Fluid inclusion automatic identification and grading method based on multi-focal plane imaging

Abstract

The invention discloses a fluid inclusion automatic identification and grading method based on multi-focal plane imaging. The method comprises the following steps: acquiring a plurality of local high-definition digital images covering the whole surface of a rock sample slice; inputting the plurality of local high-definition digital images into an optimized target detection model for identification to obtain a bounding box position and a phase state category of each fluid inclusion; calculating the actual physical size of each inclusion on the basis of the pixel size obtained by the bounding box position and in combination with calibration parameters of a microscope system, and grading the actual physical sizes by adopting a clustering algorithm; and synthesizing the plurality of local high-definition digital images into a slice panorama marked with the bounding box position of each fluid inclusion, and generating a statistical report containing the phase state category, the actual physical size and the grading information of each inclusion. According to the invention, efficient and automatic positioning, classification, grading and data statistics of the fluid inclusion in the rock sample slice are realized.

Description

Technical Field

[0001] This application relates to the field of geological and mineral analysis and rock and mineral identification technology, and more specifically, to an automatic identification and classification method for fluid inclusions based on multi-focal plane imaging. Background Technology

[0002] Fluid inclusions are important samples for studying the origin and evolution of ore-forming fluids. Analyzing their composition and temperature can reveal the origin, migration, and precipitation mechanisms of ore-forming materials. In practical studies, fluid inclusions are usually distributed at the microscopic scale in the microscopic field of rock sample thin sections. Their identification and analysis are highly dependent on microscopic imaging conditions, and traditional related research methods have significant limitations. (1) Inefficient and dependent on human experience: Traditional methods require researchers to manually move thin sections of rock samples under a microscope and search for inclusions field by field. This process is time-consuming and highly dependent on the operator's experience in adjusting the depth of focus and interpreting targets. When inclusions are small in scale and scattered in distribution, they are very easy to be missed due to improper selection of the focal plane, making it difficult to ensure the completeness and consistency of the results.

[0003] (2) Poor information integration: With the gradual introduction of component analysis methods such as laser Raman spectroscopy into modern inclusion research, traditional manual methods lack a systematic record of the spatial location information of inclusions. Inclusions found manually are difficult to accurately reproduce and locate in subsequent analysis equipment, which restricts the efficiency and reliability of joint analysis using multiple technologies.

[0004] (3) Limitations of existing image recognition methods: Although convolutional neural networks (CNNs) have made some progress in the field of image recognition, they usually adopt a "classify first, then locate" processing mode, which has low processing efficiency in microscopic images and is difficult to meet the rapid detection needs of large-area rock sample thin sections. Although the YOLO series models have shown high efficiency in target detection, their recognition stability is highly dependent on the clarity of the input image and the consistency of target features. The morphology and contrast of fluid inclusions change significantly at different focal depths, and the single focal plane image input method often cannot take into account all inclusions at the same time, resulting in small inclusions being easily missed in complex mineral backgrounds.

[0005] (4) Lack of automated size classification function: Existing methods usually rely on manual size classification of packages, which is time-consuming and prone to human error. Although some methods can perform preliminary package identification, they still lack automatic size statistics and classification functions, resulting in inconsistent analysis results and low efficiency.

[0006] In recent years, the concept of automation has been gradually introduced into fluid inclusion research. Existing technologies include schemes that employ automated scanning image acquisition techniques and utilize CNNs for inclusion identification and screening; schemes that improve the accuracy and efficiency of inclusion identification by introducing attention mechanisms and optimizing feature fusion networks; schemes that achieve preliminary automated positioning and adjustment of rock sample thin sections, replacing the traditional manual leveling process; and schemes that focus on the fine measurement of individual inclusions, achieving certain results in composition and volume measurements. However, these schemes have not effectively solved the problems of rapid positioning, counting, and hierarchical statistics of inclusions over a large area.

[0007] However, the aforementioned existing technologies mostly focus on improving image recognition algorithms or single experimental steps, and have not yet systematically solved the problems of complete acquisition, stable identification, and classification based on real physical size of inclusions under multi-depth-of-focus conditions from the perspective of the entire thin-film scale.

