A shale micro-matter intelligent detection method and system based on deep learning

Abstract

The present invention relates to the technical field of shale gas exploration, and in particular to a method and system for intelligent detection of shale microscopic matter based on deep learning. The method comprises: preparing a high-quality data set of shale scanning electron microscope, and dividing the data into quartz, pyrite, organic matter, organic matter pores and inorganic matter pores; constructing a Shale-Yolo target detection model based on a Yolov8 target detection model, relying on a transfer learning technology, loading pre-trained weights for training the prepared high-quality data set, and saving the trained network model and weight file; opening shale scanning electron microscope microscopic matter target detection software, and importing the trained network model and weight file; selecting to import a picture to realize static image detection, and selecting screen detection to realize dynamic screen real-time recognition, wherein the processing speed of a single static image is only 203.1 milliseconds, and the dynamic real-time recognition speed of the screen is as high as 91fps per second, which can fully meet the requirements of engineering applications.

Description

Technical Field

[0001] The present invention relates to the field of petroleum exploration and development and the field of rock microscopic material detection technology, and in particular to a shale microscopic material intelligent detection method and system based on deep learning. Background Art

[0002] With the rapid development of deep learning technology, target detection models have achieved remarkable results in the field of image recognition. In the field of oil exploration, shale scanning electron microscope images are an important type of image data used to observe and analyze the microstructure and composition of shale rocks. However, traditional manual analysis methods have problems such as low efficiency and strong subjectivity, and cannot meet the needs of quickly and accurately identifying shale microscopic substances.

[0003] In the prior art, Chinese patent application number: CN202210910122.9, a shale pore type detection and classification method and system based on deep learning, discloses a shale pore type detection and classification method and system based on deep learning, belonging to the field of petroleum exploration and development rock physical parameter characterization technology, the method includes: preparing a shale SEM image pore data set, dividing the data into two categories: organic pores and inorganic pores; then loading the pre-trained weights of the YOLOV3 deep convolutional neural network model, and then using the prepared shale SEM image pore data set for training, and then saving the network model and the trained weight file; finally, using the trained model, i.e., weights, to perform shale pore detection. The shale pore type detection and classification method and system have high recognition accuracy, and compared with manual recognition and traditional image processing methods for pore recognition, they are highly efficient and fast; the detection frame rate is fast, and a detection speed of 15fps can be achieved on an ordinary computing platform, which can meet laboratory detection needs.

[0004] Chinese patent application number: CN202210827143.4, a method for extracting cracks from shale electron microscope scanning images based on a deep learning segmentation network, discloses a method for extracting cracks from shale electron microscope scanning images based on a deep learning segmentation network, comprising the following steps: S1: extracting the main content part; S2: performing edge extraction on the main content part; S3: performing feature fusion on the main content and edge response; S4: performing loss function supervision on the function after feature fusion in the model training stage; the present invention realizes filtering the interference of complex background from electron microscope scanning images, accurately identifying, segmenting and extracting cracks, and increasing the channel attention mechanism to improve the extraction effect of the crack body.

[0005] Chinese Patent Application No.: CN202210508568.9, A 3D reconstruction method for digital shale core based on deep learning, discloses a 3D reconstruction method for digital shale core based on deep learning, obtains a first image data set; binarizes the first image data set to obtain a second image data set; constructs a network model based on the optical flow method and the generative adversarial network method, and uses the second image data set as a training data set, and simultaneously inputs two-dimensional Gaussian noise for sample simulation, trains the network model, and obtains an optimal model; inputs the two-dimensional plane image of the core acquired in real time into the optimal model to obtain a serialized two-dimensional core slice set; converts the two-dimensional core slice set into a three-dimensional model file in STL format, and processes the three-dimensional model file to obtain a shale digital core. The beneficial effects of the present invention are that the amount of samples collected is small, the reconstruction time is short, and the serialized two-dimensional core slice set reconstructed has good continuity and the core model has high accuracy.

