A core image achievement automatic output method based on block morphology parameter classification

Abstract

The present application belongs to the field of engineering investigation and geological engineering informatization, and discloses a core image achievement automatic output method based on block shape parameter classification, which solves the problems that the existing core image processing scheme achievement output cannot be traced back, and cannot realize the automatic organization and output of the achievement based on block shape classification. The present application takes the original core box photo bound with metadata as the starting input, successively carries out outer frame detection and positioning, core box cutting and standardization processing, completes image preprocessing, then carries out core block segmentation and result image generation, and then carries out quality control correction and effective version solidification on the segmentation result. Subsequently, the core block contour features are extracted based on the solidification result, and the shape classification is carried out. At the same time, the end point of the block is taken, and the local straight line fitting is carried out, and the fracture dip angle calculation is completed. Then, according to the shape classification and dip angle calculation result, the target block is screened, and finally the automatic output of the core processing achievement is completed according to the preset template.

Description

Technical Field

[0001] This invention belongs to the field of engineering survey and geological engineering informatization, and specifically relates to a method for automatically outputting core image results based on block morphology parameter classification. Background Technology

[0002] In water conservancy and hydropower, mining exploration, and civil engineering surveying, borehole cores are core samples reflecting the structural characteristics of underground rock masses. Parameters such as the morphology, fragmentation characteristics, and fracture dip angle of the core blocks are important bases for rock mass quality evaluation and engineering design. Traditional core logging work relies on manual interpretation and statistical analysis of the block length, morphology, and fragmentation characteristics in core photographs, followed by manual compilation into tables and maps. When faced with large-scale core photograph samples or collaborative logging scenarios involving multiple personnel, this approach generally suffers from low work efficiency, significant subjective differences in interpretation results, and high costs for result verification.

[0003] With the development of engineering information technology, end-to-end processing technology for core images has gradually advanced. Existing technologies can now identify and crop effective areas of core boxes, segment core blocks semantically / instancely, and perform random checks on segmentation results through overlay preview, forming a closed-loop process of "identification-correction-recalculation." Meanwhile, engineering applications place higher demands on the organization of core results, requiring the classification and summarization of results using "block morphology type" as an index to achieve rapid screening and location display of target blocks, and to calculate and analyze quantitative parameters such as block fracture dip angles to improve the relevance of cataloging and the efficiency of verification.

[0004] Existing core image recognition and logging solutions mostly follow a "recognition-statistics-output" chain. Their output content is mostly limited to segmentation results or general logging data, making it difficult to directly generate deliverable packages that conform to engineering archiving standards. In actual project delivery, manual secondary screening, summarization, and mapping according to project templates are still required, which easily leads to inconsistent results and information omissions. In addition, due to the influence of on-site shooting quality and the complexity of core arrangement, missegmentation and omission of segmentation during core block segmentation are objectively present. Existing solutions lack effective version management and consistency constraint mechanisms for segmentation results after manual quality control correction. It often results in situations where "correction has been completed, but subsequent statistics / export still use the old results," leading to a lack of traceability of results and further increasing the verification cost.

[0005] While some existing technologies have introduced attention mechanisms and deep networks with Transformer backbones to improve segmentation accuracy, or integrated multi-source information to achieve comprehensive judgment, none of them have solved the core problems of insufficient standardization of output results and lack of version management of correction results. They cannot meet the actual needs of engineering exploration for verifiable, traceable, efficient and standardized core logging results. Summary of the Invention

[0006] The technical problem to be solved by this invention is to provide an automatic output method for core image results based on block morphology parameter classification. This method addresses the problems of existing core image processing solutions, such as the inability to directly meet engineering archiving requirements, the lack of version management after manual quality control correction leading to untraceable results, and the inability to automate the organization and output of results based on block morphology classification, resulting in low cataloging efficiency, inconsistent results standards, and high verification costs.

[0007] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0008] An automatic output method for core image results based on block morphology parameter classification includes the following steps:

[0009] S1. Obtain core box photos and bind metadata such as borehole number and depth range to each photo to form a core box photo set with metadata association;

[0010] S2. Perform outer frame positioning, effective area cropping and standardization processing on the core box photos to obtain a set of standardized cropped effective area images of the core box;

[0011] S3. Perform core block identification, segmentation, and block numbering on the effective area image of the core box, and generate a block label image and a preview overlay image;

[0012] S4. Check and correct the segmentation results of the block label image, and solidify the corrected segmentation results as valid results;

[0013] S5. Based on the solidified segmentation results, extract the boundary contours of each core block, calculate the morphological characteristic parameters of the core block, and form a record of the block morphological characteristic parameters.

[0014] S6. Based on the morphological characteristic parameters of the core blocks, the rock core blocks are classified into different shapes. The fracture dip angle is calculated by fitting the local straight line of the end contour of the rock core block and the dip angle is graded. The morphological category, fracture dip angle value and dip angle grading result of each rock core block are obtained.

[0015] S7. Associate the morphological category, fracture dip angle value, and dip angle classification results of the core blocks with the metadata of borehole number and depth range to generate a result index and form a target block set according to preset screening conditions;

[0016] S8. Based on the results index and target block set, generate a target block list, a classification summary table, and a block location preview image;

[0017] S9. Collect the target block list, classification summary table, and block location preview image, generate result traceability record and process processing log, and package and export them according to the preset directory structure to form a result package that can be directly archived.

[0018] This solution constructs a closed-loop processing method for the entire process from core image acquisition to output of results packages, achieving standardization of core image processing and automation of results output. At the same time, the segmentation result version solidification mechanism ensures that subsequent processing is based on the latest valid results, solving the core problems of untraceable results and the need for manual secondary processing in existing technologies. With block morphology classification and fracture dip angle grading as the core of results organization, it achieves accurate indexing and rapid screening of results, greatly improving the efficiency of core logging and the consistency of results.

