A drilling core digitizing logging method based on image recognition

Abstract

This invention provides a method for digital logging of borehole cores based on image recognition, relating to the field of geological exploration technology. It aims to address the problem in existing technologies that do not integrate and unify steps and related data beyond RQD (Real Quality Diagnosis). The technical solution includes the following steps: logging various data information of the borehole cores using a mobile device at the exploration site; importing the data collected by the mobile device to a computer, performing preliminary processing on the imported data, and generating corresponding core relationship charts using a cloud-based image recognition algorithm; training a deep learning image recognition algorithm in the cloud; and digitizing and electronically storing the logged data through electronic terminal devices and a cloud-based database management system, achieving centralized storage, classification, retrieval, and backup of the data, thereby improving the efficiency and reliability of information management.

Description

Technical Field

[0001] This invention relates to the field of geological exploration technology, specifically to a method for digital logging of borehole cores based on image recognition. Background Technology

[0002] Borehole core logging is the process of systematically recording and describing the core samples obtained from boreholes. Borehole core logging generally includes the following steps: 1. Preparation: Determine the necessary tools and materials for logging, such as logging forms, core boxes, measuring tools, rulers, magnifying glasses, etc. 2. Core Processing: Number the retrieved core samples according to borehole depth and perform preliminary processing. This includes removing soil, washing the core surface to remove drilling fluid, and marking the start and end points of each core run. 3. Core Photography: Take photographs of each core box using a camera, and take close-up shots of specific parts or features. 4. Logging Measurement: Use measuring tools, such as rulers or digital measuring instruments, to measure the length and diameter of the core samples. Record the measurement results for each core segment. 5. Observation and Description: Observe and describe the details of the core samples using a magnifying glass. This includes the rock's color, structure, texture, mineral content, and other characteristics. Also, record any special structures, fractures, joints, or core variations. 6. Sampling: Select appropriate locations for sampling according to needs and sampling purpose, and record detailed sampling information. 7. Logging Table Completion: Fill in the logging tables with measurement results, observations, and descriptions. The tables typically include fields for borehole depth, per pass, diameter, recovery rate, yield rate, RQD, and observations and descriptions. 8. Logging Data Processing: Organize and summarize all logging data and digitize it for subsequent analysis and application.

[0003] Current technologies suffer from the following problems: 1. Borehole logging primarily relies on manual operation and paper-based records. Firstly, the recorded information requires subsequent organization and archiving, making retrieval difficult, easily leading to information management chaos, and posing a risk of data loss. Secondly, paper records are susceptible to human error, including illegible handwriting and readability issues, threatening the accuracy and reliability of the data. Finally, the transmission and sharing of paper records depend on physical media, resulting in difficulties in collaboration, communication delays, and low work efficiency. 2. Current core logging methods require extensive manual operation and mechanical repetition, leading to a massive workload and low efficiency in field measurements. Secondly, the use of steel tape measures or measuring tapes for repeated measurements results in low measurement accuracy, easily affected by the precision of the measuring tools themselves and subjective human factors. This method cannot provide accurate data reflecting rock quality and cannot meet the needs of precise assessment of engineering rock properties. Existing technologies such as CN113806681A, CN112241711A, and CN114387328A rely on image recognition and other methods to automate the RQD processing during the logging process, but neglect the fact that borehole logging is a systematic project that integrates multiple process steps, and does not integrate and unify other steps and related data besides RQD.

[0004] The prior art CN202210682472.4 discloses an automatic RQD logging method, system, and device for drilling cores. This patent relies on single RQD image recognition and data processing, and cannot achieve full-process digital logging through multi-platform collaboration. The patent only counts cores with a length of 10cm (including 10cm), which has certain shortcomings. In addition, the measurement basis and calculation method of RQD and other data in the prior art do not strictly follow the methods recommended by the International Society for Rock Mechanics (ISRM) and the Standardization Committee for Laboratory and Field Testing, which leads to an excessive underestimation of the quality of the rock mass. Summary of the Invention

[0005] In view of the problems existing in the prior art, this invention discloses a method for digital logging of borehole cores based on image recognition, encompassing all aspects of core logging. It comprehensively utilizes mobile, desktop, and cloud technologies, achieving full collaboration and efficient completion of borehole core logging through multi-platform cooperation. Firstly, leveraging the convenience and mobility of mobile devices, geological engineers can acquire core images and complete some logging tasks. Secondly, the larger screen space provided by computers allows for more complex task processing and data preprocessing. Finally, a cloud-trained image recognition model and more powerful computing capabilities are used to complete data analysis and processing of all core cycles, while also serving as a data storage and sharing center, ensuring data security and accessibility.