[0006] To address this problem, this patent proposes a method and system for intelligent detection of shale microscopic materials based on deep learning. By building a deep learning target detection model and combining it with a large-scale shale scanning electron microscope image data set for training, automatic recognition and detection of shale microscopic materials can be achieved. This method can not only improve the accuracy and efficiency of recognition, but also reduce the cost and time consumption of manual analysis, providing a new solution for the microstructural analysis of shale rocks. Summary of the invention

[0007] In view of this, in order to overcome the shortcomings of the above-mentioned similar technologies, the present invention proposes a shale micro-matter intelligent detection method and system based on deep learning, which can not only solve the subjectivity of manually processed images, but also greatly improve the efficiency of reservoir evaluation; compared with traditional image processing algorithms, the present invention has a stronger generalization and robustness in the recognition of different shale micro-matter. The shale micro-matter intelligent detection method and system based on deep learning improves the accuracy and speed of shale micro-matter identification, and reduces human errors and workload.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] The present invention provides a method for intelligent detection of shale microscopic matter based on deep learning, comprising the following steps:

[0010] Obtain shale SEM-Maps images;

[0011] Based on SEM-Maps images, a high-quality SEM data set of shale was produced, and the data was divided into quartz, pyrite, organic matter, organic pores, and inorganic pores;

[0012] Based on the target type and Yolov8 target detection model, a Shale-Yolo target detection model is constructed to train the prepared high-quality data set and save the trained network model and weight file;

[0013] Based on the trained network model and weight file, open the shale scanning electron microscope microscopic material target detection software, import the trained network model and weight file, choose to import pictures to achieve static target detection, and select screen detection to achieve real-time recognition of dynamic screens.

[0014] The present invention further provides a method for constructing a Shale-Yolo target detection model of the shale micro-matter intelligent detection method, including:

[0015] S2.1: Prepare high-quality shale SEM data sets and classify the data into quartz, pyrite, organic matter, organic pores, and inorganic pores;

[0016] S2.2: Based on the Yolov8 target detection model, a Shale-Yolo target detection model is constructed;

[0017] S2.3: Add a small target detection layer in the Neck part, and perform detection after splicing the shallow feature map with the deep feature map, so as to improve the model's ability to recognize microscopic substances;

[0018] S2.4: Adding the Shuffle attention mechanism module to the 10th and 20th layers of the model head can effectively capture the relationship between global and local features and improve the representation ability of the model;

[0019] S2.5: Add the SE attention mechanism module, combine the SE attention mechanism model with the C2f_SE module, and construct the C2f_SE module;

[0020] S2.6: Replace the Cf2 modules in the first, second, third and fourth layers of the head part with Cf2_SE modules;

[0021] S2.7: Add the BiFPN bidirectional feature pyramid network and merge it with the original Concat module to establish a new Concat_BiFPN module;

[0022] S2.8: Add the Concat_BiFPN module to the third and sixth layers of the head part of the model to effectively integrate feature information at different levels and improve the model's ability to detect and recognize targets.

[0023] The present invention further comprises that in S2.3, the small target detection layer performs feature extraction through convolution operation, thereby enhancing detail capture capability and multi-scale feature extraction, while also improving model robustness.

[0024] The present invention further states that in S2.4, the Shuffle attention mechanism module is based on the multi-head self-attention mechanism in the Transformer model. It generates an output sequence of the same length as the input sequence by weighted summing each position in the input sequence. In this process, the model learns the importance of each position to the output, thereby being able to capture the correlation between different parts of the input sequence.

[0025] The present invention further states that in S2.5, the SE attention mechanism module introduces an attention mechanism on each channel of the convolutional neural network, and dynamically adjusts the weight of the channel features by learning the importance weight of each channel. Specifically, the SE attention mechanism includes two steps: squeeze and excitation. In the squeeze stage, the feature map of each channel is compressed by a global average pooling operation to obtain the global description information of the channel; in the excitation stage, the activation function of each channel is learned through two fully connected layers, and then the activation function is applied to the original features to obtain a weighted feature representation.

[0026] The present invention further comprises that in S2.6, in the C2f_SE module, after a series of convolution operations and feature fusion, the feature map is connected to the SE attention mechanism. In the SE attention mechanism, the feature map of each channel is compressed and the activation function is learned to dynamically adjust the weight of the channel feature, thereby enhancing the expression ability of important features.

