Mineral exploration drilling electronic map element sorting method, system, device and medium

Abstract

The present application provides a kind of mineral exploration borehole electronic map element sorting method, system, equipment and medium, belong to geological information processing technical field, its method includes the following steps: identifying table line in mineral exploration borehole electronic map, based on table line, table is split into multiple independent text image blocks and lithology symbol image block according to the content in cell corresponding;Text in multiple independent text image blocks is recognized using optical character recognition algorithm, obtain text element in mineral exploration borehole electronic map;Convolutional neural network is trained, multiple independent lithology symbol image block is input into trained convolutional neural network, lithology symbol recognition is carried out, and different types of lithology symbol element are obtained.The present application can solve the problem that map size is large, table row and column are many, data structure is complex, and text table is dense, effectively improve the correctness of information sorting, promote the utilization efficiency and fusion sharing of mineral exploration data.

Description

Technical Field

[0001] This invention belongs to the field of geological information processing technology, specifically relating to a method, system, equipment, and medium for sorting elements in electronic maps of mineral exploration boreholes. Background Technology

[0002] Data from different stages of mineral resource exploration forms the basis for scientific decision-making and is one of the nation's fundamental strategic resources. As mineral exploration becomes increasingly challenging, abundant and high-quality mineral resource exploration data, especially deep-penetration data revealed by boreholes, has become a key focus for intelligent mineral resource exploration. Currently, a large amount of mineral resource exploration borehole data has only reached the level of digitization, not complete digitalization, becoming a bottleneck for information resource utilization and urgently needing to be addressed to unlock its value. The application of artificial intelligence theories, methods, and technologies can help fill this gap in the transition from digitized to electronic mineral exploration borehole maps, fully leveraging the value of big data in mineral resource exploration and providing crucial technical accumulation and data support for the next round of mineral resource exploration and evaluation.

[0003] In practice, electronic maps of mineral resource exploration boreholes are mostly hand-drawn or created using specialized software, leading to issues such as inconsistent preset templates and complex map structures, posing significant challenges to information extraction. Currently, most automated recognition software only achieves high accuracy for well-structured text or tables. Information sorting for unstructured elements (images, videos, audio, etc.) in electronic maps of mineral resource exploration boreholes largely relies on manual labor, resulting in low efficiency, low accuracy, and high costs, thus limiting the comprehensive utilization of geological data.

[0004] Existing methods for recognizing electronic maps focus on text recognition and digitization, achieving high accuracy in text recognition. However, for electronic maps of boreholes in mineral resource exploration, which are large in size, have many rows and columns in the tables, complex data structures, and dense text tables, existing methods cannot handle this type of map well. This limits the information mining and application of unstructured data in geological big data and is not conducive to promoting the integration and sharing of mineral exploration data. Summary of the Invention

[0005] To overcome the shortcomings of the existing technology, the present invention provides a method for sorting elements in electronic maps of mineral exploration boreholes, comprising the following steps:

[0006] Identify table lines in electronic maps of mineral exploration boreholes, and based on the table lines, split the table into multiple independent text image blocks and multiple independent lithological symbol image blocks according to the content within the cells;

[0007] An optical character recognition algorithm is used to identify the text in multiple independent text image blocks to obtain text elements in the electronic map of mineral exploration boreholes. A lithological symbol image database is constructed based on the lithological symbols appearing in the electronic map of mineral exploration boreholes. The images in the lithological symbol image database are input into a convolutional neural network to train the convolutional neural network. Multiple independent lithological symbol image blocks are then input into the trained convolutional neural network to perform lithological symbol recognition and obtain different types of lithological symbol elements.

[0008] Preferably, the step of identifying table lines in the electronic map of mineral exploration boreholes, and dividing the table into multiple independent text image blocks and lithological symbol image blocks according to the content within each cell, includes the following steps:

[0009] Identify horizontal lines, vertical lines, row numbers, and column numbers in electronic maps of mineral exploration boreholes; set windows based on thresholds; and filter out horizontal and vertical lines in cells by sliding the window.

[0010] Merge the intersections of the horizontal and vertical lines of the cells, calculate the coordinates of the intersections in the electronic map of the mineral exploration borehole, restore the table structure of the electronic map of the mineral exploration borehole based on the coordinates, and store the table structure of the electronic map of the mineral exploration borehole in the form of point coordinates and line coordinates.