[0019] Furthermore, in step S1, when binding metadata to the core box photos, the box sequence or return number information is also bound, and a consistency check is performed on the core box photos that have completed metadata binding. Samples with missing metadata, unreasonable metadata format, or duplicate depth ranges in the same borehole and indistinguishable box sequence are marked as abnormal samples and prompt information is output. Abnormal samples are processed in subsequent steps after the information is completed.

[0020] In this scheme, the metadata dimension is further refined by supplementing the box sequence / return number, so as to achieve the unique and accurate location of the core sample; unqualified inputs are filtered in advance by consistency check and anomaly marking, so as to ensure the validity of the data from the source and reduce the amount of invalid work in subsequent processing.

[0021] Furthermore, in step S2, the standardization processing of the core box photos includes: unifying the image resolution or short side size, correcting the long side direction of the core box, light denoising, or enhancing the core boundary, and the standardization processing does not change the relative structural relationship of the core blocks inside the core box; when the outer frame positioning fails or the positioning result is abnormal, the default center cropping or fixed ratio cropping is used to generate candidate cropping images, or the corresponding sample is marked as an abnormal sample and a prompt message is output.

[0022] In this scheme, operations such as unified resolution and orientation correction are used to ensure that core images captured under different conditions have the same processing benchmark, thereby improving the adaptability of the algorithm. A fallback strategy is provided for failed bounding box detection to offer alternative solutions for sample processing, while anomaly marking facilitates subsequent manual remediation, thus balancing process automation and fault tolerance.

[0023] Furthermore, in step S4, the correction methods for the segmentation results include one or more of the following: supplementing the annotation of the missing core block areas, erasing and removing the incorrectly segmented background areas, segmenting the adhered core blocks, merging the same core blocks that have been over-segmented, and adjusting the boundaries of core blocks with offset boundaries; when solidifying the corrected segmentation results, the original segmentation results are replaced by overwriting or updating, and the information of the corrected sample, the correction type, and the correction time are recorded.

[0024] In this solution, corresponding correction methods are designed for five common problems such as missed segmentation and mis-segmentation to achieve precise optimization of the segmentation results; the covering / updating replacement method avoids the storage chaos of multi-version results, and the correction records make every result adjustment traceable, meeting the quality control requirements of engineering surveys.

[0025] Further, in step S5, the morphological characteristic parameters of the core block include: one or more of the parameters reflecting the length of the core block, the parameters reflecting the width of the core block, the length-width ratio reflecting the length-width ratio relationship of the core block, and the parameters reflecting the degree of regularity of the core block shape.

[0026] In this solution, the block morphology is characterized from different dimensions based on four types of parameters: length, width, ratio, and degree of regularity, avoiding the classification limitations of a single parameter.

[0027] Further, in step S6, the core blocks are morphologically classified based on the block morphological characteristic parameters, including:

[0028] The minimum bounding rectangle is obtained based on the boundary contour of the core block, and the length L and width W of the core block are obtained from the minimum bounding rectangle and satisfy L≥W>0. The length-width ratio R = L / W is calculated, and the core blocks are morphologically classified according to the threshold interval of the length-width ratio. At the same time, an area threshold Amin is set, and the core blocks with an area A < Amin are classified as fragmented or marked as invalid blocks.

[0029] In this solution, the accurate extraction of the block length and width is achieved through the minimum bounding rectangle. The threshold classification based on the length-width ratio enables the morphological determination to have a clear quantitative standard, avoiding subjective determination errors; the setting of the area threshold effectively filters the interference of noise fragments on the classification results, further improving the accuracy and effectiveness of the morphological classification.

[0030] Further, the morphological classification of the core blocks according to the threshold interval of the length-width ratio includes:

[0031] When R≥6, the core block is a long columnar shape;

[0032] When 3≤R<6, the core block is a columnar shape;

[0033] When 1.8≤R<3, the core block is a short columnar shape;

[0034] When 1.2≤R<1.8, the core block is a semi-columnar / cake shape;

[0035] When 1≤R<1.2, the core block is a fragmented shape.

[0036] This solution provides specific threshold ranges for aspect ratio, enabling block shape classification to have a feasible execution standard, ensuring consistency of classification results for different operators and in different project scenarios, and improving the practicality and scalability of the method.

[0037] Furthermore, in step S6, the fracture dip angle is calculated and dip angle classification is completed by fitting the local straight line of the core block end profile, including:

[0038] First, obtain the point set of the outer contour of the core block and determine the reference positions of the leftmost and rightmost ends. Then, select the local point set of the end near the reference positions and perform line fitting to obtain the end direction vector.

[0039] Next, calculate the angle between the end direction vector and the vertical direction, normalize the angle to 0°~90°, and take the larger value of the angle between the two ends of the core block as the representative dip angle.

[0040] Finally, the dip angles are divided into five levels according to the preset range: gentle dip, moderately gentle dip, moderately steep dip, steep dip, and near vertical dip.

[0041] In this scheme, the crack dip angle is accurately calculated by fitting local point sets at the ends. The maximum value of the included angle at both ends is taken as the representative dip angle to make the judgment result more in line with the actual needs of the project. The five-level dip angle classification makes the crack dip angle result structured, which facilitates the subsequent statistical analysis of the results and enhances the reference value of the results for engineering design.

[0042] Furthermore, in step S7, the result index is organized according to one or more dimensions, such as borehole number, depth range, core block morphology category, and core block fracture dip angle classification category.

