for identifying formation cracks. In this embodiment, the low-resolution feature map is upsampled to restore the resolution, so that the resolution of the upsampled feature map is consistent with the resolution of the corresponding input sub-image, thereby ensuring that the resolution of the output sub-image is consistent with the resolution of the input sub-image; in this embodiment, the fused feature map fuses local features and global features, so it can better describe the cracks in the target formation, and thus can more accurately identify the cracks.

[0158] Embodiment Seven

[0159] This embodiment is a further refinement of any of the above embodiments. This embodiment is an optional way to cut the filled image to obtain multiple input sub-images, specifically including:

[0160] Keep the cutting width consistent with the image shooting width, and perform longitudinal overlapping cutting on the filled image within the image shooting height according to the preset step size and the preset cutting height to obtain multiple input sub-images; the overlapping height is the difference between the preset cutting height and the preset step size.

[0161] Among them, the preset step size refers to the distance between the initial height of the previous input sub-image and the initial height of the next input sub-image preset.

[0162] Among them, the overlapping height refers to the overlapping height between two adjacent input sub-images.

[0163] Among them, the preset cutting height refers to the cutting height of each input sub-image set in advance. Exemplarily, it can be 1m or other heights, and the preset cutting height is less than the image shooting height.

[0164] Specifically, keep the cutting width consistent with the image capture width, that is, for the filled image, keep the horizontal width unchanged and perform cutting longitudinally.

[0165] Exemplarily, assume that the image capture width of the filled image is 0 - 5m, occupying 5m, the image capture height is 0 - 200m, the preset cutting height is 2m, and the preset step size is 1m. Perform longitudinal overlapping cutting on the above-mentioned filled image. The image capture width of the first input sub-image obtained is 0 - 5m, and the image capture height is 0 - 2m. According to the preset step size (1m), the image capture width of the second input sub-image is 0 - 5m, and the image capture height is 1 - 3m.

[0166] In this embodiment, the input sub-images obtained by overlapping cutting can be expressed as: , where refers to the th input sub-image, is the initial height of the image capture height corresponding to the imaging logging image, refers to the captured height occupied by the input sub-image in the corresponding imaging logging image, refers to the preset cutting height, refers to the preset step size. According to the above example, the captured height occupied by the first input sub-image is , that is, the captured height occupied by the first input sub-image in the corresponding imaging logging image is 0 - 2m; the captured height occupied by the second input sub-image is .

[0167] Figure 5 FIG. is a schematic diagram of overlapping cutting provided in Embodiment 7. As Figure 5 shown, it includes the filled image 501 and multiple input sub-images 502 after cutting.

[0168] In Figure 5 , assume that the image capture width of the filled image is 0 - 5m, occupying 5m in the X direction, the image capture height is 0 - 20m, the preset cutting height is 2m, and the preset step size is 1m. The image capture width of the first input sub-image obtained is 0 - 5m, and the image capture height is 0 - 2m. The image capture width of the second input sub-image obtained is 0 - 5m, and the image capture height is 1 - 3m. Among them, the overlapping height of the first input sub-image and the second input sub-image is the image capture height 1 - 2m, totaling 1m, which is the preset step size. By analogy, the image capture width of the last input sub-image obtained is 0 - 5m, and the image capture height is 18 - 20m.

[0169] Among them, the filled image and the input sub-images are three-channel images.

[0170] It should be noted that in this embodiment, it also includes the training method of the formation fracture recognition model that has been trained to convergence, and the following steps:

[0171] Step 1: Obtain an initial model; the initial model includes initial model parameters.

[0172] Among them, the initial model can be any version of the medical image segmentation model (U-net neural network) that has been trained to convergence.

[0173] Step 2: Input at least one training image and its corresponding label image into the initial model for training to obtain a formation fracture recognition model that has been trained to convergence.

[0174] In one way, inputting at least one training image and its corresponding label image into the initial model for training to obtain a formation fracture recognition model that has been trained to convergence specifically includes the following steps:

[0175] Step 1: Batch at least one training image and its corresponding label image to obtain multiple batches of training image sets.