[0006] The technical solution adopted includes the following steps:

[0007] Step 1: At the exploration and construction site, use a mobile device to record various data information from the borehole core.

[0008] Step 2: Import the data collected on the mobile device to the computer, perform preliminary processing on the imported data, and use cloud-based image recognition algorithms to generate corresponding core relationship charts.

[0009] Step 3: Train the deep learning image recognition algorithm in the cloud; calculate the core relationship data for each round according to the corresponding formula based on different shape contours and their combination relationships, and output the core image recognition results and core relationship data to the computer. This invention improves information management and storage optimization: through electronic terminal equipment and cloud database management system, the cataloged data is digitized and electronic, realizing centralized storage, classification, retrieval and backup of data, and improving the efficiency and reliability of information management.

[0010] As a preferred embodiment of the present invention, step 1 further includes the following steps:

[0011] Step 1.1: Clean the surface of the core sample with water;

[0012] Step 1.2: Insert a core tag with identification function at the depth of each termination hole;

[0013] Step 1.3: Enter the project information for the rock core;

[0014] Step 1.4: Input the borehole information for the rock core;

[0015] Step 1.5: Take photos of the rock core and perform preliminary processing on the photos;

[0016] Step 1.6: Segment the rock core according to its material composition and degree of weathering;

[0017] Step 1.7: Record the fracture development of each segment of the core.

[0018] Step 1.8: Select representative soil and rock samples for sampling;

[0019] Step 1.9: Export the compiled information to the computer in the form of a compressed file.

[0020] As a preferred embodiment of the present invention, step 2 further includes the following steps:

[0021] Step 2.1: Perform secondary processing on the core images;

[0022] Step 2.2: Add titles to the core images after secondary processing and generate a title bar at the top;

[0023] Step 2.3: Import drilling cycle information;

[0024] Step 2.4: Call the cloud-trained recognition algorithm to perform core identification processing on the current borehole;

[0025] Step 2.5: Manually correct any erroneous information.

[0026] Step 2.6: Generate the corresponding core relationship chart data and store and back up the data.

[0027] As a preferred embodiment of the present invention, step 3 further includes the following steps:

[0028] Step 3.1: Conduct recognition training on the core plate in the cloud to ensure that the core plate can be accurately recognized, and then use OCR text recognition to recognize the text information on the core plate to ensure that the information is correct;

[0029] Step 3.2: Conduct recognition training on the core box in the cloud to ensure accurate segmentation and recognition of the core box and the core. Then, establish a planar coordinate network based on the length and width dimensions of the core box to lay the foundation for core length recognition. Step 3.3: Conduct recognition training on cores with different shapes and contours in the cloud. Then, accurately divide the contours identified by the algorithm and calculate the sampling rate, acquisition rate, and RQD for each round using different formulas. Step 3.4: Output the core recognition results and the calculated core relationship table for each round to the computer. Color-code the contours of each core according to its category and annotate the information within the core contours.

[0030] As a preferred embodiment of the present invention, step 3.3 includes rectangular contours, triangular contours, split contours, and irregular contours.

[0031] The mathematical expression for calculating the rectangular outline is:

[0032] Adoption rate = Ln / Lm

[0033] Acquisition rate = Ln / Lm

[0034]

[0035] The mathematical expression for calculating a triangular-like contour is:

[0036] Adoption rate = (Ln1 + Ln2) / Lm

[0037] Acquisition rate = (Ln1 + Ln2) / Lm

[0038]

[0039] The mathematical expression for calculating the outline of a split shape is:

[0040] Adoption rate = (Ln1 + Ln2) / (2 * Lm)

[0041] Acquisition rate = (Ln1 + Ln2) / (2 * Lm)

[0042]

[0043] The mathematical expression for an irregular contour is:

[0044] Adoption rate = (Ln*a) / Lm

[0045] Acquisition rate = 0

[0046] RQD = 0

[0047] m represents the cycle number; Lm represents the drilling length of the m-th cycle; Ln, Ln1, and Ln2 are the lengths of a single core sample; a represents the reduction factor.