[0043] The preset screening conditions include one or more of the following: single core block morphology category, single fracture dip angle classification category, and combination of core block morphology category and fracture dip angle classification category.

[0044] In this solution, the multi-dimensional results index enables precise retrieval of core results under multiple conditions, significantly improving the efficiency of reviewing and verifying results; the diverse screening conditions make the formation of target block sets more in line with actual engineering needs, enabling targeted results output and avoiding the generation of invalid results.

[0045] Furthermore, in step S8, the classification summary table is generated by cross-summarizing one or more of the following dimensions: the number or proportion of each core block morphology category summarized by borehole and depth segment, the number or proportion of each fracture dip angle classification category summarized by borehole and depth segment, and the combination dimension of core block morphology category and fracture dip angle classification category.

[0046] In step S9, the result traceability record includes at least the borehole number, depth range, effective segmentation result version, key parameters for core block morphology classification and fracture dip angle determination, and result generation time.

[0047] In this scheme, the multi-caliber summary table makes the statistical analysis of rock core results more comprehensive, reflecting the characteristics of rock mass structure from different dimensions and providing richer evidence for engineering evaluation; the standardized content of traceability records makes the results traceable throughout the entire process, and any result file can be traced back to the input and processing conditions, which greatly improves the efficiency of result verification and the level of quality control.

[0048] The beneficial effects of this invention are:

[0049] (1) Improved traceability and verifiability of results:

[0050] By using a version solidification and consistency control mechanism after the quality control correction of the segmentation results, the corrected results are used as the sole basis for subsequent processing. At the same time, a full-process traceability record and process log are generated, which ensures the traceability of the results from a mechanism perspective and significantly reduces the cost of review.

[0051] (2) Standardization and automation of output results, improving compilation efficiency:

[0052] By classifying block morphology and fracture dip angle as control variables for the organization and output of results, a target block list, summary table, and location preview image that meet the requirements of engineering archiving are automatically generated. The results are then packaged and exported according to a preset directory structure, eliminating the need for manual secondary screening, tabulation, and mapping. This reduces subjective differences and omission risks caused by human intervention, achieves uniformity in results, and significantly improves the efficiency of core logging and results delivery.

[0053] (3) The engineering applicability and relevance of the results have been enhanced:

[0054] The system calculates morphological features at the block outline level and completes regular quantitative classification. At the same time, it performs local fitting and hierarchical output of the end fracture dip angle, so that the results not only include the identification results of the core blocks, but also the structured judgment information that can be directly used for engineering rock mass quality evaluation and design. The multi-dimensional result index and target block screening function realize the rapid location and retrieval of specific types of blocks, improving the targeting of the catalog and the efficiency of subsequent engineering analysis.

[0055] (4) The processing flow is robust and adaptable to actual engineering scenarios:

[0056] By setting up anomaly detection, marking, and fallback strategies at each key step, it can effectively address practical engineering problems such as differences in on-site shooting quality, complex core arrangement, and segmentation anomalies. It also supports parameter calculations at both pixel and actual size scales, adapting to core image processing under different shooting conditions, and possesses good engineering practicality and scalability. Attached Figure Description

[0057] Figure 1 This is a technical roadmap for the present invention.

[0058] Figure 2 This is a flowchart of the automatic output method for core image results based on block morphology parameter classification in an embodiment of the present invention. Detailed Implementation

[0059] This invention aims to provide an automatic output method for core image results based on block morphology parameter classification. This method is designed for borehole core logging and delivery scenarios. Using core photographs as input, it standardizes the outer frame area, segments and labels core blocks, implements closed-loop manual quality control correction, identifies and statistically analyzes block morphology categories, and further associates the classification results with metadata such as depth / bore number. It outputs a directly archiveable result package according to a preset result template, improving logging efficiency, consistency, and verifiability. This addresses the problems of existing core image processing solutions where the output cannot directly meet engineering archiving requirements, the lack of version management after manual quality control correction leading to untraceable results, and the inability to automate the organization and output of results based on block morphology classification, resulting in low logging efficiency, inconsistent result definitions, and high verification costs.

[0060] Its core idea is to standardize core image processing, automate output, and ensure traceable quality control. It constructs a closed-loop processing method around core box images, encompassing acquisition, preprocessing, segmentation, quality control, feature analysis, classification, output, and archiving. A segmentation result version solidification mechanism addresses the lack of management of existing technical correction results, ensuring the effectiveness of subsequent processing and the traceability of results. Quantitative classification of core block morphology features and structured grading of fracture dip angles are used as core result organization variables. The block determination results are deeply correlated with metadata such as borehole and depth, achieving automated, structured organization and standardized output. This ultimately forms a result package directly usable for engineering archiving, fundamentally solving the problems of low efficiency, large subjective differences in results, and high verification costs associated with traditional core logging, allowing core image processing results to directly meet the actual needs of engineering exploration.

[0061] The technical route of the present invention is as follows: Figure 1As shown, starting with the original core box photo bound to metadata as the initial input, the process sequentially performs outline detection and positioning, core box cropping and standardization. After image preprocessing, core block segmentation and result image generation are performed. The segmentation results are then subject to quality control correction and valid version solidification. Subsequently, based on the solidified results, the contour features of the core blocks are extracted and morphological classification is performed. At the same time, the blocks are sampled at the ends and fitted with local straight lines, and the fracture dip angle is calculated. Then, the target blocks are screened according to the morphological classification and dip angle calculation results. Finally, the core processing results are automatically output according to the preset template, forming a closed-loop core image processing and result generation system.

