Rock core image recognition method based on deep learning

Through the core image recognition method based on deep learning, the problem of poor core photo acquisition accuracy and reliance on human eyes in engineering geological drilling is solved, and the core types are automatically identified, which improves the recognition efficiency and accuracy.

CN119006898APending Publication Date: 2024-11-22CHINA RAILWAY SIYUAN SURVEY & DESIGN GRP CO LTD +1
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202411023906.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

In engineering geological drilling, core photos are acquired relying on manual shooting, resulting in poor accuracy and low recognition rate. In addition, traditional core recognition relies on human eye observation, which is subjective and difficult to ensure accuracy and real-time.

Method used

The core image recognition method based on deep learning is adopted to automatically identify core types by collecting core images, standardizing processing, and building recognition models, reducing artificial interference and improving image clarity.

Benefits of technology

It effectively improves the efficiency and accuracy of core image recognition, overcomes manual judgment errors, and achieves higher recognition accuracy and real-time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006898A_ABST
    Figure CN119006898A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of image recognition, and discloses a rock core image recognition method based on deep learning, and the method comprises the steps: collecting a rock core image; performing standardization processing on the rock core image to obtain a rock core image histogram; constructing a rock core identification model based on the rock core image histogram; and identifying a to-be-identified rock core image through the rock core identification model to obtain an identification frame and a rock core type. According to the invention, the efficiency and accuracy of rock core image recognition can be effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of image recognition technology, and in particular to a core image recognition method based on deep learning. Background Art

[0002] Engineering geological drilling is one of the important means for various projects to obtain underground rock and soil information. Drilling core photos can directly reflect the actual situation of underground rock and soil, and are important original data for preserving geological information. At present, in engineering geological drilling work, the acquisition of core photos is still mainly based on manual shooting, which consumes a lot of manpower. In addition, due to different shooting equipment, shooting angles, height conditions, etc., the obtained core photos have poor accuracy and low recognition rate. After the drilling work is completed, it is difficult to verify the stratigraphic information.

[0003] Secondly, the shooting of core photos is affected by the environment, and there are a series of uncertain factors such as different light intensities, mutual occlusion between objects, shadow coverage, etc., which will also cause the core photos to be less clear. In addition, adjacent projects in the same area are often instructive, and the drilling data of adjacent projects can be referred to in the early stage of the project research. Therefore, it is also necessary to use pre-processing methods to standardize the core photos accumulated in the early stage to a certain format for unified use.

[0004] In addition, traditional core identification mainly relies on human visual observation, analysis and extraction of rock features, and on this basis, the types of cores are distinguished. However, the accuracy of judgment depends on personal geological level and engineering experience, which is highly subjective. Different personnel in the same project have different judgment standards, so the quality and accuracy cannot be guaranteed, and it is difficult to obtain real-time feedback.

[0005] Therefore, how to effectively improve the efficiency and accuracy of core image recognition is a problem that needs to be solved urgently.

[0006] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0007] The main purpose of this application is to provide a core image recognition method based on deep learning, aiming to solve the technical problem of how to effectively improve the efficiency and accuracy of core image recognition.

[0008] To achieve the above objectives, the present application proposes a core image recognition method based on deep learning, the method comprising:

[0009] Acquisition of core images;

[0010] Performing standardization processing on the core image to obtain a core image histogram;

[0011] constructing a core recognition model based on the core image histogram;

[0012] The core image to be identified is identified by the core identification model to obtain an identification frame and a core type.

[0013] In one embodiment, the step of normalizing the core image to obtain the core boundary includes:

[0014] Performing perspective correction on the core image to obtain a corrected core image;

[0015] De-noising the corrected core image to obtain a target core image;

[0016] Determine the core boundary according to the target core image;

[0017] The core image is segmented and spliced ​​according to the core boundary to obtain a core image histogram.

[0018] In one embodiment, denoising the corrected core image to obtain a target core image includes:

[0019] Performing edge detection on the corrected core image to obtain image edge information and a preliminarily denoised core image;

[0020] The core image after the preliminary denoising is subjected to secondary denoising according to the image edge information to obtain a target core image.

[0021] In one embodiment, determining the core boundary according to the target core image includes:

[0022] Determine a standard straight line and a connected domain according to the target core image;

[0023] Extract the edge straight line of the core column according to the connected domain to obtain the actual straight line;

[0024] The standard straight line and the actual straight line are combined to obtain a core boundary.

[0025] In one embodiment, constructing a core recognition model based on the core image histogram includes:

[0026] Construct an initial core recognition model based on the yoloV8 neural network;

[0027] Preprocessing is performed according to the core image histogram to obtain a data set;

[0028] The initial core recognition model is trained and verified according to the data set to obtain a core recognition model.

[0029] In one embodiment, the preprocessing is performed according to the core image histogram to obtain a data set, including:

[0030] Clipping the core image histogram to obtain a clipped core image;

[0031] Performing lithology labeling on the cropped core image to obtain a labeled core image;

[0032] A data set is determined according to the annotated core image histogram.