[0048] The beneficial effects of this invention are as follows: This invention improves information management and storage optimization: By using electronic terminal equipment and cloud database management system, the cataloged data is digitized and electronic, realizing centralized storage, classification, retrieval and backup of data, thereby improving the efficiency and reliability of information management.

[0049] Furthermore, this invention improves data quality and accuracy: digital records eliminate handwriting and readability issues, reduce human error, and ensure data accuracy and reliability.

[0050] Furthermore, this invention facilitates data sharing and collaboration: digital records can be instantly shared and collaborated on via networks and cloud platforms, promoting real-time communication, cooperation, and decision-making among team members, and improving work efficiency and teamwork capabilities.

[0051] Furthermore, this invention can reduce costs and increase efficiency: it solves the problem of manually and mechanically measuring the length of rock cores on site, and automatically identifies and processes the length of rock cores through rock core image recognition technology, realizing the automated calculation of rock core relationship-related data, improving work efficiency and reducing labor costs.

[0052] Furthermore, this invention enables refined identification and calculation: by identifying the rock core contour, various types of rock cores are subdivided, and different formulas are used to calculate various data of the rock core for different situations, thereby improving the accuracy of the data and providing accurate basis for rock mass integrity evaluation, engineering geological classification of dam foundation rock mass, classification of tunnel surrounding rock, and classification of slope rock mass structure. Attached Figure Description

[0053] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0054] Figure 1 This is a schematic diagram of the various modules for drilling and recording on the mobile device of the present invention. Figure 1 ;

[0055] Figure 2 This is a schematic diagram of the various modules for drilling and recording on the mobile device of the present invention. Figure 2 ;

[0056] Figure 3 This is a schematic diagram of the various modules for drilling and recording on the mobile device of the present invention. Figure 3 ;

[0057] Figure 4 This is a schematic diagram of the various modules for drilling and recording on the mobile device of the present invention. Figure 4 ;

[0058] Figure 5 This is a schematic diagram of the various modules for drilling and recording on the mobile device of the present invention. Figure 5 ;

[0059] Figure 6 This is a schematic diagram of the various modules for drilling and recording on the mobile device of the present invention. Figure 6 ;

[0060] Figure 7 This invention relates to the computer-based drilling logging operation interface and recognition results. Figure 1 ;

[0061] Figure 8 This invention relates to the computer-based drilling logging operation interface and recognition results. Figure 2 ;

[0062] Figure 9 The rock core label of this invention and its identification effect Figure 1 ;

[0063] Figure 10 The rock core label of this invention and its identification effect Figure 2 ;

[0064] Figure 11 This invention provides information on core contour image recognition. Figure 1 ;

[0065] Figure 12 This invention provides information on core contour image recognition. Figure 2 ;

[0066] Figure 13This invention provides information on core contour image recognition. Figure 3 ;

[0067] Figure 14 This invention provides information on core contour image recognition. Figure 4 ;

[0068] Figure 15 This invention provides information on core contour image recognition. Figure 5 ;

[0069] Figure 16 This invention provides information on core contour image recognition. Figure 6 ;

[0070] Figure 17 This invention provides information on core contour image recognition. Figure 7 ;

[0071] Figure 18 This invention provides information on core contour image recognition. Figure 8 ;

[0072] Figure 19 This invention provides information on core contour image recognition. Figure 9 ;

[0073] Figure 20 This invention provides information on core contour image recognition. Figure 10 ;

[0074] Figure 21 This invention provides information on core contour image recognition. Figure 10 one;

[0075] Figure 22 This invention provides information on core contour image recognition. Figure 10 two;

[0076] Figure 23 This invention provides information on core contour image recognition. Figure 10 three;

[0077] Figure 24 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0078] Example 1

[0079] like Figures 1 to 24 As shown, this invention discloses a method for digital logging of borehole cores based on image recognition. The technical solution adopted includes the following steps:

[0080] Step 1: At the exploration site, use a mobile device to record the borehole core data. This includes entering information such as the project details and exploration borehole information, taking core photos, describing the core, describing fractures, and providing sampling information. See the attached document for the operation interface of each module and related borehole recording information. Figures 1 to 6 After the on-site data recording is completed, export all the data collected on the mobile device to the computer.