[0027] The present invention further states that in S2.7, the BiFPN bidirectional feature pyramid network combines the idea of ​​a bidirectional information transfer mechanism and a feature pyramid network.

[0028] The feature pyramid network includes multiple feature pyramid levels of different scales, which are used to extract feature information of different scales; the bidirectional information transmission mechanism means that in the feature pyramid network, feature information can be transmitted from bottom to top and from top to bottom to achieve cross-level feature fusion and information transmission.

[0029] The present invention further comprises that in S2.8, the Concat_BiFPN module combines these different feature fusion methods, and splices feature information of different scales and levels through the Concat operation, thereby achieving a more comprehensive and rich feature representation.

[0030] The present invention provides a method for intelligent detection of shale micro-materials based on deep learning. The method can identify different shale micro-materials from shale SEM-Maps images, and has a strong ability to characterize features of different scales. It has high detection accuracy while ensuring that the detection speed meets the real-time requirements. The pre-trained shale micro-material recognition model, combined with functions such as image import and real-time detection screen, can effectively improve the recognition accuracy and speed of shale micro-materials, reduce human errors and workload; the algorithm has high recognition accuracy, and is more efficient and faster than manual recognition and traditional image processing methods for identifying pores; the detection frame rate is fast, with a single static image processing speed of only 203.1 milliseconds, and a screen dynamic real-time recognition speed of up to 91fps per second, which can fully meet the needs of engineering applications.

[0031] The present invention further provides that the method for using the shale scanning electron microscope microscopic material target detection software comprises the following steps:

[0032] S9.1: Open the Shale SEM Microscopic Material Target Detection Software, click the "Model" button, and import the weight file of the trained Shale-Yolo target detection model;

[0033] S9.2: Click the "Detect Screen" button to select the imported screen image and start real-time detection of the screen image to achieve dynamic target detection;

[0034] S9.3: Click the “Impert Image” button to select the imported image to realize static target detection.

[0035] In shale reservoir evaluation, the imaging method has important significance and role. The imaging method can provide intuitive visual information, and can deeply understand the key parameters of shale reservoirs such as pore structure, mineral composition, and fracture distribution, providing an important reference for shale reservoir evaluation. It can also quantitatively analyze the microstructure of shale reservoirs, and help researchers identify and characterize microscopic features, thereby achieving the purpose of assisting geological modeling and oil and gas exploration.

[0036] The present invention also provides a shale micro-matter intelligent detection system based on deep learning, which is realized by the above-mentioned shale micro-matter intelligent detection method based on deep learning.

[0037] The present invention designs a method and system for intelligent detection of shale micro-matter based on deep learning, which has important significance and innovative value. First of all, in the process of constructing the Shale-Yolo target detection model, the system introduces a series of innovative designs, adding small target detection layers, Shuffle attention mechanism modules, SE attention mechanism modules, BiFPN bidirectional feature pyramid networks, etc. to different parts of the model. These designs effectively improve the model's recognition and characterization capabilities of micro-matter, making the detection results more accurate and comprehensive. In addition, the system also proposes a new C2f_SE module, which dynamically adjusts the weights of channel features by introducing the SE attention mechanism, thereby enhancing the expression ability of important features and further improving the performance and effect of the model. By utilizing deep learning technology, especially constructing the Shale-Yolo target detection model, intelligent detection of shale micro-matter is realized, and the traditional manual detection method is transformed into an automated and efficient detection process, which greatly improves the accuracy and reliability of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only used for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. In the accompanying drawings:

[0039] Figure 1 A process for producing a data set for the present invention;

[0040] Figure 2 This is the structure diagram of the small target detection layer;

[0041] Figure 3 This is the module structure diagram of the Shuffle attention mechanism;

[0042] Figure 4 This is the module structure diagram of SE attention mechanism;

[0043] Figure 5 This is the BiFPN bidirectional feature pyramid network structure diagram;

[0044] Figure 6 It is a structural schematic diagram of the present invention;

[0045] Figure 7 is a basic flow chart of the present invention;