[0033] In one embodiment, the training and verification of the initial core identification model according to the data set to obtain the core identification model includes:

[0034] Dividing the data set according to a preset ratio to obtain a training set and a test set;

[0035] Training the initial core recognition model using the training set to obtain a trained initial core recognition model;

[0036] Verifying the trained initial core recognition model through the test set to obtain the model accuracy;

[0037] When the model accuracy reaches a preset accuracy threshold or the current training times reaches a preset training times threshold, the trained initial core recognition model is used as the core recognition model;

[0038] When the model accuracy reaches a preset accuracy threshold and the current training times do not reach a preset training times threshold, the parameters of the initial core identification model are adjusted and the step of training the initial core identification model through the training set to obtain a trained initial core identification model is performed.

[0039] In addition, to achieve the above purpose, the present application also proposes a core image recognition device based on deep learning, and the core image recognition device based on deep learning includes:

[0040] An acquisition module, used for acquiring core images;

[0041] A processing module, used for performing standardization processing on the core image to obtain a core image histogram;

[0042] A construction module, used for constructing a core recognition model based on the core image histogram;

[0043] The identification module is used to identify the core image to be identified through the core identification model to obtain an identification frame and a core type.

[0044] In addition, to achieve the above-mentioned objectives, the present application also proposes a core image recognition device based on deep learning, the device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the core image recognition method based on deep learning as described above.

[0045] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the core image recognition method based on deep learning as described above are implemented.

[0046] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the core image recognition method based on deep learning as described above.

[0047] The present application provides a core image recognition method based on deep learning. The present application first collects core images, thereby accurately capturing the microstructure and features of the core and reducing the interference of human photography; the core images are standardized to obtain a core image bar graph, thereby improving image clarity; a core recognition model is constructed based on the core image bar graph, thereby effectively improving model accuracy; the core image to be recognized is recognized by the core recognition model to obtain an identification frame and a core type, thereby effectively improving the efficiency and accuracy of core image recognition.

[0048] In summary, the present application obtains a core image bar graph by collecting core images and performing standardized processing, and then constructs a core recognition model based on the core image bar graph to identify the core type of the core image to be identified, thereby overcoming the technical defects of errors in manual judgment of rock and soil geological categories, low recognition efficiency and accuracy, and can effectively improve the efficiency and accuracy of core image recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0050] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0051] Figure 1A flowchart diagram of the first embodiment of the core image recognition method based on deep learning provided in this application;

[0052] Figure 2 This is a schematic diagram of the structure of a core image automatic acquisition device of a core image recognition method based on deep learning according to the first embodiment of the present application;

[0053] Figure 3 A flow chart of the second embodiment of the core image recognition method based on deep learning provided in this application;

[0054] Figure 4 A core image histogram of a core image recognition method based on deep learning according to the second embodiment of the present application;

[0055] Figure 5 A flowchart diagram of the third embodiment of the core image recognition method based on deep learning provided in this application;

[0056] Figure 6 This is a schematic diagram of the module structure of a core image recognition device based on deep learning in an embodiment of the present application;

[0057] Figure 7 Schematic diagram of the device structure of the hardware operating environment involved in the core image recognition method based on deep learning in the embodiment of the present application.

[0058] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0059] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0060] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0061] The main solution of the embodiment of the present application is: collecting core images; standardizing the core images to obtain a core image bar graph; building a core recognition model based on the core image bar graph; and recognizing the core image to be recognized by the core recognition model to obtain an identification frame and a core type.

[0062] Traditional core identification mainly relies on human observation, analysis and extraction of rock features, and on this basis, the core types are distinguished. However, the accuracy of judgment depends on personal geological level and engineering experience, which is highly subjective. In addition, different personnel in the same project have different judgment standards, and the quality and accuracy cannot be guaranteed, and it is difficult to get real-time feedback. Therefore, how to effectively improve the efficiency and accuracy of core image recognition is a problem that needs to be solved urgently.

[0063] The present application acquires core images and performs standardized processing to obtain a core image bar graph, and then constructs a core recognition model based on the core image bar graph to identify the core type of the core image to be identified. This overcomes the technical defects of errors in manual judgment of rock and soil geological categories, low recognition efficiency and accuracy, and can effectively improve the efficiency and accuracy of core image recognition.

[0064] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a core image recognition device based on deep learning, etc. The following takes the core image recognition device based on deep learning as an example to illustrate this embodiment and the following embodiments.

[0065] Based on this, the present application embodiment provides a core image recognition method based on deep learning, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the core image recognition method based on deep learning in this application.

[0066] In this embodiment, the core image recognition method based on deep learning includes steps S10 to S40:

[0067] Step S10, collecting core images.

[0068] It should be noted that core images are an important carrier for recording core information. By analyzing core images, various attribute information of the core can be obtained, such as the color, composition, structure, texture, type, etc. of the core. The core image automatic acquisition device is used to acquire images of the core and obtain the original image data of the core, providing basic data for subsequent analysis and processing.

[0069] It can be understood that the engineering geological drilling core images are collected in a standardized manner through the core image automatic acquisition equipment to obtain high-definition core images, which are stored in the cloud according to the hole number and depth, reducing the interference of human shooting and realizing the online storage of core images.