[0081] Step 2: Import the collected data onto the computer. First, rotate and correct some borehole photos to ensure that each core box photo is a top-down, vertically captured image. Edit the project name, exploration stage, borehole name, borehole depth range, and core box number into image titles to generate titled images. Next, use a cloud-based image recognition algorithm to identify the core outline and length, and finally generate a corresponding core relationship diagram. The corresponding computer interface and recognition results are attached. Figure 7 , Figure 8 .

[0082] Step 3 involves training a deep learning image recognition algorithm in the cloud, including core tags, core boxes, and core outlines. Based on the core image data uploaded to the computer, the recognized core tags are divided into different rounds. A planar coordinate network is established based on the length and width of the core boxes to calculate the length of a single core and the advance distance per round. The core outlines are further subdivided and marked with different colors, classifying rocks and overburden layers according to their material composition. Rock outlines can be further subdivided into rectangular, parallelogram-like, trapezoidal, triangular, and irregular gravel shapes. Based on the different shapes and their combinations, the core relationship data for each round is calculated using corresponding formulas. Finally, the core image recognition results and core relationship data are output to the computer.

[0083] Step 1 also includes the following steps:

[0084] Step 1.1, Clean the core: Wash the core with water to remove dirt, drilling fluid and other impurities from the surface of the core, ensuring that the core is clean, aesthetically pleasing and easy to photograph.

[0085] Step 1.2, Inserting the core plate: Insert a specially made core plate at the depth of the final hole for each cycle. The core plate has cycle information and hole depth information for easy identification and reading. The color of the core plate is different from that of ordinary cores for easy and accurate identification.

[0086] Step 1.3, Project Information Entry: Create a project on the mobile device, enter the project name, industry, and exploration stage, and import the corresponding EXCEL spreadsheet for strata and lithology.

[0087] Step 1.4, Drilling Information Entry: Create a new borehole within the project and enter information such as borehole number, borehole inclination, borehole direction, borehole depth, drilling unit, logger, and on-site checker.

[0088] Step 1.5, Core photography: Take photos of all cores from directly above each core box, and perform preliminary cropping and correction on the core photos.

[0089] Step 1.6, Core Description: The logging boreholes are segmented according to material composition and weathering degree. Information such as bottom depth, stratigraphic code, lithology name, hardness, structure, and texture of each segment is recorded. Special adverse geological phenomena such as faults, karst, weak interlayers, and interlayer shear zones can be recorded in detail. Finally, the core sampling status is summarized.

[0090] Step 1.7, Fracture Description: Record the fracture development of each segment, including the number of fractures, the depth of the developed pores, the fracture dip angle, the straightness, the roughness, the width, and the characteristics of the filling material, and automatically calculate the fracture density of that segment.

[0091] Step 1.8, Sampling Record: Select representative soil and rock masses for sampling, record the sampling number, type, field name, and take corresponding photos.

[0092] Step 1.9, Data Export: Export all the compiled information to the computer in the form of a compressed file for the corresponding operations.

[0093] Step 2 also includes the following steps:

[0094] Step 2.1, Image Correction: Import the data recorded on the mobile device and perform secondary correction on the borehole core photos. After clicking "Correct Image", select the four corner points of the image in sequence (the selected points will be red). After selecting the four points, the image will be automatically cropped to form a new image.

[0095] Step 2.2, Add image title: Click the "Custom Image Title" function button in the upper right corner, select the text information to be displayed in the title bar in the pop-up window, obtain the corresponding title content, and automatically generate the title bar at the top of the borehole photo.

[0096] Step 2.3, Import Recursion Information: Click the "Import Recursion" button, select the recursion information file for the current borehole, and import it.

[0097] Step 2.4, Image Recognition: After clicking the "Recognize" button, the cloud-trained recognition algorithm will be invoked to perform core identification on the current borehole. During the recognition process, the intelligent recognition progress bar will change as the recognition progresses. The recognition process ends when the intelligent recognition progress bar reaches 100%.