[0011] Based on the coordinates, the electronic map of the mineral exploration borehole is split into multiple independent text image blocks and lithological symbol image blocks.

[0012] Preferably, before identifying the table lines in the electronic map of mineral exploration boreholes, the method further includes preprocessing the electronic map of mineral exploration boreholes, including the following steps:

[0013] Red stamp detection is performed on electronic maps of mineral exploration boreholes. Pixels with red channels exceeding the threshold are extracted, assigned as background values, and the red stamps are removed.

[0014] By setting a threshold, the electronic map of mineral exploration boreholes after removing red stamps is binarized and morphologically operated to eliminate noise and ambiguity in the electronic map of mineral exploration boreholes.

[0015] The page orientation of the electronic map of mineral exploration boreholes after noise and ambiguity removal is detected and rotated, and the image is vertically projected. The peak value of the histogram of the projected image is analyzed, and the electronic map of mineral exploration boreholes with non-positive orientation is corrected.

[0016] The tables in the adjusted electronic map of mineral exploration boreholes are extracted, the distorted tables are corrected, and other background areas are cropped.

[0017] Preferably, the step of performing binarization and morphological operations on the electronic map of mineral exploration boreholes after removing red stamps by setting a threshold specifically involves: converting the color image into a grayscale image, binarizing the grayscale image by setting a threshold, enhancing image features through dilation and erosion operations, and eliminating noise and blur in the electronic map of mineral exploration boreholes.

[0018] Preferably, after obtaining different types of lithological symbol elements, the method further includes correcting the text elements based on the obtained different types of lithological symbol elements, including the following steps:

[0019] The correct value of the strata is determined by the confidence scores of textual elements and lithological symbol elements.

[0020] The accuracy of the identification content in the electronic maps of mineral exploration boreholes is checked through manual interpretation.

[0021] Preferably, the step of determining the correct value of the stratum based on the confidence scores of text element recognition and lithological symbol elements specifically involves: determining whether the confidence score of text element recognition is greater than a set threshold; if so, the stratum is considered to be the stratum described in the text element; if the confidence score of text element recognition is not greater than the set threshold, determining whether the confidence score of lithological symbol elements is greater than the set threshold; if so, the stratum is considered to be the stratum corresponding to the lithological image classification result; if the confidence score of lithological symbol elements is still not greater than the set threshold, the text recognition result is used as a temporary output, and manual inspection is indicated during manual verification.

[0022] This invention also provides a mineral exploration borehole electronic map element sorting system, comprising:

[0023] The table splitting module is used to identify table lines in electronic maps of mineral exploration boreholes and split the table into multiple independent text image blocks and multiple independent lithology symbol image blocks according to the content in the cells based on the table lines.

[0024] The element sorting module is used to identify text in multiple independent text image blocks using an optical character recognition algorithm to obtain text elements in the electronic map of mineral exploration boreholes. Based on the lithological symbols appearing in the electronic map of mineral exploration boreholes, a lithological symbol image database is constructed. The images in the lithological symbol image database are input into a convolutional neural network to train the convolutional neural network. Multiple independent lithological symbol image blocks are input into the trained convolutional neural network to perform lithological symbol recognition and obtain different types of lithological symbol elements.

[0025] The present invention also provides a computer device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute the method for sorting elements of the electronic map of the mineral exploration borehole.

[0026] The present invention also provides a computer-readable storage medium storing a computer program adapted for loading by a processor to execute the method for sorting elements of electronic maps of mineral exploration boreholes.

[0027] The mineral exploration borehole electronic map element sorting method, system, equipment, and medium provided by this invention have the following beneficial effects:

[0028] This invention identifies table lines in electronic maps of mineral exploration boreholes and, based on these lines, divides the table into multiple independent text image blocks and lithological symbol image blocks according to the content within each cell. This process effectively separates elements within the electronic maps of mineral exploration boreholes, ensuring that the text segments within the table are independent and preventing misidentification due to closely spaced numbers and text in the table structure. It solves the problems of large map size, numerous table rows and columns, complex data structures, and dense text tables, thus facilitating the accurate identification of different types of elements in complex electronic maps of mineral exploration boreholes.