[0070] It is worth noting that the automatic core image acquisition equipment includes: a detachable bracket, a high-definition camera, a level, a display screen, a storage module, a communication module, and a cloud server. The bracket is used to support the main part of the equipment, including other parts except the bracket. The bracket types include a tripod bracket and a clamp bracket. The tripod bracket is used for flat terrain, and the clamp bracket is used for steep mountainous terrain. The high-definition camera is an image acquisition device with a pixel of not less than 8 million. The model is not required here. The level is used to adjust the relative level of the equipment. The display screen is used to display the power and the number of stored images; the storage module is used to temporarily store core images, and the communication module is used to upload core images to the cloud. The cloud server stores all collected core images according to the hole number and depth, and can also import completed core images through an external interface.

[0071] like Figure 2 As shown, Figure 2 The invention is a schematic diagram of the structure of the automatic core image acquisition device, which includes: a detachable bracket (1), a high-definition camera (2), a level (3), a display screen (4), a storage module (5), and a communication module (6).

[0072] It is worth mentioning that the core image acquisition process includes: fixing the detachable bracket on the site, or clamping it on the short side of the core box, using a spirit level to adjust the horizontal state of the equipment, binding the equipment according to the drilling hole number and hole depth, pre-collecting the core image with a high-definition camera, adjusting and fixing the bracket height according to the image acquisition situation, setting a timed collection of core images, and after the collection is completed, wirelessly transmitting the images to the cloud server through the communication module. The cloud server stores all collected core images according to the hole number and depth, and can also import completed core images through an external interface.

[0073] In the specific implementation, the core image automatic acquisition equipment is supported in the middle of the site with a tripod bracket. In areas with limited site space, a clamp-type bracket can be selected to fix it on the core box. A QR code is made according to the drilling hole number and drilling depth to bind the equipment. Secondly, when collecting images for the first time, the distance between the high-definition camera and the core box needs to be adjusted to ensure that all four corner points of the core box are in the entry, and the level is adjusted to ensure that the equipment is in a horizontal state and the camera angle is horizontal. Finally, the core image automatic acquisition device is used to obtain high-definition core images, which are classified and stored according to different hole numbers and depths, and uploaded to the cloud processor through wireless transmission signals. When the on-site signal is not good, it can be temporarily stored in the device and automatically uploaded after the signal is restored.

[0074] Step S20, performing standardization processing on the core image to obtain a core image histogram.

[0075] It should be noted that during the acquisition process of core images, due to the influence of factors such as acquisition equipment and acquisition environment, there may be problems such as low image quality and noise interference. Therefore, it is necessary to standardize the core images to improve the image quality and clarity.

[0076] It can be understood that the core image histogram is an image form that intuitively displays the core information by arranging and combining the core images according to certain rules after standardization. Through the core image histogram, the hierarchical structure, color change, component distribution and other characteristics of the core can be more clearly observed and analyzed, providing an important basis for subsequent core identification.

[0077] In a specific implementation, the standardized processing of the core image includes steps such as correction, denoising, and enhancement, which are not specifically limited in this embodiment. These steps can effectively eliminate noise and interference in the image and improve the clarity and contrast of the image. The core image can also be cropped, scaled, and other operations to make the size and proportion of different core images consistent, which is convenient for subsequent processing and analysis. According to the depth information of the core image, the processed core images are arranged in depth order to form a core image bar graph.

[0078] Step S30: constructing a core recognition model based on the core image histogram.

[0079] It should be noted that the core recognition model is a neural network model used to recognize and classify core images. By learning and analyzing the core image histogram, the core recognition model can learn a variety of feature information of the core, thereby realizing automatic recognition and classification of the core.

[0080] In the specific implementation, the core recognition model can be constructed using deep learning technology, such as the yoloV8 neural network. By building a suitable network structure and parameter configuration, the model can learn effective feature representations from the core image histogram and output corresponding recognition results. In order to improve the recognition performance and generalization ability of the model, data enhancement, transfer learning and other technologies can also be used to optimize and improve the model.

[0081] Step S40: The core image to be identified is identified by using the core identification model to obtain an identification frame and a core type.

[0082] It should be noted that the trained core recognition model can automatically identify and classify new core images. The core image to be identified is input into the core recognition model. The model will identify the image based on the learned feature representation and output the identification frame and core type of each core. The identification frame is a rectangular frame used to mark the position and range of the core in the image, and the core type is derived based on the classification results of the core features by the model.

[0083] It is understandable that the core recognition model will extract features and classify the input core image to be identified, and then output the corresponding identification frame and core type according to the classification result. This automatic recognition and classification method can greatly improve the efficiency and accuracy of core recognition and reduce the possibility of manual intervention and errors.

[0084] The overall process of the core image recognition method based on deep learning includes: the core image acquisition equipment collects raw data, the core image is uploaded to the cloud server, and the core image standardization system processes the image data, including: image perspective correction technology corrects the image, Canny edge detection algorithm denoises, morphological secondary denoises and compiles algorithms to complete core image segmentation, splicing, and cropping, establish a core image data set, establish a core image recognition model based on the yoloV8 network, perform model training, model evaluation and model testing. The model evaluation includes evaluation based on the loss function, model accuracy and recall rate. Finally, the intelligent recognition of core images based on the yoloV8 network is completed.