[0176] Among them, each batch of training image sets includes at least one training image and its corresponding label image. Among them, each batch of training image sets includes a preset batch quantity of training data. It should be noted that the training data includes the training image and its corresponding label image. The next batch of training image sets includes some training data of the previous batch. The quantity of the above-mentioned part of the training data is the preset overlapping quantity, and the above-mentioned part of the training data is the preset overlapping quantity of the training data at the end after sorting in the previous batch of training image sets. Exemplarily, assume that there are a total of 5 training images and their corresponding label images (i.e., 5 training data, where the training data is arranged according to the serial number), the preset overlapping quantity is 2, and the preset batch quantity is 3. Then, batch according to the preset overlapping quantity and the preset batch quantity. The first batch of training image sets includes 1-3 training data, the second batch of training image sets includes 2-4 training data, and the third batch of training image sets includes 3-5 training data. Among them, the second batch of training image sets includes 2 training data overlapping with the first batch of training image sets, that is, the second training data and the third training data.

[0177] It should be noted that the label image in this embodiment is obtained by performing label marking on the corresponding training image using a preset marking software and performing binarization processing. The label image is a binarized image.

[0178] Among them, the label image may include the fractures of the corresponding training image. The fracture is a two-dimensional fracture, including information such as the shape, length, and width of the fracture.

[0179] Among them, the label marking is to mark the cracks in the training images to outline the cracks.

[0180] Step 2: Input the current batch of training image sets into the initial model for training in the order of batches, and obtain the loss values corresponding to each training data in the current batch of image sets.

[0181] Step 3: Calculate the average loss value of the loss values corresponding to each training data in the current batch of image sets.

[0182] Step 4: If the average loss value of the current batch of training image sets is less than the saved optimal average loss value, update the saved optimal average loss value to the current average loss value, take the current average loss value as the optimal average loss value, and retain the optimized model parameters corresponding to the above current average loss value.

[0183] In one way, when the current batch of training image sets trains the initial model, each training data in the current batch of training image sets can optimize the model parameters, so each training data can correspond to its own optimized model parameters (that is, corresponding to at least one optimized model parameter). Furthermore, when retaining the optimized model parameters corresponding to the current average loss value, the optimal loss value can be determined from the loss values corresponding to each training data, and the optimized model parameters of the training data corresponding to the optimal loss value are determined as the optimized model parameters corresponding to the current average loss value to be retained and retained.

[0184] In another way, for the current batch of training image sets, the optimized model parameters corresponding to the last training data in it can be determined as the optimized model parameters corresponding to the current average loss value to be retained and retained.

[0185] Step 5: Continue to input the next batch of training image sets into the model with the optimized model parameters in Step 4 above, and continue to optimize the model parameters on the basis of the above optimized model parameters until the next batch of training image sets is trained, obtain the average loss value of the next batch and the corresponding optimized model parameters, take the next batch of training image sets as the current training image sets, and continue to execute Step 4 and Step 5 until all batches of training image sets are trained to obtain the optimal optimized model parameters. The above optimal optimized model parameters are the formation fracture recognition models trained to convergence.

[0186] It should be noted that in order to improve the generalization and accuracy of the formation fracture recognition model, some training images may not include cracks, and the corresponding label images are blank images. Furthermore, in actual applications, if there is no crack in the input sub-image, the formation fracture recognition model can accurately generate an output sub-image as a blank image, that is, it will not be misrecognized.

[0187] Among them, the loss value can be calculated based on the binary cross-entropy loss function.

[0188] This embodiment provides a method for identifying formation fractures. In this embodiment, overlapping cutting can ensure the integrity of fracture information in the input sub-image.

[0189] This embodiment provides a complete schematic diagram of formation fracture identification. Figure 6 It is a complete schematic diagram of formation fracture identification provided for Embodiment 7.

[0190] Such as Figure 6 In [the figure], it includes an application stage and a training stage. Among them, in the application stage, it includes the imaging logging image 601 of the target formation, the image 602 after filling the blank area in the imaging logging image 601, the input sub-image 603 after overlapping cutting of the filled image 602, the formation fracture identification model 604 that has been trained to convergence in this application, and the output sub-image 605 obtained by inputting the input sub-image 603 into the formation fracture identification model 604.