[0062] Example

[0063] This embodiment provides a method for automatically outputting core image results based on block morphology parameter classification. See [link to relevant documentation]. Figure 2 It includes the following implementation process:

[0064] S1. Core Image Acquisition and Metadata Registration:

[0065] In this step, core box photos are acquired, and each photo is ensured to clearly correspond to information such as "borehole number, depth range, and box sequence." This provides accurate input for subsequent frame inspection, segmentation, and output, avoiding sample mismatch and inconsistent data archiving. In one exemplary implementation, the following sub-steps are included:

[0066] S11. Core Box Photo Acquisition: Take photos of the core box of the borehole to be processed to obtain core box photos. When taking photos, try to ensure that the boundaries of the core box are complete and visible, and that the core block is clear and not obstructed; a small amount of background is allowed, but it should not affect the identification of the main body of the core box.

[0067] S12. Metadata collection: Collect at least two pieces of information for each photo, including the corresponding borehole number, the depth start and end range of the corresponding box, and the box sequence (or return number); preferably, both the borehole number and the depth range are required to ensure a unique correspondence.

[0068] S13. Consistent Binding of Photographs and Metadata: Establish a fixed association between the aforementioned metadata and the corresponding photographs. Preferably, file naming or directory structures are used to carry the metadata, enabling subsequent processing programs to automatically read and identify the borehole and depth information corresponding to the photograph. In one preferred approach, the photograph filename includes the borehole number and depth range, with box sequence as an optional supplementary field. In another approach, the photograph filename remains unchanged, the metadata is recorded in a registry table, and the registry table indicates the correspondence between the photograph filename (or full path) and the metadata.

[0069] S14. Consistency Check and Anomaly Marking: Perform a consistency check on the bound photos and metadata. If there are missing borehole numbers, missing depth ranges, or obviously unreasonable formats, or duplicate depth ranges under the same borehole and the box sequence cannot be distinguished, the corresponding photos will be marked as an anomaly and a prompt message will be output. Anomalies can proceed to subsequent steps after the information is completed.

[0070] S15. Output Results: The output includes a set of core box images with clear correspondences between borehole numbers and depth ranges, as well as corresponding metadata records (which can be in the form of filenames / directories or registration forms). The above output will serve as the input for step S2.

[0071] S2. Outer frame inspection and standardized cutting:

[0072] In this step, "effective region images that can be used for subsequent segmentation and analysis" are automatically obtained from the core box photographs, reducing the impact of background interference and shooting differences, and ensuring that the size, orientation, and content range of the output images for the same batch of samples are relatively consistent. In an exemplary implementation, the following sub-steps are included:

[0073] S21. Outer Frame Localization: Perform outer frame detection on the core box image obtained in step S1 to identify the outer boundary of the core box or the boundary of the effective working area, thereby determining the effective area to be cropped. Outer frame localization is preferably based on edge features, rectangular structure features, or color / brightness differences, or target detection can be performed using deep learning to ensure that the output area covers the main body of the core box as much as possible and excludes redundant background.

[0074] S22. Effective Area Cropping: The original photo is cropped based on the outer frame positioning results to obtain a cropped image containing only the effective area of ​​the core box. In a preferred embodiment, a small amount of margin is left outside the outer frame boundary during cropping to avoid the edge information of the core box being truncated due to overly tight boundaries.

[0075] S23. Standardization Processing: Standardize the cropped image to ensure consistent and stable input for subsequent steps. The standardization processing includes one or a combination of the following:

[0076] 1) Standardize image resolution or shorter side size;

[0077] 2) Orientation correction (to ensure the long side of the core box is aligned);

[0078] 3) Lightweight noise reduction / enhancement (improving the distinguishability of core boundaries).

[0079] Standardization does not change the relative structural relationships of the core blocks inside the core box.

[0080] S24. Failure Fallback and Anomaly Marking: When the bounding box detection fails or the detection result is obviously abnormal (e.g., the cropped area is too small, does not cover the core box body, or the cropped area is severely deviated), a fallback strategy is implemented: candidate cropping images are generated using default center cropping or fixed-ratio cropping; or the sample is marked as an anomaly, and a prompt message is output, allowing manual correction before proceeding to subsequent steps. Anomaly marking prevents erroneous cropping from directly entering subsequent segmentation and causing a chain of errors.

[0081] S25. Output results: The output includes: a set of images of the effective area of ​​the core box after standardization and cropping, and the correspondence with the metadata such as "drill hole number, depth range, box sequence" in step S1 remains unchanged, so as to serve as the input for step S3 (core block segmentation and label map generation).

[0082] S3. Core block segmentation and label generation:

[0083] In this step, the core blocks are identified and separated from the background by processing the standardized cropped image obtained in step S2. Each segment (or block) of the core is distinguished, forming the basic data required for subsequent morphological feature extraction and classification. In an exemplary implementation, the following sub-steps are included:

[0084] S31. Block Recognition and Segmentation: Core block segmentation is performed on the standardized cropped image to distinguish between core and non-core regions. Furthermore, the separated core blocks are divided into multiple independent blocks. Segmentation can be achieved using deep learning-based or image feature-based methods. This invention does not limit the specific algorithm; the key is to output stable block region results.

[0085] S32. Block Numbering and Differentiation: The segmented block regions are numbered to ensure that different blocks can be distinguished and referenced in subsequent processing. In a preferred embodiment, interconnected regions are considered as the same block; when adhesion causes connectivity anomalies, they can be output as the same block and corrected in subsequent quality control steps.

[0086] S33. Generate Label Images: Generate label images (i.e., image data that uses different labels to distinguish different blocks) from the segmentation and numbering results. These labels will be used for subsequent calculations of block size, morphological parameters, and classification. The label images should maintain the same size correspondence with the original cropped images for verification and positioning.