[0085] The present embodiment provides a core image recognition method based on deep learning. The present embodiment first collects core images, thereby accurately capturing the microstructure and features of the core and reducing the interference of human photography; the core images are standardized to obtain a core image bar graph, thereby improving image clarity; a core recognition model is constructed based on the core image bar graph, thereby effectively improving model accuracy; the core image to be recognized is recognized by the core recognition model to obtain an identification frame and a core type, thereby effectively improving the efficiency and accuracy of core image recognition.

[0086] In summary, this embodiment acquires core images and performs standardization processing to obtain a core image histogram, and then constructs a core recognition model based on the core image histogram to identify the core type of the core image to be identified, thereby overcoming the technical defects of errors in manual judgment of rock and soil geological categories, low recognition efficiency and accuracy, and can effectively improve the efficiency and accuracy of core image recognition.

[0087] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can refer to the above introduction, and will not be repeated later. Figure 3, the step S20 further includes steps S201-S204:

[0088] Step S201, performing perspective correction on the core image to obtain a corrected core image.

[0089] It should be noted that perspective correction refers to the geometric transformation of images to correct image deformation caused by shooting angle or lens distortion. Since the core may have irregular shapes and surface features, perspective deformation is easily generated when shooting. Therefore, the image perspective correction technology is used to perform secondary correction on the collected core images to correct the errors caused by the shooting angle.

[0090] It can be understood that the essence of perspective correction is to project a two-dimensional image onto a three-dimensional observation plane and then convert it back to two-dimensional coordinates. The perspective correction equation is as follows:

[0091]

[0092] Among them, (x, y, 1) is the two-dimensional pixel coordinate before perspective correction, M is the perspective correction matrix, and (x*, y*, z*) is the three-dimensional coordinate after perspective correction.

[0093] Step S202, denoising the corrected core image to obtain a target core image.

[0094] It should be noted that the collected core images may contain some noise, such as dust, scratches, etc., which will affect the image quality and recognition effect. Therefore, after perspective correction, the corrected core images need to be denoised to eliminate these noises and improve the image quality.

[0095] It can be understood that the denoising of the corrected core image may include preliminary denoising based on the Canny edge detection algorithm and secondary denoising based on morphology.

[0096] In a feasible implementation, step S202 may include: performing edge detection on the corrected core image to obtain image edge information and a core image after preliminary denoising; and performing secondary denoising on the core image after preliminary denoising according to the image edge information to obtain a target core image.

[0097] It should be noted that the image edge information refers to the pixel position and grayscale change of the edge of the object in the image. The edge detection of the corrected core image is performed, that is, the preliminary denoising of the corrected core image is performed based on the Canny edge detection algorithm, which can identify and retain the edge information in the image and remove some noise at the same time.

[0098] It is understandable that when the Canny edge detection algorithm is used to process the core image, the Canny operator combines steps such as Gaussian filtering, gradient calculation, non-maximum suppression and double thresholds to effectively extract image edge information, including: using Gaussian filters to achieve image denoising, reducing the impact of invalid edge information on edge extraction; finding the gradient of the image and extracting the edge of the denoised image based on the image pixel gradient. The Canny operator edge detection algorithm defines image edge extraction as calculating the position with the largest intensity of grayscale change in the image, while ignoring the influence of the color space pixel value. The Sobel operator is used to calculate the gradient of the image in the x and y directions, as follows:

[0099]

[0100] Among them, d x is the gradient of the image in the x direction, d y is the gradient of the image in the y direction, S is the gradient strength, and θ is the angle between the gradient direction of the image at a certain point and the positive x-axis.

[0101] After gradient calculation, the edge information of the image can be obtained, but in reality there is other texture information. The non-maximum suppression technique is applied to filter non-edge pixels, thereby eliminating the blurred boundaries and making the edges clear. In this process, the maximum value of the gradient on each pixel is retained and other values ​​are filtered out. The double threshold judgment criterion is applied to determine the possible boundaries. That is, an upper threshold and a lower threshold are set artificially. If the pixel in the image is greater than the upper threshold, it is considered to be a boundary. If it is less than the lower threshold, it is definitely not a boundary. Those between the two are alternatives (weak boundaries). A weak boundary connected to a certain pixel position and a strong boundary is considered to be a boundary, and the remaining weak boundaries are deleted. Finally, the edge information of the image and the core image after preliminary denoising are obtained.

[0102] It is worth noting that secondary denoising is performed on the core image after preliminary denoising based on the image edge information, and morphological operations can be used to further eliminate the remaining noise and improve the quality and clarity of the image. Secondary denoising can include morphological operations such as opening and closing operations, which can remove invalid edges while retaining the original target edge information to smooth the image and remove small noise points. After secondary denoising, the target core image obtained will have better visual effects and higher recognition accuracy.

[0103] Step S203: determining the core boundary according to the target core image.

[0104] It should be noted that the core boundary refers to the boundary between the core and the background in the core image. Determining the core boundary is one of the key steps in core image analysis, which helps to extract the characteristic information of the core and perform subsequent recognition, classification and other operations.

[0105] In a feasible implementation, step S203 may include: determining a standard straight line and a connected domain according to the target core image; extracting a straight line from the edge of the core column according to the connected domain to obtain an actual straight line; and merging the standard straight line and the actual straight line to obtain a core boundary.