[0191] In Figure 6 In [the figure], the input sub-image 603 is input into the formation fracture identification model 604, and its corresponding output sub-image 605 is output from the formation fracture identification model 604.

[0192] Among them, the fractures in the output sub-image 605 are two-dimensional and complete curves.

[0193] It should be noted that in Figure 6 In [the figure], taking one input sub-image 603 as an example for illustration, in actual applications, multiple input sub-images 603 can be obtained after overlapping cutting of the filled image 602. The processing methods of the multiple input sub-images 603 are the same as those of the input sub-image 603 in Figure 6 In [the figure], and will not be elaborated here.

[0194] In Figure 6 In [the figure], it also includes another input sub-image 606, which is obtained by filling the blank area of the remaining imaging logging images and overlapping cutting. Among them, more complete fractures are shown in the other input sub-image 606.

[0195] In Figure 6 In [the figure], the other input sub-image 606 is input into the formation fracture identification model 604, and a first sub-image 607 of two-dimensional fractures is obtained. Among them, the first sub-image 607 is a two-dimensional fracture with a complete fracture morphology.

[0196] Among them, a more complete crack is shown in another input sub-image 606, while no complete crack is shown in the input sub-image 603. However, when the input sub-image 603 is input into the formation crack recognition model 604, an output sub-image 605 with a complete crack form can still be output in this application. Therefore, this application can also recognize complex cracks in complex formations, accurately identify the corresponding cracks, and be able to recognize the complete form of complex cracks.

[0197] In Figure 6 , a traditional recognition model 608 is also included. When the input sub-image 603 is input into the traditional recognition model 608, a second sub-image 609 is output.

[0198] Among them, the crack form in the second sub-image 609 is incomplete, showing intermittent cracks. There is a large gap in the crack form between the second sub-image 609 and the output sub-image 605, indicating that the output sub-image 605 can recognize the complete form of the crack, thereby improving the accuracy.

[0199] In Figure 6 , a logging software 610 is also included.

[0200] Among them, when the input sub-image 603 is input into the logging software 610, a third sub-image 611 is output.

[0201] Among them, the crack in the third sub-image is a one-dimensional crack, and the one-dimensional crack includes a crack form and a length.

[0202] In Figure 6 , it can be seen that the crack in the output sub-image 605 is a two-dimensional crack, which includes a crack form, a length, and a width; while the crack in the third sub-image 611 is a one-dimensional crack, which includes a crack form and a length. Since the output sub-image 605 shows two dimensions, the output sub-image 605 is more accurate than the third sub-image 611.

[0203] In Figure 6 , a formation crack image 612 obtained by splicing the output sub-image 605 and other output sub-images 605 (not shown in the figure) corresponding to the imaging logging image 601 is also included.

[0204] Furthermore, the crack width, crack density, and crack length included in the formation crack image 612 are analyzed to obtain a crack analysis diagram 613.

[0205] In Figure 6 the training stage of, a medical image segmentation model 614, at least one training image 615, and its corresponding label image 616 are also included.

[0206] In Figure 6Among them, the training image 615 is obtained by filling the blank area of the original training image 617 and then cutting it.

[0207] Among them, the original training image 617 is the imaging logging image of the training formation.

[0208] Among them, the training image 615 is labeled with a preset labeling software to obtain the training labeled image 618.

[0209] Among them, the training labeled image 618 is binarized to obtain the label image 616 corresponding to the training image 615.

[0210] Furthermore, training data is obtained according to the above method to obtain at least one training image 615 and its corresponding label image 616, and at least one training image 615 and its corresponding label image 616 are used to train the medical image segmentation model 614 to obtain the formation fracture recognition model 604 that has been trained to convergence. Then, the formation fracture recognition model 604 that has been trained to convergence is used to identify the fractures in the target formation. In Figure 6 one example is used to illustrate one piece of training data. In the actual training stage, at least one piece of training data is required.