[0098] Step 2.5, Manual Error Correction: After recognition is completed, recognition errors caused by the clarity of the photo can be corrected manually to ensure the accuracy of the information.

[0099] Step 2.6, Data Storage and Backup: Generate the corresponding core relationship diagram and store all electronic data of the borehole (bent borehole core photos, core descriptions, fracture descriptions, sampling information, etc.) in the cloud project folder for easy access and retrieval later.

[0100] Step 3 also includes the following steps:

[0101] Step 3.1 involves cloud-based training on specially designed core tags to ensure accurate recognition regardless of their location or arrangement within the core box. A consistent color coding system is then implemented to pinpoint the drilling position for each run. Finally, the hole depth and drilling progress for each run are determined using OCR text recognition of the core tag information or an externally imported core run table, ensuring a one-to-one correspondence between the core tags and the data. See the attached image for details on the core tags and their recognition performance. Figure 9 , Figure 10 .

[0102] Step 3.2 involves training the core boxes for identification in the cloud, including empty core boxes of various colors and materials with different numbers of layers, as well as core boxes filled with cores of various colors and patterns, to ensure accurate segmentation and identification of the core boxes and cores. Then, a planar coordinate network is established based on the length and width dimensions of the core boxes (generally with the lower left corner of the core box as the origin), laying the foundation for core length identification.

[0103] Step 3.3 involves training the algorithm to identify cores with different shapes and contours in the cloud. The identified contours are then accurately categorized into rectangular, parallelogram-like, trapezoidal, triangular, and irregular gravel and overburden layers. Following the methods recommended by the International Society for Rock Mechanics (ISRM) and the Standardization Committee for Laboratory and Field Testing, the length of a single core should be measured along its centerline. Furthermore, cores that can be assembled into a columnar shape and are greater than 10 cm under certain conditions should also be included in the RQD (Recovery Rate) statistics, rather than the common practice in China of only including columnar cores larger than 10 cm in each advance. The length of the centerline of each core contour is accurately read from the plane coordinate network as the core length. Different formulas are used below to calculate the recovery rate, yield rate, and RQD for each advance, depending on various practical situations.

[0104] Case 1: For cases involving partially complete columns such as rectangles, trapezoids, and parallelograms, as well as cases where these three shapes are joined together horizontally, see Appendix. Figures 11 to 16 Recovery rate, yield, and RQD are all calculated independently based on the centerline length of a single core sample. The mathematical formulas can be uniformly expressed as follows:

[0105] Adoption rate = Ln / Lm

[0106] Acquisition rate = Ln / Lm

[0107]

[0108] Case 2: When a triangular-like shape is combined with other triangular-like, trapezoidal-like, or parallelogram-like shapes to form a complete or more complete columnar shape, see Appendix. Figure 17 , Figure 18 The two core samples should be treated as a single unit, and their recovery rate, yield, and recovery quantity (RQD) should be determined by the sum of their centerline lengths. The mathematical formula can be uniformly expressed as:

[0109] Adoption rate = (Ln1 + Ln2) / Lm

[0110] Acquisition rate = (Ln1 + Ln2) / Lm

[0111]

[0112] Scenario 3: When the core sample is split into two parts by a near-vertical fracture, see attached... Figure 19 These two semi-columnar cores should be treated as a whole, and their recovery rate, yield, and RQD should be considered based on the average of their centerline lengths.

[0113] Adoption rate = (Ln1 + Ln2) / (2 * Lm)

[0114] Acquisition rate = (Ln1 + Ln2) / (2 * Lm)

[0115]

[0116] Case 4: For irregularly shaped gravel that cannot be formed into a column, and for overburden layers with mixed compositions (such as silty clay, sand, gravelly soil, gravel, sand and gravel pebbles, etc.), since only the extraction rate is considered, it can be simplified into a rectangular shape and its length reduced. See appendix. Figures 20 to 23 .