[0008] In summary, existing technologies still have long-standing unresolved technical problems in fluid inclusion research: (1) Under the microscopic imaging conditions of rock sample thin sections, fluid inclusions are highly sensitive to depth of focus, and existing identification methods based on single-focal-plane images are difficult to maintain stable identification of small inclusions under multi-depth of focus conditions; (2) Although microscopic images usually include scale information, existing methods still require experimenters to manually read the scale and measure the size of each inclusion, making it difficult to achieve batch, standardized acquisition and objective grading of the real physical size in the automatic identification process; (3) There is a lack of automated solutions that can integrate image acquisition, target identification, size analysis and result output at the full thin section scale, which makes it difficult to maintain consistency and traceability of detection results across different fields of view.

[0009] Therefore, there is an urgent need in the field for an automated method for identifying, classifying, and grading fluid inclusions that can cover the entire process from sample imaging to data output, in order to solve the above-mentioned technical problems. This is the core problem that this invention aims to solve. Summary of the Invention

[0010] To address the aforementioned issues, this application provides an automatic identification and classification method for fluid inclusions based on multi-focal plane imaging, aiming to solve problems such as low identification efficiency, large subjective bias, and discontinuous information links in thin sections.

[0011] The first aspect of this invention provides an automatic identification and classification method for fluid inclusions based on multi-focal plane imaging, comprising: By automatically scanning and stacking multi-focal plane images, multiple local high-resolution digital images covering the entire surface of a thin rock section are acquired, and the location information at the time of acquisition of each image is recorded. The multiple local high-definition digital images are input into the optimized target detection model for identification to obtain the bounding box position and phase category of each fluid inclusion; wherein, the optimized target detection model is built based on a single-stage detection architecture and introduces a channel feature adaptive enhancement module and a multi-scale spatial feature fusion module into its network; Based on the pixel dimensions obtained from the bounding box positions and combined with the calibration parameters of the microscope system, the actual physical dimensions of each inclusion are calculated, and a clustering algorithm is used to classify the actual physical dimensions. The multiple local high-definition digital images are combined to form a thin-slice panoramic image with the location of the boundary boxes of each fluid inclusion marked, and a statistical report containing the phase category, actual physical size and classification information of each inclusion is generated.

[0012] In one optional implementation, the acquisition of images via automatic scanning and multi-focal plane image stacking specifically includes: Control the three-axis stage to move along a preset serpentine path, and perform the following operations at each scanning stop point: drive the three-axis stage to move in the Z-axis direction, acquire at least three sub-images with different focal planes, and stack and fuse these sub-images to generate a local high-definition digital image of the entire surface of the rock sample thin section. Among them, the local high-definition digital images generated by adjacent scanning stops have an overlap area of ​​15%-20%.

[0013] In one optional implementation, the channel feature adaptive enhancement module performs global average pooling on the input feature map, performs one-dimensional convolution on the obtained pooling result and generates a channel weight vector through the Sigmoid activation function, and multiplies the channel weight vector with the input feature map.

[0014] In one optional implementation, the multi-scale spatial feature fusion module performs multiple max pooling operations with kernels of different sizes on the input feature map in parallel, and concatenates the pooling results with the input feature map along the channel dimension.

[0015] In one optional implementation, the use of a clustering algorithm for grading specifically involves using the K-means clustering algorithm to cluster the actual physical dimensions of all detected fluid inclusions, and automatically classifying them into different size levels based on the clustering results.

[0016] In one alternative implementation, the size class includes at least: less than 5 micrometers, 5 to 10 micrometers, and greater than 10 micrometers.

[0017] In one optional implementation, the synthesis of the panoramic thin-film image labeled with the location of each fluid inclusion bounding box specifically involves: stitching together multiple local high-definition digital images according to their location information at the time of acquisition to form a complete thin-film digital image, and overlaying a label layer composed of the bounding box location and phase category on it.