[0211] Embodiment 8

[0212] The following is the device embodiment of the present application. Figure 7 It is a schematic structural diagram of a formation fracture recognition device provided for Embodiment 8. As Figure 7 shown, the formation fracture recognition device 700 includes the following modules:

[0213] The acquisition module 701 is used to acquire the imaging logging image corresponding to the target formation; the imaging logging image is at least one; the target formation includes deep and ultra-deep complex tight sandstone oil and gas reservoir formations; the imaging logging image includes the micro-resistivity imaging logging map under oil-based drilling fluid;

[0214] The obtaining module 702 is used to obtain a plurality of corresponding input sub-images based on each imaging logging image for each imaging logging image;

[0215] The recognition module 703 is used to generate output sub-images corresponding to the plurality of input sub-images based on the plurality of input sub-images corresponding to the imaging logging image and the formation fracture recognition model that has been trained to convergence to complete fracture recognition; the output sub-image includes the fracture situation corresponding to the input sub-image; the output sub-image is generated after the formation fracture recognition model extracts local features and global features of the input sub-image; the formation fracture recognition model is generated by training a training set; the training set includes at least one training image and its corresponding label image; the label image includes the fractures existing in the training image; the label image is obtained by labeling the training image with a preset labeling software and then binarizing it.

[0216] A splicing module 704, configured to splice multiple output sub-images corresponding to an imaging logging image for each imaging logging image, so as to obtain a formation fracture image corresponding to the imaging logging image.

[0217] An acquisition module 701, when acquiring an imaging logging image corresponding to a target formation, is specifically configured to:

[0218] Acquire electrical imaging data corresponding to the target formation;

[0219] Acquire an imaging logging image based on the electrical imaging data.

[0220] An obtaining module 702, when obtaining multiple input sub-images based on the imaging logging image, is specifically configured to:

[0221] Fill the blank area to obtain a filled image;

[0222] Cut the filled image to obtain multiple input sub-images.

[0223] An obtaining module 702, when cutting the filled image to obtain multiple input sub-images, is specifically configured to:

[0224] Keep the cutting width consistent with the image shooting width, and perform longitudinal overlapping cutting on the filled image within the image shooting height according to a preset step length and a preset cutting height, so as to obtain multiple input sub-images; the overlapping height is the difference between the preset cutting height and the preset step length.

[0225] An identification module 703, when generating output sub-images corresponding to multiple input sub-images based on an imaging logging image and a formation fracture identification model that has been trained to convergence to complete fracture identification, is specifically configured to:

[0226] Batch input multiple input sub-images corresponding to the imaging logging image into the formation fracture identification model;

[0227] Use the formation fracture identification model to generate low-resolution feature maps corresponding to multiple input sub-images;

[0228] Use the formation fracture identification model to generate output sub-images corresponding to multiple input sub-images based on each low-resolution feature map to complete fracture identification; the output sub-image is a binary image.

[0229] An identification module 703, when using the formation fracture identification model to generate low-resolution feature maps corresponding to multiple input sub-images, is specifically configured to:

[0230] The first convolutional layer in the formation fracture recognition model is used to extract local features from each input sub-image respectively, so as to obtain local feature maps corresponding to the input sub-images; the local features include edge and corner features in the input sub-image; the first convolutional layer is located in the encoder;

[0231] The pooling layer in the formation fracture recognition model is used to downsample each local feature map respectively, so as to capture abstract features and generate corresponding low-resolution feature maps.

[0232] The recognition module 703, when using the formation fracture recognition model to generate output sub-images corresponding to multiple input sub-images based on each low-resolution feature map to complete fracture recognition, is specifically used for:

[0233] The upsampling layer in the formation fracture recognition model is used to upsample the low-resolution feature map to identify the upsampled feature map; the resolution of the upsampled feature map is the same as that of the corresponding input sub-image;

[0234] The second convolutional layer in the formation fracture recognition model is used to extract global features from each input sub-image in multiple hierarchical orders; the second convolutional layer is located in the decoder; the global features include appearance and shape features of the input sub-image; the second convolutional layer includes at least one;

[0235] Fuse each global feature into the corresponding upsampled feature map to generate a fused feature map corresponding to each input sub-image; the fused feature map includes local features and global features;

[0236] Determine each fused feature map as the corresponding output sub-image to complete fracture recognition.