[0029] This invention identifies text information by recognizing text image blocks and lithological symbol information by recognizing lithological symbol image blocks. By separately recognizing tables, text, and lithological symbols, this invention enables the effective utilization of all elements of a map, significantly improving the accuracy of information sorting and promoting the efficiency and fusion of mineral exploration data. Attached Figure Description

[0030] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 A schematic flowchart illustrating an embodiment of a method for sorting element information in electronic maps of mineral exploration boreholes, provided for the implementation of this invention;

[0032] Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S1.

[0033] Figure 3 for Figure 1 A detailed flowchart illustrating the implementation steps of step S2.

[0034] Figure 4 for Figure 1A detailed flowchart illustrating the implementation steps of step S4.

[0035] Figure 5 for Figure 1 A detailed flowchart illustrating the implementation steps of step S5.

[0036] Figure 6 This diagram illustrates common text recognition errors in ordinary OCR algorithms, as well as errors detected using corpus, rules, and image recognition.

[0037] Figure 7 The sorting results of electronic map elements from mineral exploration boreholes, including: Figure 7 (b) and (c) are respectively Figure 7 Enlarged view of the two red boxes on the left and right in (a). Detailed Implementation

[0038] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0039] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the technical solution of this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0040] Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. In the description of this invention, it should be noted that, unless otherwise explicitly specified or limited, the terms "connected" or "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances. In the description of this invention, unless otherwise stated, "a plurality of" means two or more, which will not be elaborated further here.

[0041] Example

[0042] This invention provides a method for sorting elements in electronic maps of mineral exploration boreholes, specifically as follows: Figure 1 As shown, it includes the following steps:

[0043] Step S1: Preprocess the electronic maps of mineral exploration boreholes.

[0044] Electronic maps of mineral exploration boreholes can be stored electronically as graphic files, used to display stratigraphic and sample attribute information of the boreholes. They are primarily electronic maps obtained by scanning original paper maps and can be image files (such as JPEG, PNG, TIF, etc.) or portable files (PDF format). The files contain all visual elements of the electronic map, such as text, lines, and symbols, and can be in color or monochrome format. Electronic maps of mineral exploration boreholes reflect the properties of different strata and rocks traversed by the boreholes during mineral exploration. They comprehensively record basic borehole information and geological information for each layer, including stratum name, thickness, lithological description, sampling location, and sample attributes, presented in an intuitive graphical format. In geological exploration and mineral development, electronic maps of mineral exploration boreholes are crucial for geological engineers and researchers in geological interpretation and mineral assessment.

[0045] Understandably, the preprocessing of electronic maps of mineral exploration boreholes directly impacts the accuracy and efficiency of subsequent information extraction. In the process of automatically recognizing the content of text and image tables and sorting element information, determining the table's boundaries is a prerequisite for accurately identifying different geological information elements (such as stratigraphic lines and lithological symbols). Only by clearly defining the boundary range can elements within the boundary be accurately analyzed and processed. Before recognizing the content of text and image tables, the original image needs to be processed through red stamp detection, binarization, and morphological operations to eliminate interference from non-map elements. Only by resolving noise and blurring issues in historical maps and avoiding misjudgment and extraction of irrelevant information can the accuracy and correctness of recognition be improved. During recognition, it is necessary to determine the correct page orientation and adjust the appropriate page angle. Only by obtaining the corrected image can the correct table structure and text information be extracted.

[0046] Step S2: Identify the table lines in the electronic map of mineral exploration boreholes, and based on the table lines, split the table into multiple independent text image blocks and multiple independent lithology symbol image blocks according to the content in the cells.

[0047] Among them, the table refers to the framework structure that makes up the electronic map of mineral exploration boreholes. It is mainly composed of horizontal and vertical line segments. The rows and columns of the table in the map may not be consistent. There may be relatively complex merging and splitting relationships, and the length of the lines in each row or column may be inconsistent.