[0018] A second aspect of this invention provides an automatic identification and classification system for fluid inclusions based on multi-focal plane imaging, the system comprising: The image acquisition module is used to acquire multiple local high-resolution digital images covering the entire surface of a thin rock sample section through automatic scanning and multi-focal plane image stacking, and to record the location information of each image acquisition. The identification and classification module is used to input the multiple local high-definition digital images into the optimized target detection model for identification, and obtain the bounding box position and phase category of each fluid inclusion; wherein, the optimized target detection model is built based on a single-stage detection architecture, and a channel feature adaptive enhancement module and a multi-scale spatial feature fusion module are introduced into its network; A size grading module is used to calculate the actual physical size of each inclusion based on the pixel size obtained from the bounding box position and in conjunction with the calibration parameters of the microscope system, and to grade the actual physical size using a clustering algorithm; and The results output module is used to synthesize the multiple local high-definition digital images into a thin-film panoramic image with the location of the boundary boxes of each fluid inclusion marked, and to generate a statistical report containing the phase category, actual physical size and classification information of each inclusion.

[0019] A third aspect of the present invention provides an electronic device, characterized in that it includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements an automatic identification and classification method for fluid inclusions based on multi-focal plane imaging.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, a method for automatic identification and classification of fluid inclusions based on multi-focal plane imaging is provided.

[0021] This application has at least the following advantages or beneficial effects: (1) The efficiency and throughput of analysis have been improved by orders of magnitude. Through the fully automated image acquisition, recognition and analysis process, the high-throughput batch processing of massive fluid inclusions has been achieved, shortening the traditional "hour-level" analysis cycle to "minute-level", improving efficiency by tens of times and breaking the efficiency bottleneck of manual methods.

[0022] (2) Ensure the accuracy and objectivity of the analysis results. Based on the improved deep learning model for identification and classification, avoid misjudgment and missed detection caused by human subjective experience and visual fatigue. Combine quantitative methods such as clustering algorithms to ensure that the size classification and statistical results are consistent and repeatable, and provide reliable geological research data.

[0023] (3) The analysis results are intuitive and provide strong decision support. It automatically generates a panoramic view of the thin section with integrated spatial location and a comprehensive report and visualization chart containing the type, size and grade of the inclusions, making the micro data clear and easy to understand, greatly improving the efficiency of result interpretation, and providing a reference for subsequent experiments.

[0024] (4) The technology has broad applicability and promising application prospects. As a general fluid inclusion identification and analysis solution, it is not limited to specific mineral types or geological backgrounds. It can be widely applied to the identification and processing of various minerals and has broad potential for technology transfer and industrialization.

[0025] (5) To achieve traceability and experimental reproduction of the spatial location of the package, researchers can control the stage to move automatically to the target package location by associating the identification results with the spatial coordinates of the three-axis stage, so as to achieve accurate reproduction and positioning, and provide reliable support for subsequent component analysis. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart of an embodiment of the present application of an automatic identification and classification method for fluid inclusions based on multi-focal plane imaging; Figure 2 This is a structural diagram of an automatic identification and classification system for fluid inclusions based on multi-focal plane imaging, as proposed in an embodiment of this application. Figure 3 This is a schematic diagram of a scanning path proposed in an embodiment of this application; Figure 4 This is a schematic diagram of a photo stack according to an embodiment of this application; Figure 5 This is a flowchart illustrating the functional structure and data processing of a detection module according to an embodiment of this application; Figure 6 This is a schematic diagram of a panoramic image of a thin rock section with marked fluid inclusion locations, as proposed in an embodiment of this application. Figure 7 This is an example of an embodiment of the present application that provides an output report of fluid inclusion identification, size, and classification results; Figure 8 This is a schematic diagram of an electronic device according to this application. Detailed Implementation

[0028] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0029] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of an automatic identification and classification method for fluid inclusions based on multi-focal plane imaging proposed in this application. Figure 1 As shown, an automatic identification and classification method for fluid inclusions based on multi-focal plane imaging is characterized by comprising: S100: Through automatic scanning and multi-focal plane image stacking, it acquires multiple local high-definition digital images covering the entire surface of a rock sample thin section, and records the location information of each image acquisition. Specifically, the process of acquiring images through automatic scanning and multi-focal plane image stacking includes: Control the three-axis stage to move along a preset serpentine path, and perform the following operations at each scanning stop point: drive the three-axis stage to move in the Z-axis direction, acquire at least three sub-images with different focal planes, and stack and fuse these sub-images to generate a local high-definition digital image of the entire surface of the rock sample thin section. Among them, the local high-definition digital images generated by adjacent scanning stops have an overlap area of ​​15%-20%.