[0117] Adoption rate = (Ln*a) / Lm

[0118] Acquisition rate = 0

[0119] RQD = 0

[0120] Where m represents the run number; Lm represents the drilling length of the m-th run; Ln, Ln1, and Ln2 are the lengths of a single core; and a represents the reduction factor. These four scenarios basically cover all situations in the actual exploration and production process. All the recovery rates, yield rates, and RQD mentioned above are only independent data for a single block (or two blocks combined). To calculate the recovery rate, yield rate, and RQD of a single run, simply sum up all the cases (independent recovery rate, independent yield rate, independent RQD) that occurred in that run.

[0121] Step 3.4: Output the core identification results and the calculated core relationship table for each round to the computer. Color-code the outline of each core according to its category, and mark the round, number of cores, length and other information within the core outline.

[0122] There are many tools for implementing image recognition algorithms in this technical solution, such as CNN, OpenCV, TensorFlow, PyTorch, etc.

[0123] While the specific embodiments of the present invention have been described in detail above, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention, and modifications or variations without creative effort are still within the protection scope of the present invention.

Claims

1. A method for digital logging of borehole cores based on image recognition, characterized in that, Includes the following steps: Step 1: At the exploration and construction site, use a mobile device to record various data information from the borehole core. Step 2: Import the data collected on the mobile device to the computer, perform preliminary processing on the imported data, and use cloud-based image recognition algorithms to generate corresponding core relationship charts. Step 3: Train the deep learning image recognition algorithm in the cloud; calculate the core relationship data for each round according to the corresponding formula based on different shape contours and their combination relationships, and output the core image recognition results and core relationship data to the computer. Step 1 also includes the following steps: Step 1.1: Clean the surface of the core sample with water; Step 1.2: Insert a core tag with identification function at the depth of each termination hole; Step 1.3: Enter the project information for the rock core; Step 1.4: Input the borehole information for the rock core; Step 1.5: Take photos of the rock core and perform preliminary processing on the photos; Step 1.6: Segment the rock core according to its material composition and degree of weathering; Step 1.7: Record the fracture development of each segment of the core. Step 1.8: Select representative soil and rock samples for sampling; Step 1.9: Export the compiled information to the computer as a compressed file; Step 3 also includes the following steps: Step 3.1: Conduct recognition training on the core plate in the cloud to ensure that the core plate can be accurately recognized, and then use OCR text recognition to recognize the text information on the core plate to ensure that the information is correct; Step 3.2: Conduct recognition training on the core box in the cloud to ensure that the core box and the core can be accurately separated and identified. Then, establish a planar coordinate network based on the length and width of the core box to lay the foundation for core length recognition. Step 3.3: Conduct identification training on rock cores with different shapes and contours in the cloud, then accurately divide the contours identified by the algorithm, and use different formulas to calculate the sampling rate, acquisition rate and RQD for each round; Step 3.4: Output the core identification results and the calculated core relationship table for each round to the computer, color-code the outline of each core according to its category, and annotate the information within the core outline; Step 3.3 includes rectangular contours, triangular contours, split contours, and irregular contours. The mathematical expression for the rectangular contour is: Utilization rate = Ln / Lm; Acquisition rate = Ln / Lm; ； The mathematical expression for calculating a triangular-like contour is: Adoption rate = (Ln1 + Ln2) / Lm; Acquisition rate = (Ln1 + Ln2) / Lm; ； The mathematical expression for calculating the outline of a split shape is: Adoption rate = (Ln1 + Ln2) / (2 * Lm); Acquisition rate = (Ln1 + Ln2) / (2 * Lm); ； The mathematical expression for an irregular contour is: Adoption rate = (Ln*a) / Lm; Acquisition rate = 0; RQD=0; m represents the cycle number; Lm represents the drilling length of the m-th cycle; Ln, Ln1, and Ln2 are the lengths of a single core sample; a represents the reduction factor.

2. The method for digital logging of borehole cores based on image recognition according to claim 1, characterized in that, Step 2 also includes the following steps: Step 2.1: Perform secondary processing on the core images; Step 2.2: Add titles to the core images after secondary processing and generate a title bar at the top; Step 2.3: Import drilling cycle information; Step 2.4: Call the cloud-trained recognition algorithm to perform core identification processing on the current borehole; Step 2.5: Manually correct any erroneous information. Step 2.6: Generate the corresponding core relationship chart data and store and back up the data.

Images