[0237] The formation fracture recognition device provided in this embodiment can execute the method provided in the above method embodiment, and its implementation principle and technical effect are similar, which will not be elaborated here in this embodiment.

[0238] Figure 8 It is a schematic structural diagram of the electronic device provided in this application. As Figure 8 shown, the electronic device 800 provided in this embodiment includes: at least one processor 801 and a memory 802. Optionally, the electronic device 800 further includes a communication component 803. Among them, the processor 801, the memory 802, and the communication component 803 are connected through a bus 804.

[0239] In the specific implementation process, at least one processor 801 executes the computer execution instructions stored in the memory 802, so that at least one processor 801 executes the above method.

[0240] The specific implementation process of the processor 801 can refer to the above method embodiment, and its implementation principle and technical effect are similar, which will not be elaborated here in this embodiment.

[0241] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the invention can be directly implemented by a hardware processor, or can be implemented by a combination of hardware and software modules in the processor.

[0242] The memory may include a random access memory (RAM), and may also include a non-volatile memory (NVM), such as at least one disk memory.

[0243] The bus may be an industry standard architecture (ISA) bus, a peripheral component interconnect (PCI) bus, an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, the buses in the drawings of this application are not limited to only one bus or one type of bus.

[0244] This application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0245] This application also provides a computer-readable storage medium, in which computer-executable instructions are stored, and when the processor executes the computer-executable instructions, the above method is implemented.

[0246] The above-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk. The readable storage medium can be any available medium accessible by a general-purpose or special-purpose computer.

[0247] An exemplary readable storage medium is coupled to a processor such that the processor can read information from and write information to the readable storage medium. Of course, the readable storage medium can also be part of the processor. The processor and the readable storage medium can be located in an Application Specific Integrated Circuits (ASIC). Of course, the processor and the readable storage medium can also exist as discrete components in a device.

[0248] The division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings or direct couplings or communication connections shown or discussed between each other can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0249] The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0250] In addition, in each embodiment of the present invention, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0251] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in each embodiment of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0252] Those of ordinary skill in the art will understand that all or part of the steps to implement the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps including the above method embodiments; and the foregoing storage medium includes: various media such as ROM, RAM, magnetic disks, or optical discs that can store program codes.

[0253] Finally, it should be noted that those skilled in the art will readily conceive of other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention, which follow the general principles of the present invention and include known common knowledge or conventional technical means in the technical field not disclosed by the present invention. It is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is only limited by the appended claims.

Claims

1. A method for identifying formation fractures, characterized in that, Including: Obtaining an imaging logging image corresponding to a target formation; the imaging logging image is at least one; the target formation includes deep and ultra-deep complex tight sandstone oil and gas reservoir formations; the imaging logging image includes a micro-resistivity imaging logging map under oil-based drilling fluid; For each imaging logging image, obtaining a corresponding plurality of input sub-images based on the imaging logging image; Based on the plurality of input sub-images corresponding to the imaging logging image and a trained-to-converge formation fracture identification model, generating output sub-images corresponding to the plurality of input sub-images to complete fracture identification; the output sub-image includes the fracture condition corresponding to the input sub-image; the output sub-image is generated after the formation fracture identification model extracts local features and global features of the input sub-image; the formation fracture identification model is generated by training with a training set; the training set includes at least one training image and its corresponding label image; the label image includes fractures existing in the training image; the label image is obtained by performing label marking on the training image using a preset marking software and performing binary processing; For each imaging logging image, splicing the plurality of output sub-images corresponding to the imaging logging image to obtain a formation fracture image corresponding to the imaging logging image.

2. The method according to claim 1, characterized in that, The obtaining of the imaging logging image corresponding to the target formation includes: Obtaining the electrical imaging data corresponding to the target formation; Based on the electrical imaging data, obtaining the imaging logging image.