[0030] In this embodiment, this step aims to acquire a local high-resolution digital image covering the entire surface of the rock sample thin section without omissions and with clear images at all focal depths, and to associate it with location information to provide a reliable data foundation for subsequent identification. The specific steps are as follows: Equipment preparation and section fixation: A three-axis automatic stage capable of micro-stepping along the XYZ axes is selected. It communicates with the microscopic imaging system, allowing precise control of the spatial position of the rock sample section and recording its coordinates. The prepared rock sample section is fixed in the center of the stage, ensuring the section is flat and without displacement.

[0031] Scanning path planning and movement control: The control program automatically generates a serpentine scanning path based on the shape and size parameters of the rock sample thin section, such as... Figure 3As shown, the path starts at "Start" and moves in a serpentine pattern along the surface of the thin section to "Finish," ensuring no observation field is missed. The stage is controlled to move along this path, and the spacing between adjacent scanning stops is set according to the microscope's field of view size to ensure that the subsequently generated local images meet the overlap requirements.

[0032] Multi-focal-plane image acquisition and stacking: At each scanning stop, the stage is driven to move in micro-steps along the Z-axis (vertical direction) to perform autofocus, and the image acquisition unit acquires at least three sub-images of different focal planes. At each scanning stop, the spatial coordinates of the three-axis stage are associated with the acquired image data and stored to accurately reproduce the position of the package in subsequent analysis. During the acquisition process, the system evaluates the sharpness of each focal-plane sub-image in real time. If the sharpness difference between adjacent sub-images exceeds a preset threshold, the number of focal planes acquired is automatically increased until the effective depth of field of view is covered. Subsequently, all sub-images at the same stop are stacked and fused to generate a single clear, locally high-resolution digital image, such as... Figure 4 As shown, the stacked image can simultaneously present inclusions at different depths of focus, avoiding missed detections on a single focal plane.

[0033] Image numbering and location association: For local high-resolution digital images after stacking, they are uniformly numbered according to their row and column positions in the scanning path (e.g., "6-2" represents the 6th row and 2nd column field of view), and the number is associated with the spatial coordinate information of the three-axis stage when the image was acquired and stored to provide a basis for subsequent spatial reproduction and positioning.

[0034] Overlap area control: A 15%-20% overlap area is maintained between the local high-definition digital images generated by adjacent scanning stops to ensure that no boundary information is lost during subsequent panoramic image stitching and to guarantee the continuity of the full-slice image.

[0035] Full thin section image sequence acquisition: Repeat the above scanning, acquisition, and stacking operations until the stage completes all movement along the preset path to obtain a local high-definition digital image sequence covering the entire surface of the rock sample thin section.

[0036] S200: Input the multiple local high-definition digital images into the optimized target detection model for identification to obtain the bounding box position and phase category of each fluid inclusion; wherein, the optimized target detection model is built based on a single-stage detection architecture and introduces a channel feature adaptive enhancement module and a multi-scale spatial feature fusion module into its network; Specifically, the channel feature adaptive enhancement module performs global average pooling on the input feature map, performs one-dimensional convolution on the obtained pooling result and generates a channel weight vector through the Sigmoid activation function, and multiplies the channel weight vector with the input feature map.

[0037] The multi-scale spatial feature fusion module performs multiple max pooling operations with kernels of different sizes in parallel on the input feature map, and concatenates the pooling results with the input feature map in the channel dimension.

[0038] In this embodiment, this step employs an optimized target detection model based on a single-stage detection architecture to identify the image sequence acquired in S100, outputting the bounding box position and phase category of the inclusion, thus solving the challenge of identifying small inclusions in complex backgrounds. The specific steps are as follows: Model Foundation: The target detection model is built on a single-stage detection architecture, possessing efficient processing capabilities for "simultaneous localization and classification," adaptable to the rapid detection needs of large-area rock sample thin sections. The model integrates a feature extraction unit, a feature enhancement unit (corresponding to a channel feature adaptive enhancement module), a multi-scale feature fusion unit (corresponding to a multi-scale spatial feature fusion module), and a detection output unit. These units work collaboratively to complete the recognition task, such as... Figure 5 As shown.