[0087] S34. Generate a preview overlay image for inspection: Overlay the block boundaries and numbering information onto the cropped image to generate a preview image, which is used to visually check whether there are obvious omissions, missegments, block boundary offsets, or large-area background misidentifications in the segmentation effect, thereby providing a basis for subsequent quality control corrections.

[0088] S35. Anomaly Handling and Output Results: When a significant anomaly occurs in the segmentation results (e.g., almost no blocks are identified, block areas largely cover the background, block boundaries are severely misaligned), record the anomaly and output a prompt message; if necessary, the sample can be marked as a sample to be reviewed to avoid the abnormal results directly entering the subsequent statistical and output stages.

[0089] The output of this step includes a block label image and its corresponding preview overlay image, maintaining a consistent correspondence with the sample metadata, and serves as the input for step S4.

[0090] S4. Segmentation result quality control correction and version fixation:

[0091] In this step, the segmentation results from step S3 are checked and corrected to rectify issues such as incorrect segmentation, missed segmentation, block adhesion, and unreasonable fracture segmentation. The corrected results are used as the sole basis for subsequent processing, thereby ensuring the accuracy and consistency of subsequent morphological feature extraction, morphological classification, and output. In an exemplary implementation scheme, the following sub-steps are included:

[0092] S41. Segmentation Result Check: Compare the preview overlay image output in step S3 with the corresponding cropped image to determine if the segmentation result falls under any of the following categories:

[0093] 1) Core blocks were not identified or were not completely identified (missed segments);

[0094] 2) The background was incorrectly identified as a rock core (mis-segmentation);

[0095] 3) Adjacent blocks are connected together and not separated (adhesive);

[0096] 4) The same block is unreasonably divided into multiple segments (over-segmentation);

[0097] 5) The block boundary deviates significantly from the actual edge (boundary offset).

[0098] S42. Manual or Rule-Based Correction: When the above-mentioned anomalies exist, provide correction methods to adjust the segmentation results so that the block boundaries and number of blocks conform to the actual situation. Correction methods should include at least one of the following:

[0099] 1) Add supplementary annotations to the missing segments so that the missing blocks are included;

[0100] 2) Erase or remove mis-segmented regions to prevent the background from being included;

[0101] 3) Perform a segmentation operation on the adhered blocks to separate adjacent blocks;

[0102] 4) Merge over-segmented blocks to restore the original block to its original state;

[0103] 5) Adjust the boundaries of the offset blocks to make the boundaries fit the real blocks.

[0104] S43. Result Fixing and Replacement: The corrected segmentation result is saved as a new valid result and fixed by "replacing the original result," so that subsequent steps will default to reading the corrected result instead of continuing to use the original result. In a preferred approach, the corrected result and the original result use the same naming and storage rules. By overwriting or updating, the system can automatically identify the current valid version, avoiding the problem of multiple results for the same image and subsequent reading confusion.

[0105] S44. Consistency Assurance and Recording: After completing the replacement and solidification, record the basic information of this correction, including the sample being corrected, the correction type (missed segmentation / incorrect segmentation / adhesion / oversegmentation / boundary offset, etc.), the correction time, etc., and output a prompt message. When the same sample is used again in subsequent calculations or exports, the latest solidified result should be directly called to ensure that the output results are consistent with the correction results, facilitating verification and traceability.

[0106] S45. Output results: The output is the corrected and solidified segmentation results (as input for subsequent morphological feature extraction and classification), and the corresponding inspection / correction record information; for samples where no abnormalities are found, the results of step S3 are directly used as valid inputs for subsequent steps.

[0107] S5. Block contour extraction and morphological feature calculation:

[0108] In this step, based on the segmentation results after solidification in step S4, the system automatically acquires the boundary information of each core block and automatically calculates parameters that reflect the block's shape characteristics, providing a basis for subsequent morphological classification. In one exemplary implementation, the following sub-steps are included:

[0109] S51. Block Boundary Extraction: Read the solidified segmentation results and automatically extract the boundary contour or circumscribed range of each core block in the image to obtain the boundary shape information of the block. When there are broken boundaries, obvious jagged edges, or small holes, smoothing or small-area repair processing can be performed to ensure that the boundary description is stable and reliable.

[0110] S52. Size and Shape Parameter Calculation: Based on the boundary shape of each block, the shape characteristic parameters of the block are automatically calculated. The shape characteristic parameters include at least one of the following or a combination thereof:

[0111] 1) Parameters reflecting "length": such as block length, long side dimension, and main direction dimension;

[0112] 2) Parameters reflecting "width / thickness": such as block width and short side dimensions;

[0113] 3) Parameters reflecting "proportional relationships": such as aspect ratio;

[0114] 4) Parameters reflecting the "degree of regularity of shape": such as roundness, elongation, and degree of boundary curvature. These parameters are used to describe the shape characteristics of the block, such as "whether it is slender, flat, or broken and irregular."

[0115] S53. Scale Conversion: When a scale reference exists in the image, the above parameters are automatically converted from pixel scale to actual size scale to ensure consistent results for different photos and shooting distances. If a scale reference is lacking, relative comparisons can be made at the pixel scale, and relative thresholds or normalization methods can be used to reduce the impact of shooting differences during subsequent classification.

[0116] S54. Result Summarization and Saving: The morphological feature parameters corresponding to each block are automatically recorded and summarized to form a block parameter table, maintaining a one-to-one correspondence with the original image and block number for direct retrieval in subsequent steps. In a preferred embodiment, the positional range information of the block in the image is also retained for subsequent location display and verification of the filtering results.

[0117] S55. Output results: The output includes the morphological characteristic parameters of each core block, as well as the necessary block location / outline information; the above output serves as the input for step S6.