[0106] It should be noted that, based on the principle of "each section of the core column is of equal width", the two straight lines at the top and bottom of the detection area are first detected, and the upper and lower edges are equally divided according to the core width, so that multiple groups of straight lines can be obtained, namely standard straight lines, and a core column area is between two adjacent standard straight lines.

[0107] It can be understood that by compiling an algorithm to calculate and obtain the connected domain, further extracting the edge straight line of the core column, that is, the actual straight line, and merging the standard straight line and the actual straight line, the core boundary can be obtained.

[0108] It is worth noting that a connected domain refers to an area in an image where pixel values ​​are continuous and adjacent, and is usually used to describe an object or target in an image. In a core image, a connected domain can represent the boundary or outline of a core column. In order to obtain a connected domain, image segmentation techniques can be used, such as a threshold-based segmentation method, an edge detection algorithm, or a region growing algorithm. These methods can divide an image into different regions based on the image's grayscale, color, texture, and other features, and extract the boundary of the target region.

[0109] It is worth noting that in the processing of core images, since the boundaries of core columns are usually clear, edge detection algorithms can be used to obtain connected domains. By detecting edge pixels in the image and connecting them into continuous lines, the boundary information of the core column can be obtained, and the boundary of the core column can be quickly located.

[0110] In a specific implementation, after determining the connected domain, the edge straight line of the core column can be further extracted. This usually involves fitting or approximating the connected domain to obtain a smoother and more accurate edge straight line. By selecting appropriate fitting algorithms and parameters, a straight line closer to the actual core boundary can be obtained, thereby improving the accuracy and reliability of core image analysis.

[0111] Step S204, segmenting and splicing the core image according to the core boundary to obtain a core image histogram.

[0112] In this embodiment, after the core boundary is determined, the core image is segmented and spliced ​​to form a core image histogram, so as to more intuitively display the structure and characteristics of the core, such as Figure 4 As shown, Figure 4The core image is a column chart. During the segmentation process, the core image can be divided into different regions according to the core boundary, and each region corresponds to a core column. After the division is completed, each region can be processed and analyzed separately to extract the characteristic information of the core column, such as color, texture, shape, etc.

[0113] It can be understood that the stitching step is to stitch the segmented core column images according to their original positions in the core to form a complete core image column chart, which helps to maintain the relative positions and relationships between the core columns, so that the column chart can more accurately reflect the overall structure and characteristics of the core. During the stitching process, it is necessary to keep the size and proportion of the core column image consistent to avoid deformation or distortion.

[0114] In this embodiment, the core image is standardized by performing perspective correction, denoising, segmentation and splicing on the core image to obtain a core image histogram, which can improve the quality and clarity of the image.

[0115] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the core image recognition method based on deep learning of the present application. More forms of simple transformations based on this technical concept are all within the scope of protection of the present application.

[0116] Based on the first embodiment of the present application, in the third embodiment of the present application, the same or similar contents as those in the first embodiment can be referred to the above description, and will not be described in detail later. Figure 5 , the step S30 further includes steps S301-S303:

[0117] Step S301, constructing an initial core recognition model based on the yoloV8 neural network.

[0118] It should be noted that the yoloV8 neural network is a real-time target detection algorithm, which has been widely used in the field of image processing due to its high efficiency and accuracy.

[0119] It can be understood that the yoloV8 neural network, as the basis for building the initial core recognition model, can improve the speed and accuracy of core image recognition.

[0120] In the specific implementation, in the process of building the initial core recognition model, it is necessary to first make necessary adjustments and optimizations to the yoloV8 neural network to meet the specific needs of core image recognition. This includes adjusting the network structure, optimizing parameter settings, etc., so as to better extract the feature information in the core image and accurately identify the type and location of the core.

[0121] In addition, in order to train the initial core recognition model, a large number of core image datasets need to be prepared. These datasets should contain core images of different types and features, and be accurately annotated so that the model can learn the accurate feature representation of the core.

[0122] Step S302: preprocessing is performed according to the core image histogram to obtain a data set.

[0123] It should be noted that after obtaining the core image histogram, in order to further improve the accuracy of core identification, the histogram needs to be preprocessed to generate a data set for training. The preprocessing may include operations such as image enhancement, size normalization, and data enhancement, which are not specifically limited in this embodiment.

[0124] It is understandable that image enhancement technology can enhance the feature information in the histogram, such as contrast enhancement, color correction, etc., to improve the quality and recognizability of the image. This helps the model to better extract the features of the core and improve the accuracy of recognition. Size normalization is to adjust the histograms of different sizes to a uniform size for the convenience of model processing and recognition. Through size normalization, the impact of image size differences on model performance can be eliminated and the generalization ability of the model can be improved. Data enhancement technology can generate more training samples by rotating, scaling, cropping and other operations on the histogram to increase the robustness and generalization ability of the model, which helps the model to maintain a high recognition accuracy when facing core images of different angles and scales.

[0125] It is worth noting that after preprocessing, the obtained data set will be used to train the initial core recognition model so that it can accurately identify the category and characteristics of the core. During the training process, the model will continue to learn and optimize to adapt to different types of core images and gradually improve the accuracy and stability of recognition.