[0039] The adaptive enhancement module for channel features is implemented as follows: This module performs global average pooling on the input feature map extracted by the model to compress the spatial dimension and retain channel feature information; it performs one-dimensional convolution on the pooling result to adaptively calibrate the channel response; it generates a channel weight vector through the Sigmoid activation function, which assigns high weights to feature channels related to fluid inclusions and low weights to feature channels corresponding to mineral background textures; the channel weight vector is multiplied with the input feature map to enhance key features of inclusions and suppress background interference, thereby improving the distinguishability of inclusions in complex backgrounds.

[0040] Multi-scale spatial feature fusion module implementation: This module performs multiple max pooling operations with kernels of different sizes in parallel on the input feature map, capturing the detailed features of small inclusions and the overall features of larger inclusions respectively; the pooling results are concatenated with the original input feature map in the channel dimension to fuse multi-scale spatial context information, enabling the model to stably identify fluid inclusions of different sizes, especially improving the detection capability of small inclusions (<5μm).

[0041] Identification and classification process: The local high-definition digital image sequence acquired by S100 is input into the optimized target detection model. The model performs feature extraction, enhancement and fusion processing, and outputs the bounding box coordinates (location information) of each fluid inclusion and the probability value of the corresponding phase category in parallel. The system determines the final phase classification result based on the probability value, including gas-liquid two-phase inclusions, pure liquid phase inclusions and inclusions containing daughter mineral phases, and summarizes them to form an intermediate dataset containing inclusion number, bounding box position and phase category.

[0042] S300: Based on the pixel size obtained from the bounding box position and combined with the calibration parameters of the microscope system, calculate the actual physical size of each inclusion, and use a clustering algorithm to classify the actual physical size; Specifically, the method of using clustering algorithms for grading involves using the K-means clustering algorithm to cluster the actual physical dimensions of all detected fluid inclusions, and automatically classifying them into different size levels based on the clustering results.

[0043] The size categories include at least: less than 5 micrometers, 5 to 10 micrometers, and greater than 10 micrometers.

[0044] In this embodiment, this step automatically calculates and classifies the actual physical size of the package based on the recognition results, ensuring that the classification results are objective and consistent.

[0045] Pixel size extraction: Extract the pixel size parameters of each fluid inclusion bounding box in the image from the intermediate dataset output by S200, including the pixel length in the horizontal and vertical directions.

[0046] Actual physical size conversion: The system calls pre-stored microscope imaging magnification parameters and pixel-physical size calibration coefficients to automatically convert the extracted pixel size, obtaining the true physical size data of each inclusion (e.g., "9.6 × 9.8 μm"). This process requires no manual intervention, avoiding subjective errors caused by manually reading scale bars and measuring boundaries, and ensuring the consistency of size data at the full-section scale.

[0047] K-means clustering grading: The K-means clustering algorithm was used to perform unsupervised modeling of the true physical dimensions of all inclusions. Grading thresholds were automatically assigned based on the natural distribution characteristics of the size data, resulting in at least three size categories: less than 5 micrometers (relatively small), 5 to 10 micrometers (medium), and greater than 10 micrometers (relatively large). The grading results were adaptively generated from the sample data, avoiding the subjectivity of manually set thresholds, and achieving over 98% consistency with manually graded results. This grading was used to facilitate the selection of appropriately sized inclusions for subsequent experiments.

[0048] S400: Combine the multiple local high-definition digital images to form a thin-film panoramic image with the location of the boundary boxes of each fluid inclusion marked, and generate a statistical report containing the phase category, actual physical size and classification information of each inclusion.

[0049] Specifically, the process of synthesizing a panoramic thin-slice image with the bounding box positions of each fluid inclusion is as follows: multiple local high-definition digital images are stitched together according to their position information at the time of acquisition to form a complete thin-slice digital image, and a label layer composed of the bounding box positions and phase categories is superimposed on it.