[0118] S6. Classification of Block Morphology and Determination of Cracks:

[0119] In this step, based on the block morphology characteristics obtained in step S5, the system automatically assigns a block morphology category to each core block; simultaneously, based on the local straight-line fitting results of the block boundary, it automatically assigns a fracture dip angle (or end section dip angle) category for that block, thus providing directly usable judgment results for subsequent result screening and statistical output. In an exemplary implementation, the following sub-steps are included:

[0120] S61. Block Morphology Classification: For each core block mask region obtained in step S5, extract its outer contour point set, and calculate the block length based on the outer contour. ,width Aspect Ratio And the degree of regularity of shape, used to automatically classify and obtain the block shape category.

[0121] Wherein, the length With width It can be obtained through the minimum bounding rectangle of the block outline, satisfying... and define the aspect ratio. .

[0122] The block shape categories include at least: long columnar, columnar, short columnar, semi-columnar / pancake-shaped, and fragmented. In a preferred embodiment, a shape based on aspect ratio is used. The threshold determination and rule combination methods are classified, for example:

[0123] 1) When It was determined to be a long columnar shape;

[0124] 2) When It was determined to be columnar;

[0125] 3) When It was determined to be a short columnar shape;

[0126] 4) When It was determined to be semi-cylindrical / cake-shaped;

[0127] 5) When It was determined to be a fragmented or near-equiaxed block.

[0128] Furthermore, to reduce the interference of noise fragments on classification, an area threshold can be set. :when If the block is not found, it should be directly classified as a fragment or marked as an invalid block and the reason should be recorded.

[0129] It should be understood that the above thresholds are preferred examples used to ensure that the classification results are reproducible and interpretable. Without departing from the concept of this invention, the thresholds can be adjusted according to engineering scale, imaging resolution, and core box specifications. The key is that the classification results have a consistent correspondence with the block morphology characteristics and can be used for subsequent "rapid screening of target blocks and output of results."

[0130] S62. Crack Determination Point Selection and Local Fitting: For each block's outer contour point set, local boundary point sets are automatically selected near the left and right ends of the block, and straight line fitting is performed on these local point sets to obtain the direction vectors of the left and right ends, which serve as an approximate representation of the direction of the crack (or end section) at that end. Specifically, this includes:

[0131] 1) Obtain the block's external contour point set ;

[0132] 2) Determine the reference positions of the leftmost and rightmost ends of the block, respectively. and In a preferred embodiment, the end bandwidth is set. Select from the set of contour points that satisfy The points are used as the local point set on the left. Select the one that satisfies The points are used as the local point set on the right. ,in It can be taken as the block width of ;

[0133] 3) For each , Perform a line fitting to obtain the corresponding direction vector. , To improve robustness, least squares fitting, total least squares fitting, or principal direction fitting based on PCA can be used.

[0134] When the number of local point sets is insufficient (e.g.) or When ), it can be increased Alternatively, expand the neighborhood for compensation; if still insufficient, record the end as "unfittable" and mark the tilt angle calculation as invalid or use only the result of the fittable side.

[0135] S63. Crack dip angle calculation and value rules: Calculate the angles between the left and right end direction vectors and the vertical direction respectively to obtain the left dip angle. With right tilt angle The larger of the two values ​​is taken as the representative tilt angle of the block. The included angle is limited to .

[0136] In a preferred embodiment, let the end direction vector be... The angle between it and the vertical direction is defined as:

[0137]

[0138] when season Based on this, we obtained the following respectively. , , and take:

[0139]

[0140] in, This represents the angle of inclination.

[0141] S64. Fracture dip angle classification: based on representative dip angle Automatic classification is performed to obtain the fracture dip angle category. This includes at least:

[0142] 1) Slowly tilted;

[0143] 2) Slopes gently;

[0144] 3) : steeply sloping;

[0145] 4) : steep slope;

[0146] 5) : Nearly vertical.

[0147] S65. Output Results: The output should include at least the "block shape category + fracture dip angle value + fracture dip angle category" for each block, and maintain consistency with the block mask number, image sample and depth position; at the same time, output visualization results for verification, which should include at least the block number and its representative dip angle / dip angle category labeled in the preview image to support manual sampling, quality verification and subsequent screening statistics.

[0148] S7. Automated organization of results driven by classification outcomes:

[0149] In this step, based on the block morphology classification results and fracture dip angle determination results from step S6, each block is automatically grouped and organized according to borehole, depth segment, and category, forming a result organization structure that can be directly output later, reducing manual sheet-by-sheet sorting and manual summarization. In an exemplary implementation, the following sub-steps are included:

[0150] S71. Automatic association of positioning information: The system automatically reads the borehole number, depth range, box sequence and other information corresponding to each photo, and binds them with the "morphological category, fracture dip angle value and classification" of each block in the photo. At the same time, it retains the position range of the block in the photo for subsequent one-click positioning and verification.

[0151] S72. Automatic Generation of Results Index: The system automatically generates a results index according to preset organization rules, organizing the block results according to at least one of the following dimensions:

[0152] 1) By borehole number;

[0153] 2) By depth range;

[0154] 3) By block shape category;

[0155] 4) Classification by fracture dip angle.

[0156] The result index can automatically locate all block sets under any specified borehole, specified depth range, specified shape category, or specified dip angle classification.

[0157] S73. Automatically generate target block sets (for subsequent output): The system automatically generates one or more target block sets based on preset filtering conditions, for example:

[0158] 1) Select only blocks of a certain shape category;

[0159] 2) Select only blocks with a specific fracture dip angle for grading;

[0160] 3) Blocks that simultaneously meet the morphological category and fracture dip angle classification conditions.