[0126] In a feasible implementation, step S302 may include: cropping the core image histogram to obtain a cropped core image; annotating the lithology of the cropped core image to obtain an annotated core image; and determining a data set based on the annotated core image histogram.

[0127] It should be noted that the collected core image histograms are cropped to a uniform size, cropped into core images of 480×480 size, and the lithology of the core images is annotated to form a data set based on a neural network deep learning model.

[0128] It is understandable that by clipping the core image histogram, the redundant parts of the image can be removed, and only the part containing the core can be retained, thereby reducing the amount of calculation and improving the recognition accuracy. At the same time, the size of the clipped core image is unified, which is convenient for subsequent processing and analysis. In the clipping process, the appropriate clipping method and parameters can be selected according to the actual situation of the core and the recognition requirements to ensure that the quality of the clipped image meets the requirements.

[0129] It is worth noting that lithology annotation refers to marking and classifying the lithology of different areas in the image so that the model can learn the characteristics and classification information of the core. During the annotation process, it is necessary to accurately identify the areas of different lithology in the image and assign corresponding labels or categories to them. The accuracy and consistency of annotation are crucial for subsequent training and recognition.

[0130] In the specific implementation, when determining the data set, it is necessary to comprehensively consider factors such as the number, diversity, and annotation quality of the core images. By reasonably dividing the training set and the test set, it can be ensured that the model fully learns the characteristics of the core during the training process and verifies its recognition performance during the test process. At the same time, by optimizing and expanding the data set, the robustness and generalization ability of the model can be further improved.

[0131] It is worth noting that the core image dataset of the deep learning model should meet certain quantitative requirements, and the different lithology types that need to be identified should also meet certain quantitative requirements. In principle, the larger the dataset, the better the recognition effect of the model.

[0132] Step S303: training and verifying the initial core recognition model according to the data set to obtain a core recognition model.

[0133] It should be noted that in the process of training and verifying the initial core recognition model, the data set must first be divided into a training set and a validation set for model training and performance evaluation, respectively. During the training process, the model will learn to extract features from core images and classify and identify them based on the annotation information. Through continuous iteration and optimization, the model can gradually improve the accuracy and stability of recognition. During the verification process, the model will identify the core images in the verification set and compare them with the actual annotation results to evaluate the performance of the model. By calculating indicators such as accuracy and recall, we can fully understand the recognition effect of the model and make further adjustments and optimizations as needed.

[0134] It is worth noting that in order to ensure the reliability and stability of the model, some advanced training techniques and methods, such as regularization and optimization algorithms, can be used in the training process to reduce overfitting problems and improve the generalization ability of the model. At the same time, by monitoring and adjusting the training process, potential problems can be discovered and solved in a timely manner, thus ensuring that the model can meet the expected performance requirements.

[0135] In a feasible implementation manner, step S303 may include: dividing the data set according to a preset ratio to obtain a training set and a test set; training the initial core identification model through the training set to obtain a trained initial core identification model; verifying the trained initial core identification model through the test set to obtain a model accuracy; when the model accuracy reaches a preset accuracy threshold or the current number of training times reaches a preset training times threshold, using the trained initial core identification model as the core identification model; when the model accuracy reaches the preset accuracy threshold and the current number of training times does not reach the preset training times threshold, adjusting the parameters of the initial core identification model and executing the step of training the initial core identification model through the training set to obtain a trained initial core identification model.

[0136] It should be noted that the training set and the test set can be randomly divided in a 9:1 manner, and this embodiment does not impose any specific limitation on this.

[0137] It is understandable that the optimal number of iterations can be determined by presetting the training number threshold, i.e., the optimal number of iterations, and using the loss function, model accuracy, recall performance, and other evaluation models to determine the optimal number of iterations for model training. Among them, the loss function calculation for the core image recognition model based on the yoloV8 neural network includes classification and regression branches. The classification branch uses VFL-Loss, and the regression branch uses Distribution Focal Loss and CIOU Loss. The three loss functions are calculated according to certain weights.

[0138] In a specific implementation, the initial core image recognition model is trained using the training set. The hyperparameter settings of the training may be 100 iterations, 32 batches, 0.01 initial learning rate, 0.01 periodic learning rate, etc., which are not specifically limited in this embodiment.

[0139] It is worth noting that as the total number of epochs of deep learning model training increases for each test set, the accuracy and recall performance of the model will also change synchronously. As the number of epochs increases, the accuracy and recall of the model show a gradual upward trend, but when the number of epochs reaches a certain level, the accuracy and recall of the model are stable. Therefore, the optimal number of epochs for model training is determined based on the relationship between accuracy, recall and epochs.

[0140] It is worth noting that in this embodiment, binary cross entropy can be used as the object detection loss function. For the prediction results of the core image, the model calculates the overlap rate (I O U), and convert it into a probability value, i.e., the model accuracy, I O The value range of U is 0 to 1. The closer its value is to 1, the more accurate the target detection is. The accuracy threshold can be 0.9. If I O U>0.9, the core image recognition is considered accurate, and the overlap rate (I O U) is as follows:

[0141]

[0142] Among them, I O U is the overlap rate, that is, the model accuracy, S A is the marking box, S B For the real frame.