[0050] In this embodiment, this step records the spatial coordinate information of each fluid inclusion, binds it to the position information of each local high-definition digital image, stores the spatial coordinates and generates a spatial positioning dataset, so that subsequent experiments can accurately reproduce the position of the fluid inclusion based on these spatial coordinate information.

[0051] Thin section panoramic image synthesis: Based on the image IDs and spatial coordinate information stored in the S100, all local high-resolution digital images are stitched together to form a complete digital image covering the entire surface of the rock sample thin section; an inclusion bounding box location and phase category annotation layer output from the S200 is overlaid on this image to generate a thin section panoramic image labeled with inclusion IDs, locations, and phases, such as... Figure 6 As shown, the panoramic image clearly marks the location and key information of each package.

[0052] Statistical report generation: Integrating the phase classification results of S200, the actual physical dimensions and grading results of S300, and the image numbering of S100, structured statistical reports are generated, such as... Figure 7 As shown, the report includes the unique identifier of the inclusion, the image number it belongs to, the phase category, the actual physical size, and the size classification, providing a clear basis for the selection of subsequent experimental subjects.

[0053] Spatial reproduction and positioning support: The spatial coordinate information in the statistical report is linked with the control system of the three-axis stage. When it is necessary to perform subsequent component analysis on a specific inclusion (such as laser Raman spectroscopy test), the coordinate information in the report can be used to control the stage to automatically move to the location of the target inclusion, so as to achieve accurate reproduction and positioning of the microscope field of view, forming a closed-loop support of "identification-screening-reproduction".

[0054] Please refer to Figure 2 , Figure 2 This is a structural diagram of an automatic fluid inclusion identification and classification system based on multi-focal plane imaging, proposed in an embodiment of this application. Figure 2 As shown, this disclosure also provides an automatic identification and classification system for fluid inclusions based on multi-focal plane imaging. The system includes: an image acquisition module 201, an identification and classification module 202, a size classification module 203, and a result output module 204; wherein, The image acquisition module 201 is used to acquire multiple local high-definition digital images covering the entire surface of a rock sample thin section through automatic scanning and multi-focal plane image stacking, and to record the location information of each image acquisition. The identification and classification module 202 is used to input the multiple local high-definition digital images into the optimized target detection model for identification, and obtain the bounding box position and phase category of each fluid inclusion; wherein, the optimized target detection model is built based on a single-stage detection architecture, and a channel feature adaptive enhancement module and a multi-scale spatial feature fusion module are introduced into its network; The size grading module 203 is used to calculate the actual physical size of each inclusion based on the pixel size obtained from the bounding box position and in combination with the calibration parameters of the microscope system, and to grade the actual physical size using a clustering algorithm; and the result output module 204 is used to synthesize the multiple local high-definition digital images into a thin-section panoramic image with the bounding box positions of each fluid inclusion marked, and to generate a statistical report containing the phase category, actual physical size and grading information of each inclusion.

[0055] This disclosure also provides an electronic device, please refer to... Figure 8 , Figure 8 This is a schematic diagram of an electronic device illustrated in an embodiment of this disclosure. For example... Figure 8 As shown, the electronic device 100 includes a memory 110 and a processor 120. The memory 110 and the processor 120 are connected via a bus for communication. The memory 110 stores a computer program that can run on the processor 120 to implement the steps in the automatic identification and classification method for fluid inclusions based on multi-focal plane imaging disclosed in this embodiment.

[0056] The disclosed embodiments also provide a computer-readable storage medium that, when the instructions in the computer-readable storage medium are executed by a processor of a computer device, enables the computer device to perform the steps in the automatic identification and classification method for fluid inclusions based on multi-focal plane imaging as described in the embodiments of this disclosure.