[0161] The target block set is used to automatically generate lists, previews, and result files in subsequent steps, avoiding omissions due to manual screening.

[0162] S74. Output Results: The output includes the result index and target block set information, and maintains a consistent correspondence with the original sample. It serves as the input for step S8 to achieve automatic generation and archiving of results.

[0163] S8. Output Generation:

[0164] In this step, based on the result index and target block set obtained from S7 automated organization, and according to preset result output rules, the system automatically generates result files that can be directly used for review and archiving. These include screening lists, summary tables, location preview images, and necessary process records, thereby reducing manual organization and re-copying of tables and charts. In one exemplary implementation, the following sub-steps are included:

[0165] S81. Determine Output Content and Scope: Based on the result index and target block set, the system determines the output content and scope to be output. The output content includes at least one of the following or a combination thereof:

[0166] 1) Summary results by block shape category;

[0167] 2) Summary results of classification by fracture dip angle;

[0168] 3) A list of target blocks that simultaneously meet the morphological category and fracture dip angle conditions;

[0169] 4) Preview of the target block's location (for quick verification).

[0170] Optionally, the system can also output length-related statistics or distribution information derived from the block parameters.

[0171] S82. Generate a list of target blocks: Automatically generate a list file for the set of target blocks. The list should include at least the corresponding borehole number, depth range, block shape category, fracture dip angle value and its classification category, and retain reference information that can be used to locate the original image (such as image file name, block number or block location description) so that reviewers can quickly find the corresponding block.

[0172] S83. Generate Summary Table: The system automatically summarizes the list according to a preset summary standard and generates a summary table file. The summary standard includes at least one of the following:

[0173] 1) Summarize the quantity / percentage of each morphological category according to borehole depth;

[0174] 2) Summarize the number / percentage of each dip angle classification according to borehole and depth segment;

[0175] 3) Cross-summarize based on the combination dimension of morphological category and tilt angle classification.

[0176] The summary table is used to reflect the differences in structural characteristics of different boreholes and different depths, which facilitates subsequent engineering analysis and report writing.

[0177] S84. Generate Positioning Preview and Verification Annotations: The system annotates the position of the target block on the original cutout drawing and automatically generates a positioning preview. The preview can be annotated with the block shape category and crack dip angle information, so that the reviewers can intuitively check whether the classification and dip angle determination are reasonable without having to go back to the intermediate process.

[0178] S85. Output organization and naming archiving: The system automatically organizes output files according to the preset directory structure and naming rules, ensuring that the output files of the same borehole and the same depth are stored together, and that the file name can reflect the borehole number, depth range and output type, so as to avoid confusion in subsequent archiving.

[0179] S86. Anomaly Handling: When a sample has missing key fields required for output, an empty target block set, or fails to generate a preview image, the system records the reason for the anomaly and outputs a prompt message; abnormal samples do not affect the generation of results for other samples, ensuring the stable completion of the batch output process.

[0180] S87. Output Results: The output includes a list of target blocks, a summary table, a location preview image, and necessary error messages; the above output serves as the input for step S9 and is used for final packaging and traceability recording.

[0181] S9. Export of Results Packages and Traceable Logs:

[0182] In this step, the output files generated by S8 are uniformly collected and exported, and key process information and anomaly information are recorded, making the output results traceable, verifiable, and reproducible, facilitating project delivery and subsequent quality inspection. In an exemplary implementation scheme, the following sub-steps are included:

[0183] S91. Collection of Results Documents: The system will collect the list, summary table, positioning preview image and abnormal prompt information output in step S8 according to the preset directory structure to ensure that the documents of the same borehole, the same depth section and the same result type are stored together to avoid the results being scattered and resulting in omissions or mismatches.

[0184] S92. Generate a traceability record: The system generates a traceability record file for this process, which describes "what inputs and processes were involved in generating this result". The traceability record includes at least the following:

[0185] 1) Corresponding borehole number, depth range, box sequence, and other information;

[0186] 2) The version of the segmentation results used in this study (i.e., the latest results after S4 solidification);

[0187] 3) Summary of key parameters used for morphological classification and fracture dip angle determination (e.g., determination threshold or classification interval).

[0188] 4) Time information and generation status of the output results.

[0189] Through the aforementioned traceability records, any output document can be traced back to its corresponding input and processing conditions.

[0190] S93. Process Log and Exception Log Output: The system records process logs during batch processing, including the execution status of each step, the number of steps processed, the reasons for skipping, and the reasons for failure. When an exception occurs, the exception type and exception location are recorded (e.g., frame detection failure, segmentation exception, preview generation failure, missing key fields, etc.) for subsequent review and remediation.

[0191] S94. Exporting the Deliverables Package: The collected deliverables files, traceability records, and log files are packaged and exported together to form a deliverables package. The deliverables package can be exported as a compressed file, and a consistent directory structure and naming rules should be maintained within the package to facilitate delivery, archiving, and review.

[0192] S95. Output Results: The output includes the final deliverables package and the corresponding traceability records and log files; the deliverables package can be directly used for project deliverable archiving, quality verification and subsequent supplementary processing.

[0193] Although embodiments of the present invention have been described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the present invention, and all such changes and alterations shall not depart from the protection scope of the present invention.