[0143] It is worth noting that known borehole core images other than the training set are used for image preprocessing to establish a borehole core training set, and the test set is used to test the recognition accuracy of the model; if the test accuracy meets the target requirement, the model test is considered successful; if it does not meet the target accuracy requirement, the model parameters need to be adjusted and the training needs to be repeated until success.

[0144] In this embodiment, the core image histogram is preprocessed to obtain a data set, and then the initial core recognition model is trained and verified based on the data set to obtain a core recognition model, which effectively improves the accuracy and generalization ability of the model prediction.

[0145] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the core image recognition method based on deep learning of the present application. More forms of simple transformations based on this technical concept are all within the scope of protection of the present application.

[0146] This application also provides a core image recognition device based on deep learning, please refer to Figure 6 , the core image recognition device based on deep learning includes:

[0147] The acquisition module 10 is used to acquire core images.

[0148] The processing module 20 is used to perform standardization processing on the core image to obtain a core image histogram.

[0149] The construction module 30 is used to construct a core recognition model based on the core image histogram.

[0150] The identification module 40 is used to identify the core image to be identified by using the core identification model to obtain an identification frame and a core type.

[0151] The present embodiment provides a core image recognition device based on deep learning. The present embodiment first collects core images, thereby accurately capturing the microstructure and features of the core and reducing the interference of human photography; the core images are standardized to obtain a core image bar graph, thereby improving image clarity; a core recognition model is constructed based on the core image bar graph, thereby effectively improving model accuracy; the core image to be recognized is recognized by the core recognition model to obtain an identification frame and a core type, thereby effectively improving the efficiency and accuracy of core image recognition.

[0152] In summary, this embodiment acquires core images and performs standardization processing to obtain a core image histogram, and then constructs a core recognition model based on the core image histogram to identify the core type of the core image to be identified, thereby overcoming the technical defects of errors in manual judgment of rock and soil geological categories, low recognition efficiency and accuracy, and can effectively improve the efficiency and accuracy of core image recognition.

[0153] Optionally, the processing module 20 is further used to perform perspective correction on the core image to obtain a corrected core image; perform denoising on the corrected core image to obtain a target core image; determine the core boundary based on the target core image; and segment and splice the core image according to the core boundary to obtain a core image bar graph.

[0154] Optionally, the processing module 20 is further used to perform edge detection on the corrected core image to obtain image edge information and a core image after preliminary denoising; and perform secondary denoising on the core image after preliminary denoising according to the image edge information to obtain a target core image.

[0155] Optionally, the processing module 20 is further used to determine a standard straight line and a connected domain according to the target core image; extract the edge straight line of the core column according to the connected domain to obtain an actual straight line; and merge the standard straight line and the actual straight line to obtain a core boundary.

[0156] Optionally, the construction module 30 is also used to construct an initial core recognition model based on the yoloV8 neural network; preprocess the core image histogram to obtain a data set; train and verify the initial core recognition model according to the data set to obtain a core recognition model.

[0157] Optionally, the construction module 30 is further used to crop the core image histogram to obtain a cropped core image; perform lithology annotation on the cropped core image to obtain a labeled core image; and determine a data set based on the labeled core image histogram.

[0158] Optionally, the construction module 30 is further used to divide the data set according to a preset ratio to obtain a training set and a test set; train the initial core identification model through the training set to obtain a trained initial core identification model; verify the trained initial core identification model through the test set to obtain a model accuracy; when the model accuracy reaches a preset accuracy threshold or the current training times reaches a preset training times threshold, use the trained initial core identification model as the core identification model; when the model accuracy reaches the preset accuracy threshold and the current training times does not reach the preset training times threshold, adjust the parameters of the initial core identification model and execute the step of training the initial core identification model through the training set to obtain the trained initial core identification model.

[0159] The core image recognition device based on deep learning provided by the present application adopts the core image recognition method based on deep learning in the above-mentioned embodiment, which can solve the technical problem of how to effectively improve the efficiency and accuracy of core image recognition. Compared with the prior art, the beneficial effects of the core image recognition device based on deep learning provided by the present application are the same as the beneficial effects of the core image recognition method based on deep learning provided by the above-mentioned embodiment, and the other technical features in the core image recognition device based on deep learning are the same as the features disclosed in the above-mentioned embodiment method, which will not be repeated here.

[0160] The present application provides a core image recognition device based on deep learning, and the core image recognition device based on deep learning includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the core image recognition method based on deep learning in the above-mentioned embodiment one.

[0161] Reference below Figure 7, which shows a schematic diagram of the structure of a core image recognition device based on deep learning suitable for implementing the embodiment of the present application. The core image recognition device based on deep learning in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 7 The deep learning-based core image recognition device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0162] like Figure 7 As shown, the core image recognition device based on deep learning may include a processing device 1001 (such as a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the core image recognition device based on deep learning are also stored. The processing device 1001, ROM1002 and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the core image recognition device based on deep learning to communicate with other devices wirelessly or by wire to exchange data. Although the core image recognition device based on deep learning with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have alternatively.