[0048] It is understandable that correctly extracting the table structure is crucial for sorting table element information in maps. Furthermore, the table segmentation directly impacts the accuracy and efficiency of subsequent information extraction. In text recognition, traditional methods, which simultaneously recognize entire pages of text, struggle with the large map size, numerous table rows and columns, complex data structures, and dense text tables found in electronic maps of mineral exploration boreholes. Using table line recognition to differentiate and identify each unit of content helps improve the accuracy of text recognition. Simultaneously, by segmenting lithological symbols based on the table structure, the segmented symbols can be independently classified into image categories.

[0049] Step S3: Use an optical character recognition algorithm to recognize the text in multiple independent text image blocks to obtain the text elements in the electronic map of mineral exploration boreholes.

[0050] The text refers to various types of text in the electronic maps of mineral exploration boreholes, including Chinese characters, English characters, numbers, and special symbols. After step S2, all text is split into independent image blocks, and character detection and recognition are performed separately for each block. However, some stratigraphic age symbols are composed of combined characters, which have poor accuracy when recognized individually. This type of symbol should be classified and information extracted using the image recognition algorithm in step S4.

[0051] Understandably, the recognition of basic text mainly relies on optical character recognition (OCR) algorithms. Existing OCR algorithms are relatively mature, and in this embodiment of the invention, the appropriate algorithm can be selected based on the specific situation and requirements.

[0052] Step S4: Train the convolutional neural network by inputting multiple independent lithological symbol image blocks into the trained convolutional neural network to perform lithological symbol recognition and obtain different types of lithological symbol elements.

[0053] Lithological symbols are a crucial component of geological information representation in electronic maps of mineral exploration boreholes. These symbols typically appear in various geometric primitives to represent different lithological information. For example, lines, dots, text, and special symbols are among the most basic primitives, appearing in different directions, lengths, sizes, and thicknesses to represent key information such as the lithology and compositional proportions of strata. For instance, dots represent sandy soil, horizontal lines represent argillaceous soil, and a regular arrangement of multiple dots represents silty soil. Besides single images, lithological symbols can also employ combinations of various graphics to form complex patterns representing mixed lithological compositions, such as siliceous limestone and sandy dolomite. In this embodiment of the invention, classifying lithological symbols in the map using a convolutional neural network-based method accurately identifies different geological strata. This helps to better reconstruct the text and graphic symbol elements in the map based on the distribution of strata.

[0054] Understandably, in step S3, some combined characters, such as combined geological age symbols, are difficult to recognize with high accuracy using conventional OCR algorithms. Therefore, this type is recognized using the same method as in step S4. By constructing a lithological image symbol library, the actual semantic information corresponding to the geological age symbols is obtained, thereby sorting the element information of this type.

[0055] Specifically, before performing image symbol recognition, a lithological image symbol library of a certain size can be established, containing the relationship between symbols and names. The same lithology may have multiple lithological symbols, therefore the lithological image symbol library needs to be continuously expanded to ensure the accuracy of lithological digitization. In this invention, after each new image is identified, classified, and the results are confirmed, the symbol and its corresponding lithology are entered into the lithological image symbol library, thereby effectively expanding the lithological image symbol library while performing recognition.

[0056] Step S5: Combine the sorting results of stratigraphic symbol information with manual review to correct the text recognition results.

[0057] Understandably, existing methods, whether based on knowledge rules or deep learning, cannot achieve completely accurate recognition results for all text, tables, and graphic symbols. Therefore, correcting the recognition results is necessary. Correction can combine manual correction and machine-assisted judgment, which can improve efficiency and reduce workload while maintaining accuracy.

[0058] Step S6: Combine knowledge rules and entity semantic recognition to perform structured storage of information sorting results.

[0059] The knowledge rules can be the spatial relationship between each element's title and the element itself. Specifically, in some embodiments, by summarizing the left-right spatial relationship between texts such as "hole number" and "drill hole number" and their actual corresponding values, the numerical text appearing to the right of "hole number" and "drill hole number" can be stored according to the attribute meaning of the hole number. Entity semantic recognition refers to an algorithm based on natural language processing (NLP) that extracts information of interest to users from text and forms structured data. This method is trained on a large amount of data, is not limited to specific row and column relationships, and has a wider range of applications, but its matching accuracy is low in some embodiments. This invention combines knowledge rules and deep learning methods, complementing each other's results, and taking the maximum matching result as the final result.