[0046] Figure 8 It is the software interface diagram of the present invention;

[0047] Fig. 9 The software detection effect diagram of the present invention (a is static image recognition, b and c are dynamic screen real-time detection). DETAILED DESCRIPTION

[0048] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to be able to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features described in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0049] The present invention is a shale micro-material intelligent detection method based on deep learning based on the Shale-Yolo target detection model. The structure diagram of the Shale-Yolo target detection model is as follows: Figure 6 The Shale-Yolo target detection model adds a small target detection layer to the Neck part of the Yolov8 target detection model. The structure of the small target detection layer is shown in Figure 2 As shown in the figure, the Shuffle attention mechanism module is added to the 10th and 20th layers of the head part. The structure of the Shuffle attention mechanism module is shown in the figure. Figure 3 As shown; the SE attention mechanism model is combined with the C2f_SE module to construct the C2f_SE module. The structure diagram of the SE attention mechanism module is as follows Figure 4 As shown in the figure, the Cf2 modules of the 1st, 2nd, 3rd and 4th layers of the head part are replaced with Cf2_SE modules; a BiFPN bidirectional feature pyramid network is added and merged with the original Concat module to establish a new Concat_BiFPN module. The BiFPN bidirectional feature pyramid network structure is shown in the figure. Figure 5 As shown in the figure, add the Concat_BiFPN module to the third and sixth layers of the head part of the model. By training the model, save the pre-trained weights. Open the shale SEM microscopic material target detection software, the software interface is as follows Figure 8 Click the "Model" button to import the trained Shale-Yolo model weight file; click the "DetectScreen" button to select the screen image to be imported, start real-time detection of the screen image, and realize dynamic target detection; click the "ImpertImage" button to select the image to be imported to realize static target detection. The detection effect is shown in the figure below. Fig. 9 shown.

[0050] The basic flow chart of the present invention is as follows Figure 7 As shown, a shale micro-matter intelligent detection method based on deep learning includes the following steps:

[0051] Obtain shale SEM-Maps images;

[0052] Based on SEM-Maps images, a high-quality SEM data set of shale was produced, and the data was divided into quartz, pyrite, organic matter, organic pores, and inorganic pores;

[0053] Based on the target type and Yolov8 target detection model, a Shale-Yolo target detection model is constructed to train the prepared high-quality data set and save the trained network model and weight file;

[0054] Based on the trained network model and weight file, open the shale scanning electron microscope microscopic material target detection software, import the trained network model and weight file, choose to import pictures to achieve static target detection, and select screen detection to achieve real-time recognition of dynamic screens.

[0055] The present invention further provides a method for constructing a Shale-Yolo target detection model of the shale micro-matter intelligent detection method, including:

[0056] S2.1: Prepare high-quality shale SEM data sets and classify the data into quartz, pyrite, organic matter, organic pores, and inorganic pores;

[0057] S2.2: Based on the Yolov8 target detection model, a Shale-Yolo target detection model is constructed;

[0058] S2.3: Add a small target detection layer in the Neck part, and perform detection after splicing the shallow feature map with the deep feature map, so as to improve the model's ability to recognize microscopic substances;

[0059] S2.4: Adding the Shuffle attention mechanism module to the 10th and 20th layers of the model head can effectively capture the relationship between global and local features and improve the representation ability of the model;

[0060] S2.5: Add the SE attention mechanism module, combine the SE attention mechanism model with the C2f_SE module, and construct the C2f_SE module;

[0061] S2.6: Replace the Cf2 modules in the first, second, third and fourth layers of the head part with Cf2_SE modules;

[0062] S2.7: Add the BiFPN bidirectional feature pyramid network and merge it with the original Concat module to establish a new Concat_BiFPN module;

[0063] S2.8: Add the Concat_BiFPN module to the third and sixth layers of the head part of the model to effectively integrate feature information at different levels and improve the model's ability to detect and recognize targets.

[0064] The present invention further comprises that in S2.3, the small target detection layer performs feature extraction through convolution operation, thereby enhancing detail capture capability and multi-scale feature extraction, while also improving model robustness.