Claims

1. A method for automatically outputting core image results based on block morphology parameter classification, characterized in that, The method includes the following steps: S1. Obtain core box photos and bind metadata such as borehole number and depth range to each photo to form a core box photo set with metadata association; S2. Perform outer frame positioning, effective area cropping and standardization processing on the core box photos to obtain a set of standardized cropped effective area images of the core box; S3. Perform core block identification, segmentation, and block numbering on the effective area image of the core box, and generate a block label image and a preview overlay image; S4. Check and correct the segmentation results of the block label image, and solidify the corrected segmentation results as valid results; S5. Based on the solidified segmentation results, extract the boundary contours of each core block, calculate the morphological characteristic parameters of the core block, and form a record of the block morphological characteristic parameters. S6. Based on the morphological characteristic parameters of the core blocks, the rock core blocks are classified into different shapes. The fracture dip angle is calculated by fitting the local straight line of the end contour of the rock core block and the dip angle is graded. The morphological category, fracture dip angle value and dip angle grading result of each rock core block are obtained. S7. Associate the morphological category, fracture dip angle value, and dip angle classification results of the core blocks with the metadata of borehole number and depth range to generate a result index and form a target block set according to preset screening conditions; S8. Based on the results index and target block set, generate a target block list, a classification summary table, and a block location preview image; S9. Collect the target block list, classification summary table, and block location preview image, generate result traceability record and process processing log, and package and export them according to the preset directory structure to form a result package that can be directly archived.

2. The method for automatically outputting core image results based on block morphology parameter classification as described in claim 1, characterized in that, In step S1, when binding metadata to core box photos, the box sequence or cycle number information is also bound. A consistency check is performed on the core box photos with completed metadata binding. Samples with missing metadata, unreasonable metadata format, or duplicate depth range in the same borehole and indistinguishable box sequence are marked as abnormal samples and prompt information is output. Abnormal samples are processed in subsequent steps after the information is completed.

3. The method for automatically outputting core image results based on block morphology parameter classification as described in claim 1, characterized in that, In step S2, the standardization processing of the core box photos includes: unifying the image resolution or short side size, correcting the long side direction of the core box, light denoising, or enhancing the core boundary, and the standardization processing does not change the relative structural relationship of the core blocks inside the core box; when the outer frame positioning fails or the positioning result is abnormal, the default center cropping or fixed ratio cropping is used to generate candidate cropping images, or the corresponding sample is marked as an abnormal sample and a prompt message is output.

4. The method for automatically outputting core image results based on block morphology parameter classification as described in claim 1, characterized in that, In step S4, the correction methods for the segmentation results include one or more of the following: supplementing the annotation of the missing core block areas, erasing and removing the incorrectly segmented background areas, segmenting the adhered core blocks, merging the same core blocks that have been over-segmented, and adjusting the boundaries of core blocks with offset boundaries. When solidifying the corrected segmentation results, the original segmentation results are replaced by overwriting or updating, and the information of the corrected sample, the correction type, and the correction time are recorded.

5. The method for automatically outputting core image results based on block morphology parameter classification as described in claim 1, characterized in that, In step S5, the morphological characteristic parameters of the core block include one or more of the parameters reflecting the length of the core block, the parameters reflecting the width of the core block, the length-width ratio reflecting the length-width ratio relationship of the core block, and the parameters reflecting the degree of regularity of the core block shape.

6. The method for automatically outputting core image results based on block morphology parameter classification as described in claim 5, characterized in that, In step S6, morphological classification of the core block is performed based on the block morphological characteristic parameters, including: Obtain the minimum bounding rectangle based on the boundary contour of the core block, obtain the length L and width W of the core block from the minimum bounding rectangle and satisfy L≥W>0, calculate the length-width ratio R = L / W, and perform morphological classification of the core block according to the threshold interval of the length-width ratio. At the same time, set the area threshold Amin, and classify the core block with an area A < Amin as fragmented or mark it as an invalid block.

7. The method for automatically outputting core image results based on block morphology parameter classification as described in claim 6, characterized in that, The morphological classification of the core block according to the threshold interval of the length-width ratio includes: When R≥6, the core block is long columnar; When 3≤R<6, the core block is columnar; When 1.8≤R<3, the core block is short columnar; When 1.2≤R<1.8, the core block is semi-columnar / cake-shaped; When 1≤R<1.2, the core block is fragmented.

8. The method for automatically outputting core image results based on block morphology parameter classification as described in claim 1, characterized in that, In step S6, the fracture dip angle is calculated by local linear fitting of the end contour of the core block and the dip angle classification is completed, including: First, obtain the outer contour point set of the core block and determine the leftmost and rightmost reference positions. Select the end local point set near the reference position and perform linear fitting to obtain the end direction vector; Then calculate the angle between the end direction vector and the vertical direction and normalize the angle to 0°~90°. Take the larger value of the angles at both ends of the core block as the representative dip angle. Finally, divide the representative dip angle into five dip angle grades of gentle dip, medium gentle dip, medium steep dip, steep dip, and near vertical according to the preset interval.

9. The method for automatically outputting core image results based on block morphology parameter classification as described in claim 1, characterized in that, In step S7, the result index is organized according to one or more dimensions of the borehole number, depth range, core block morphological category, and core block fracture dip angle classification category; The preset screening conditions include one or more of a single core block morphological category, a single fracture dip angle classification category, and a combination of the core block morphological category and the fracture dip angle classification category.

10. A method for automatically outputting core image results based on block morphology parameter classification as described in any one of claims 1 to 9, characterized in that, In step S8, the classification summary table is generated according to one or more calibers of summarizing the quantity or proportion of each core block morphological category by borehole and depth segment, summarizing the quantity or proportion of each fracture dip angle classification category by borehole and depth segment, and cross-summarizing according to the combination dimension of the core block morphological category and the fracture dip angle classification category; In step S9, the result trace record includes at least the borehole number, depth range, valid segmentation result version, key parameters of the core block morphological classification and fracture dip angle determination, and result generation time.

Images