[0057] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, systems, electronic devices, and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0058] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction set implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0059] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0060] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0061] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0062] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0063] The above provides a detailed description of an automatic identification and classification method for fluid inclusions based on multi-focal plane imaging provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and its core ideas. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for automatic identification and classification of fluid inclusions based on multi-focal plane imaging, characterized in that, include: By automatically scanning and stacking multi-focal plane images, multiple local high-resolution digital images covering the entire surface of a thin rock section are acquired, and the location information at the time of acquisition of each image is recorded. The multiple local high-definition digital images are input into the optimized target detection model for identification to obtain the bounding box position and phase category of each fluid inclusion; wherein, the optimized target detection model is built based on a single-stage detection architecture and introduces a channel feature adaptive enhancement module and a multi-scale spatial feature fusion module into its network; Based on the pixel dimensions obtained from the bounding box positions and combined with the calibration parameters of the microscope system, the actual physical dimensions of each inclusion are calculated, and a clustering algorithm is used to classify the actual physical dimensions. The multiple local high-definition digital images are combined to form a thin-slice panoramic image with the location of the boundary boxes of each fluid inclusion marked, and a statistical report containing the phase category, actual physical size and classification information of each inclusion is generated.

2. The method for automatic identification and classification of fluid inclusions based on multi-focal plane imaging according to claim 1, characterized in that, The process of acquiring images through automatic scanning and multi-focal plane image stacking specifically includes: Control the three-axis stage to move along a preset serpentine path, and perform the following operations at each scanning stop point: drive the three-axis stage to move in the Z-axis direction, acquire at least three sub-images with different focal planes, and stack and fuse these sub-images to generate a local high-definition digital image of the entire surface of the rock sample thin section. Among them, the local high-definition digital images generated by adjacent scanning stops have an overlap area of ​​15%-20%.

3. The method for automatic identification and classification of fluid inclusions based on multi-focal plane imaging according to claim 1, characterized in that, The channel feature adaptive enhancement module performs global average pooling on the input feature map, performs one-dimensional convolution on the pooling result, generates a channel weight vector through the Sigmoid activation function, and multiplies the channel weight vector with the input feature map.

4. The method for automatic identification and classification of fluid inclusions based on multi-focal plane imaging according to claim 1, characterized in that, The multi-scale spatial feature fusion module performs multiple max pooling operations with kernels of different sizes in parallel on the input feature map, and concatenates the pooling results with the input feature map in the channel dimension.

5. The method for automatic identification and classification of fluid inclusions based on multi-focal plane imaging according to claim 1, characterized in that, The method of using clustering algorithms for grading specifically involves using the K-means clustering algorithm to cluster the actual physical dimensions of all detected fluid inclusions, and automatically classifying them into different size levels based on the clustering results.

6. The method for automatic identification and classification of fluid inclusions based on multi-focal plane imaging according to claim 5, characterized in that, The size categories include at least: less than 5 micrometers, 5 to 10 micrometers, and greater than 10 micrometers.

7. The method for automatic identification and classification of fluid inclusions based on multi-focal plane imaging according to claim 1, characterized in that, The process of synthesizing a panoramic thin-slice image with the bounding box positions of each fluid inclusion is as follows: multiple local high-definition digital images are stitched together according to their position information at the time of acquisition to form a complete thin-slice digital image, and a label layer composed of the bounding box positions and phase categories is superimposed on it.

8. The automatic identification and classification system for fluid inclusions based on multi-focal plane imaging according to any one of claims 1-7, characterized in that, The system includes: The image acquisition module is used to acquire multiple local high-resolution digital images covering the entire surface of a thin rock sample section through automatic scanning and multi-focal plane image stacking, and to record the location information of each image acquisition. The identification and classification module is used to input the multiple local high-definition digital images into the optimized target detection model for identification, and obtain the bounding box position and phase category of each fluid inclusion; wherein, the optimized target detection model is built based on a single-stage detection architecture, and a channel feature adaptive enhancement module and a multi-scale spatial feature fusion module are introduced into its network; The size grading module is used to calculate the actual physical size of each inclusion based on the pixel size obtained from the bounding box position and in combination with the calibration parameters of the microscope system, and to grade the actual physical size using a clustering algorithm; and the result output module is used to synthesize the multiple local high-definition digital images into a thin-section panoramic image with the bounding box positions of each fluid inclusion marked, and to generate a statistical report containing the phase category, actual physical size and grading information of each inclusion.

9. An electronic device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the automatic identification and classification method for fluid inclusions based on multifocal plane imaging as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the automatic identification and classification method for fluid inclusions based on multi-focal plane imaging as described in any one of claims 1 to 7.

Images