[0065] The present invention further states that in S2.4, the Shuffle attention mechanism module is based on the multi-head self-attention mechanism in the Transformer model. It generates an output sequence of the same length as the input sequence by weighted summing each position in the input sequence. In this process, the model learns the importance of each position to the output, thereby being able to capture the correlation between different parts of the input sequence.

[0066] The present invention further states that in S2.5, the SE attention mechanism module introduces an attention mechanism on each channel of the convolutional neural network, and dynamically adjusts the weight of the channel features by learning the importance weight of each channel. Specifically, the SE attention mechanism includes two steps: squeeze and excitation. In the squeeze stage, the feature map of each channel is compressed by a global average pooling operation to obtain the global description information of the channel; in the excitation stage, the activation function of each channel is learned through two fully connected layers, and then the activation function is applied to the original features to obtain a weighted feature representation.

[0067] The present invention further comprises that in S2.6, in the C2f_SE module, after a series of convolution operations and feature fusion, the feature map is connected to the SE attention mechanism. In the SE attention mechanism, the feature map of each channel is compressed and the activation function is learned to dynamically adjust the weight of the channel feature, thereby enhancing the expression ability of important features.

[0068] The present invention further states that in S2.7, the BiFPN bidirectional feature pyramid network combines the idea of ​​a bidirectional information transfer mechanism and a feature pyramid network.

[0069] The feature pyramid network includes multiple feature pyramid levels of different scales, which are used to extract feature information of different scales; the bidirectional information transmission mechanism means that in the feature pyramid network, feature information can be transmitted from bottom to top and from top to bottom to achieve cross-level feature fusion and information transmission.

[0070] The present invention further comprises that in S2.8, the Concat_BiFPN module combines these different feature fusion methods, and splices feature information of different scales and levels through the Concat operation, thereby achieving a more comprehensive and rich feature representation.

[0071] The present invention provides a method for intelligent detection of shale micro-materials based on deep learning. The method can identify different shale micro-materials from shale SEM-Maps images, and has a strong ability to characterize features of different scales. It has high detection accuracy while ensuring that the detection speed meets the real-time requirements. The pre-trained shale micro-material recognition model, combined with functions such as image import and real-time detection screen, can effectively improve the recognition accuracy and speed of shale micro-materials, reduce human errors and workload; the algorithm has high recognition accuracy, and is more efficient and faster than manual recognition and traditional image processing methods for identifying pores; the detection frame rate is fast, with a single static image processing speed of only 203.1 milliseconds, and a screen dynamic real-time recognition speed of up to 91fps per second, which can fully meet the needs of engineering applications.

[0072] The present invention further provides that the method for using the shale scanning electron microscope microscopic material target detection software comprises the following steps:

[0073] S9.1: Open the shale SEM microscopic material target detection software, click the "Model" button, and import the weight file of the trained Shale-Yolo model;

[0074] S9.2: Click the "Detect Screen" button to select the imported screen image and start real-time detection of the screen image to achieve dynamic target detection;

[0075] S9.3: Click the “Impert Image” button to select the imported image to realize static target detection.

[0076] The present invention also provides a shale micro-matter intelligent detection system based on deep learning, which is realized by the above-mentioned shale micro-matter intelligent detection method based on deep learning.

[0077] The present invention discloses a method and system for intelligent detection of shale micro-materials based on deep learning. The method can identify different shale micro-materials from shale SEM-Maps images, and has a strong ability to characterize features of different scales; it has high detection accuracy while ensuring that the detection speed meets the real-time requirements. The pre-trained shale micro-material recognition model, combined with functions such as image import and real-time detection screen, can effectively improve the recognition accuracy and speed of shale micro-materials, reduce human errors and workload; the algorithm has high recognition accuracy, and is more efficient and faster than manual recognition and traditional image processing methods for identifying pores; the detection frame rate is fast, with a single static image processing speed of only 203.1 milliseconds, and a dynamic real-time screen recognition speed of up to 91fps per second, which can fully meet the needs of engineering applications.