3. The method according to claim 1, wherein The obtaining of a plurality of input sub-images based on the imaging logging image includes: Filling the blank area to obtain a filled image; Cutting the filled image to obtain a plurality of input sub-images; The cutting of the filled image to obtain a plurality of input sub-images includes: Keeping the cutting width consistent with the image shooting width, and longitudinally overlapping and cutting the filled image within the image shooting height according to a preset step length and a preset cutting height to obtain a plurality of input sub-images; the overlapping height is the difference between the preset cutting height and the preset step length.

4. The method according to claim 1, wherein The generating of output sub-images corresponding to the plurality of input sub-images based on the plurality of input sub-images corresponding to the imaging logging image and a trained-to-converge formation fracture identification model to complete fracture identification includes: Batch inputting the plurality of input sub-images corresponding to the imaging logging image into the formation fracture identification model; Using the formation fracture identification model to generate low-resolution feature maps corresponding to the plurality of input sub-images; Using the formation fracture identification model to generate output sub-images corresponding to the plurality of input sub-images based on each of the low-resolution feature maps to complete fracture identification; the output sub-image is a binary image.

5. The method according to claim 4, wherein The using of the formation fracture identification model to generate low-resolution feature maps corresponding to the plurality of input sub-images includes: Using the first convolutional layer in the formation fracture identification model to extract local features in each of the input sub-images to obtain local feature maps corresponding to the input sub-images; the local features include edge and corner features in the input sub-images; the first convolutional layer is located in the encoder; The pooling layers in the formation fracture identification model are used to downsample each of the local feature maps to capture abstract features and generate corresponding low-resolution feature maps.

6. The method according to claim 4, wherein The formation fracture identification model generates output sub-images corresponding to multiple input sub-images based on the low-resolution feature maps to complete fracture identification, including: The upsampling layer in the formation fracture identification model is used to upsample the low-resolution feature map to identify the upsampled feature map; the resolution of the upsampled feature map is the same as that of the corresponding input sub-image; The second convolutional layer in the formation fracture identification model is used to hierarchically extract the global features in each of the input sub-images; the second convolutional layer is located in the decoder; the global features include the appearance and shape features of the input sub-image; the second convolutional layer includes at least one; Each of the global features is fused into the corresponding upsampled feature map to generate a fused feature map corresponding to each of the input sub-images; the fused feature map includes local features and global features; Each of the fused feature maps is determined as the corresponding output sub-image to complete fracture identification.

7. A formation fracture identification device, characterized in that, Including: An acquisition module for acquiring imaging logging images corresponding to a target formation; the imaging logging images are at least one; the target formation includes deep and ultra-deep complex tight sandstone oil and gas reservoir formations; the imaging logging images include micro-resistivity imaging logging maps under oil-based drilling fluids; An obtaining module for, for each imaging logging image, obtaining corresponding multiple input sub-images based on the imaging logging image; An identification module for generating output sub-images corresponding to multiple input sub-images based on the multiple input sub-images corresponding to the imaging logging image and a formation fracture identification model that has been trained to convergence to complete fracture identification; the output sub-image includes the fracture condition corresponding to the input sub-image; the output sub-image is generated after the formation fracture identification model extracts local features and global features from the input sub-image; the formation fracture identification model is generated by training with a training set; the training set includes at least one training image and its corresponding label image; the label image includes the fractures existing in the training image; the label image is obtained by performing label marking on the training image using a preset marking software and performing binary processing; A splicing module for, for each imaging logging image, splicing the multiple output sub-images corresponding to the imaging logging image to obtain a formation fracture image corresponding to the imaging logging image.

8. An electronic device, characterized in that, Including: A memory, a processor; The memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory, so that the processor executes the method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that Computer execution instructions are stored in the computer-readable storage medium, and when the computer execution instructions are executed by a processor, they are used to implement the method according to any one of claims 1-6.

10. A computer program product, characterized in that, Including a computer program, which when executed by a processor implements the method according to any one of claims 1-6.

Images