[0060] Understandably, up to step S5, the identified mineral exploration borehole electronic map elements only possess content and coordinate information, lacking semantic information. Specifically, for example, during storage, it cannot automatically interpret "ZK1" as a borehole number, nor can it automatically link information within each stratigraphic row. Therefore, the stored data remains in a semi-structured state. Such data is detrimental to information sorting and effective utilization. Step S6 enables the unstructured and semi-structured mineral exploration borehole electronic map element information to be structurally sorted.

[0061] In some embodiments, the meaning of attribute information in element information may not be labeled by preset knowledge rules or relation extraction results in entity semantic recognition algorithms. In such cases, information sorting is achieved through human-computer interaction. For example, by selecting a set of keys and values, the attribute of the value is labeled. For instance, by clicking on the "ZK1" cell and the borehole number attribute in the dictionary, the value of "ZK1" is assigned to the borehole attribute.

[0062] This invention preprocesses electronic maps of mineral exploration boreholes in step S1, ensuring that images that have undergone correction, skew-correction, and image enhancement are accepted for subsequent recognition steps, effectively improving the accuracy of map element information extraction. Steps S2, S3, and S4 respectively identify and extract tables, text, and images from the maps, enabling more precise and streamlined extraction and sorting of elements from electronic maps of mineral exploration boreholes. Furthermore, the knowledge-rule-based and deep learning-based method for sorting element information from electronic maps of mineral exploration boreholes enhances semantic analysis and automated sorting capabilities of the identified text, effectively improving the accuracy and efficiency of information sorting. This achieves effective utilization of all map elements, promoting the efficiency and fusion of mineral exploration data.

[0063] For a detailed flowchart of the implementation steps of step S1, please refer to [link / reference]. Figure 2 As shown, it includes the following steps:

[0064] Step S11: If a red stamp appears in the original image, red stamp detection and removal are required to avoid interfering with the recognition results. Red stamp detection is performed using an image processing-based algorithm. Pixels with red channels greater than the threshold are extracted and assigned as background values, thereby eliminating the red stamp.

[0065] In some embodiments, the method for red stamp detection can be to identify the range of the red stamp using image processing techniques. For example, edge detection algorithms can be used to identify edge information in the image, thereby determining the range of red stamps that need to be removed. Alternatively, similarity detection methods can be used. For example, SIFT algorithms and text similarity detection algorithms can be used to obtain possible red stamps. Alternatively, pixel channel thresholding methods can be used, such as setting a threshold for the red channel to be greater than a set threshold, or setting the red channel to be greater than the green and blue channels, to filter out the range of red stamps. The first two methods are relatively effective for all images; the latter works better for color images, and the recognition and removal accuracy is greater than the former two. In the embodiments of the present invention, the appropriate method can be selected according to specific circumstances and needs.

[0066] Step S12: Perform binarization and morphological operations on the map after removing the red stamps to enhance the contrast and clarity of the image. This includes converting the color image to a grayscale image, binarizing the grayscale image through a set threshold, and enhancing image features through dilation and erosion operations to solve the noise and blur problems in historical mineral exploration borehole electronic maps.

[0067] In some embodiments, noise removal and image enhancement are performed using image morphology methods. For example, erosion operations are used to remove small objects and details, reducing the boundaries of objects in the image. For each pixel, only the minimum value in its neighborhood is retained, or dilation methods can be used to enhance object boundaries, fill small holes, and expand the boundaries of objects in the image. For each pixel, the maximum value in its neighborhood is retained, or opening operations can be used, performing erosion first and then dilation, typically used to remove small objects or noise while preserving the shape of larger objects. Alternatively, closing operations can be used, performing dilation first and then erosion, to fill small holes and breaks, typically making the edges of objects more continuous. In these embodiments, the appropriate method can be selected based on specific circumstances and requirements.

[0068] Step S13: Detect and rotate the page orientation. By projecting the image vertically, observe the distribution of text in the horizontal direction. By analyzing the peak value of the histogram of the projection image, determine the main orientation of the text and adjust the image whose orientation is not positive.

[0069] Step S14: Extract the table region, use a deep learning-based table recognition algorithm (PaddlePaddle PP-YOLOv2) to correct the distorted table, extract the basic range of the table region, and crop other background regions.