[0078] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) containing computer-usable program code.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A shale micro-matter intelligent detection method based on deep learning, characterized in that: The following steps are involved: Obtain shale SEM-Maps images; Produce high-quality shale SEM data sets and classify the data into quartz, pyrite, organic matter, organic pores, and inorganic pores; Build the Shale-Yolo target detection model to train the prepared high-quality data set and save the trained network model and weight files; Open the shale SEM microscopic material target detection software, import the trained network model and weight file, select import image to achieve static target detection, select screen detection to achieve real-time recognition of dynamic screen; The method for constructing the Shale-Yolo target detection model includes: S2.1: Prepare high-quality shale SEM data sets and classify the data into quartz, pyrite, organic matter, organic pores, and inorganic pores; S2.2: Based on the Yolov8 target detection model, a Shale-Yolo target detection model is constructed; S2.3: Add a small target detection layer in the Neck part, and perform detection after concatenating the shallow feature map with the deep feature map; S2.4: Add the Shuffle attention mechanism module to the 10th and 20th layers of the model head to capture the relationship between global and local features; S2.5: Add the SE attention mechanism module, combine the SE attention mechanism model with the C2f_SE module, and construct the C2f_SE module; S2.6: Replace the Cf2 modules in the first, second, third and fourth layers of the head part with Cf2_SE modules; S2.7: Add the BiFPN bidirectional feature pyramid network and merge it with the original Concat module to establish a new Concat_BiFPN module; S2.8: Add the Concat_BiFPN module to the third and sixth layers of the head part of the model to integrate feature information at different levels; In S2.4, the Shuffle attention mechanism module is based on the multi-head self-attention mechanism in the Transformer model, which generates an output sequence with the same length as the input sequence by weighted summing each position in the input sequence; In S2.7, the BiFPN bidirectional feature pyramid network combines a bidirectional information transfer mechanism and a feature pyramid network; The feature pyramid network includes multiple feature pyramid levels of different scales, which are used to extract feature information of different scales; the bidirectional information transmission mechanism is specifically that in the feature pyramid network, feature information can be transmitted from bottom to top and from top to bottom, so as to realize cross-level feature fusion and information transmission.

2. According to claim 1, a method for intelligent detection of shale microscopic matter based on deep learning is characterized in that: In S2.3, the small target detection layer performs feature extraction through convolution operation to capture details and multi-scale features.

3. The method for intelligent detection of shale microscopic materials based on deep learning according to claim 1 is characterized in that: In S2.5, the SE attention mechanism module introduces an attention mechanism on each channel of the convolutional neural network, and dynamically adjusts the weight of the channel feature by learning the importance weight of each channel; The SE attention mechanism consists of two steps: squeeze and excitation; In the squeeze stage, the feature map of each channel is compressed through the global average pooling operation to obtain the global description information of the channel; in the excitation stage, the activation function of each channel is learned through two fully connected layers, and the activation function is applied to the original features to obtain the weighted feature representation.

4. The method for intelligent detection of shale microscopic matter based on deep learning according to claim 1 is characterized in that: In S2.6, in the C2f_SE module, the feature map is connected to the SE attention mechanism after a series of convolution operations and feature fusion; In the SE attention mechanism, the feature map of each channel is compressed and the activation function is learned, and the weight of the channel features is dynamically adjusted to enhance the expressiveness of important features.

5. The method for intelligent detection of shale microscopic matter based on deep learning according to claim 1 is characterized in that: In S2.8, the Concat_BiFPN module is used to combine different feature fusion methods and splice feature information of different scales and levels through the Concat operation.

6. The method for intelligent detection of shale microscopic matter based on deep learning according to claim 1 is characterized in that: The method for using the shale scanning electron microscope microscopic material target detection software comprises the following steps: S9.1: Open the Shale SEM Microscopic Material Target Detection Software, click the "Model" button, and import the weight file of the trained Shale-Yolo target detection model; S9.2: Click the "Detect Screen" button to select the imported screen image and start real-time detection of the screen image to achieve dynamic target detection; S9.3: Click the "Impert Image" button to select the imported image to realize static target detection.

7. A deep learning-based intelligent detection system for shale micro-matter, which is realized by the deep learning-based intelligent detection method for shale micro-matter described in any one of claims 1-6.

Images