[0163] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0164] The core image recognition device based on deep learning provided by the present application adopts the core image recognition method based on deep learning in the above-mentioned embodiment, which can solve the technical problem of how to effectively improve the efficiency and accuracy of core image recognition. Compared with the prior art, the beneficial effects of the core image recognition device based on deep learning provided by the present application are the same as the beneficial effects of the core image recognition method based on deep learning provided by the above-mentioned embodiment, and the other technical features in the core image recognition device based on deep learning are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0165] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0166] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0167] The present application provides a computer-readable storage medium having computer-readable program instructions (i.e., computer programs) stored thereon, and the computer-readable program instructions are used to execute the deep learning-based core image recognition method in the above-mentioned embodiment.

[0168] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0169] The above-mentioned computer-readable storage medium may be included in the core image recognition device based on deep learning; or it may exist independently without being assembled into the core image recognition device based on deep learning.

[0170] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by a core image recognition device based on deep learning, the core image recognition device based on deep learning: collects core images; performs standardization processing on the core images to obtain a core image histogram; constructs a core recognition model based on the core image histogram; and recognizes the core image to be recognized through the core recognition model to obtain an identification frame and a core type.

[0171] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0172] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0173] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0174] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned core image recognition method based on deep learning, and can solve the technical problem of how to effectively improve the efficiency and accuracy of core image recognition. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the core image recognition method based on deep learning provided in the above-mentioned embodiment, and will not be repeated here.

[0175] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the deep learning-based core image recognition method as described above.

[0176] The computer program product provided in this application can solve the technical problem of how to effectively improve the efficiency and accuracy of core image recognition. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the core image recognition method based on deep learning provided in the above embodiment, which will not be repeated here.

[0177] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A core image recognition method based on deep learning, characterized in that: The method comprises: Acquisition of core images; Performing standardization processing on the core image to obtain a core image histogram; constructing a core recognition model based on the core image histogram; The core image to be identified is identified by the core identification model to obtain an identification frame and a core type.

2. The method according to claim 1, characterized in that The step of performing standardization processing on the core image to obtain a core image histogram includes: Performing perspective correction on the core image to obtain a corrected core image; De-noising the corrected core image to obtain a target core image; Determine the core boundary according to the target core image; The core image is segmented and spliced ​​according to the core boundary to obtain a core image histogram.

3. The method according to claim 2, characterized in that The step of denoising the corrected core image to obtain a target core image includes: Performing edge detection on the corrected core image to obtain image edge information and a preliminarily denoised core image; The core image after the preliminary denoising is subjected to secondary denoising according to the image edge information to obtain a target core image.

4. The method according to claim 2, characterized in that Determining the core boundary according to the target core image includes: Determine a standard straight line and a connected domain according to the target core image; Extract the edge straight line of the core column according to the connected domain to obtain the actual straight line; The standard straight line and the actual straight line are combined to obtain a core boundary.

5. The method according to claim 1, characterized in that The constructing of the core recognition model based on the core image histogram includes: Construct an initial core recognition model based on the yoloV8 neural network; Preprocessing is performed according to the core image histogram to obtain a data set; The initial core recognition model is trained and verified according to the data set to obtain a core recognition model.

6. The method according to claim 5, characterized in that The preprocessing is performed according to the core image histogram to obtain a data set, including: Clipping the core image histogram to obtain a clipped core image; Performing lithology labeling on the cropped core image to obtain a labeled core image; A data set is determined according to the annotated core image histogram.

7. The method according to claim 5, characterized in that The training and verification of the initial core identification model according to the data set to obtain the core identification model includes: Dividing the data set according to a preset ratio to obtain a training set and a test set; Training the initial core recognition model using the training set to obtain a trained initial core recognition model; Verifying the trained initial core recognition model through the test set to obtain the model accuracy; When the model accuracy reaches a preset accuracy threshold or the current training times reaches a preset training times threshold, the trained initial core recognition model is used as the core recognition model; When the model accuracy reaches a preset accuracy threshold and the current training times do not reach a preset training times threshold, the parameters of the initial core identification model are adjusted and the step of training the initial core identification model through the training set to obtain a trained initial core identification model is performed.

8. A core image recognition device based on deep learning, characterized in that: The core image recognition device based on deep learning includes: An acquisition module, used for acquiring core images; A processing module, used for performing standardization processing on the core image to obtain a core image histogram; A construction module, used for constructing a core recognition model based on the core image histogram; The identification module is used to identify the core image to be identified through the core identification model to obtain an identification frame and a core type.

9. A core image recognition device based on deep learning, characterized in that: The deep learning-based core image recognition device includes: a memory, a processor, and a deep learning-based core image recognition program stored in the memory and executable on the processor, wherein the deep learning-based core image recognition program is configured to implement the deep learning-based core image recognition method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a core image recognition program based on deep learning, and when the core image recognition program based on deep learning is executed by the processor, the core image recognition method based on deep learning as described in any one of claims 1 to 7 is implemented.

Citation Information

Patent Citations

  • Rock soil automatic recognition naming method and system based on deep learning

    CN110490100A

  • Intelligent rock core box, rock core image acquisition system and rock core intelligent logging method

    CN115753612A

  • Rock core histogram generation method and device, terminal equipment and storage medium

    CN116630351A