[0070] This invention performs preprocessing on the map, including red stamp removal, binarization and image enhancement, orientation detection and correction, and table region extraction. The result is a clear image with offset interference and noise interference removed. This means that the accuracy of feature recognition can be improved automatically and in a streamlined manner through the preprocessing.

[0071] For a detailed flowchart of the implementation steps of step S2, please refer to [link / reference]. Figure 3 As shown, it includes the following steps:

[0072] Step S21: Identify horizontal and vertical lines, obtain the number of rows and columns of the image, set the window according to the threshold, and filter out the horizontal and vertical lines by sliding the window. When identifying horizontal lines, the height of the window is 1 and the width is the number of columns divided by 20. When identifying vertical lines, the width of the window is 1 and the height is the number of rows divided by 40.

[0073] In some embodiments, the recognition accuracy of horizontal and vertical lines in a table can be improved by adjusting the window threshold, depending on the specific situation and requirements, as the line thickness and fragmentation may affect the accuracy of the line recognition.

[0074] Step S22: Merge the intersections of the horizontal and vertical lines, calculate the coordinates of the intersections in the image, restore the electronic map table structure of the mineral exploration borehole, and store the map table structure in the form of point coordinates and line coordinates.

[0075] Step S23: Based on the segmentation coordinates of the table lines, split the original image into local regions and save them.

[0076] This invention identifies table lines and extracts table element information. The key step is ensuring the table is upright and the lines are clear and identifiable. This requires table correction, distortion correction, and table enhancement based on image and morphology algorithms. The result is a data structure with table position coordinates, indicating the table's position and shape. This means that tables can be decomposed using line recognition methods, addressing the difficulty of recognizing complex tables in mineral exploration borehole electronic maps. Based on the table lines, cells are split, and fragmented text is input into an OCR algorithm for recognition. This means that each text and graphic fragment will be independently recognized and matched, solving the problem of dense text easily sticking together and misaligned lines in general models.

[0077] This invention uses an OCR algorithm to recognize text in tables. The key to this step is to ensure that the text segments within the table are independent and that the table structure does not cause adjacent numbers and text to stick together, leading to incorrect recognition. This means that table segmentation based on table line recognition is necessary.

[0078] For a detailed flowchart of the implementation steps of step S4, please refer to [link / reference]. Figure 4 As shown, it includes the following steps:

[0079] Step S41: Construct a lithological image symbol library of a certain size for the lithological symbols that may appear in the electronic maps of mineral exploration boreholes. The lithological image symbol library is saved by <mineral lithology name, lithological symbol image>.

[0080] Step S42: Divide the lithological image symbol library into training set, validation set and test set according to the proportion. Extract image features through convolutional neural network, construct lithological symbol classification model, and use cross-entropy loss function and optimization algorithm to update model parameters.

[0081] Step S43: Use the trained model to predict the lithological symbol category in the input mineral exploration borehole electronic map, the input category and its predicted probability.

[0082] Step S44: After sorting the element information of each mineral exploration borehole electronic map, the sorted lithological images and their corresponding strata are added to the lithological image symbol library for updating the recognition model.

[0083] This invention achieves the identification and extraction of image element information from electronic maps of mineral exploration boreholes by recognizing lithological symbol images. For image classification based on Convolutional Neural Networks (CNNs), various appropriate CNN architectures can be selected, including ResNet, Inception, and VGG. The results of each round of recognition are added to a lithological image symbol library. Training can be repeated every fixed number of new images to expand the dataset and improve model accuracy. Image recognition also helps correct text recognition errors, improving the accuracy of element recognition.

[0084] For a detailed flowchart of the implementation steps of step S5, please refer to [link / reference]. Figure 5 As shown, it includes the following steps:

[0085] Step S51: Determine the correct value of the stratum based on the confidence scores of text element recognition and symbol recognition. Step S51 specifically includes:

[0086] Step S511: Determine whether the confidence level of text element recognition is greater than the set threshold. If so, the stratum is considered to be the stratum described in the text.

[0087] Step S512: If the confidence level of text element recognition is not greater than the set threshold, then determine whether the confidence level of lithological symbol element is greater than the set threshold. If so, then the stratum is considered to be the stratum corresponding to the lithological image classification result.

[0088] Step S513: If the confidence level of the lithological symbol element is still not greater than the set threshold, the text recognition result is used as a temporary output, and manual inspection is required during manual verification.

[0089] This invention integrates the confidence scores of text element recognition and lithological symbol element recognition to determine the stratigraphic lithology identification results, which helps improve the identification accuracy of strata that are difficult to identify and whose element information is critical, such as... Figure 6 As shown.

[0090] Step S52: Using a manual interactive approach, manually interpret the text information in the electronic map of the mineral exploration borehole, including basic borehole information, formation depth information, formation description information, and parameter information.

[0091] Step S6 includes the following steps:

[0092] Step S61: Construct a knowledge rule base based on common key-value spatial relationships found in electronic maps of mineral exploration boreholes. In these maps, key-value spatial relationships refer to the relationships between various geological and exploration data and spatial locations. Here, the "key" typically refers to the identifier of a specific location or feature in geographic information (such as borehole number, coordinates, etc.), while the "value" represents the specific data or attributes associated with that location (such as geological strata, mineral content, lithological characteristics, etc.). The core of key-value spatial relationships is how data from different locations in space are interconnected and how spatial analysis reveals the regularities between these data. For example, in mineral exploration, the location of a borehole (such as borehole number or coordinates) is a "key," while the associated "value" is the mineral content at different depths within the borehole. By visualizing this data in the electronic map, the mineral distribution of different boreholes can be displayed, and the relationship between mineral content and borehole depth or geographical location can be analyzed.

[0093] Step S62: For each segment of recognized text and image, the classification information is stored in a structured manner according to knowledge rules and semantic entity recognition and relation extraction algorithms.

[0094] The beneficial effects of this invention are as follows: It preprocesses electronic maps of mineral exploration boreholes; it uses a table line recognition algorithm to split elements in the map, obtaining intelligent recognition results of table information and independent text blocks and stratigraphic segments; it uses an optical character recognition algorithm to recognize text in the map, obtaining intelligent recognition results of text information; it uses a deep learning-based image recognition algorithm to recognize lithological symbols in the map, obtaining intelligent recognition results of graphic symbol information; it combines stratigraphic symbol information sorting results with manual review to correct the text recognition results; and it combines knowledge rules and entity semantic recognition to structurally store the information sorting results. Thus, through a process from preprocessing to digitization and then to structuring, the entire process of sorting electronic maps of mineral exploration boreholes can be automated. By identifying tables, text, and maps step by step, all elements of the maps can be effectively utilized. The identification results of one element can effectively improve the identification results of other elements, solving the problems of large map size, many rows and columns in tables, complex data structures, and dense text tables. By combining knowledge rules, deep learning, and manual verification, the semantic analysis and automated sorting of the identified text can effectively improve the accuracy of information sorting and promote the utilization efficiency and integration and sharing of mineral exploration data.

[0095] Based on the same inventive concept, embodiments of the present invention provide a sorting result for element information of electronic maps of mineral exploration boreholes, such as... Figure 7 As shown, where, Figure 7 (b) and (c) are respectively Figure 7 Enlarged view of the two red boxes on the left and right in (a). Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Therefore, it is intended that all changes falling within the meaning and scope of equivalents of the application be incorporated into the invention.

[0096] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined in this invention may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention.

[0097] This invention also provides a mineral exploration borehole electronic map element sorting system, comprising:

[0098] The table splitting module is used to identify table lines in electronic maps of mineral exploration boreholes and split the table into multiple independent text image blocks and multiple independent lithology symbol image blocks according to the content in the cells based on the table lines.

[0099] The element sorting module is used to identify text in multiple independent text image blocks using an optical character recognition algorithm to obtain text elements in the electronic map of mineral exploration boreholes. Based on the lithological symbols appearing in the electronic map of mineral exploration boreholes, a lithological symbol image database is constructed. The images in the lithological symbol image database are input into a convolutional neural network to train the convolutional neural network. Multiple independent lithological symbol image blocks are input into the trained convolutional neural network to perform lithological symbol recognition and obtain different types of lithological symbol elements.

[0100] The present invention also provides a computer device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to perform a method for sorting elements of electronic maps of mineral exploration boreholes.

[0101] The present invention also provides a computer-readable storage medium storing a computer program adapted for loading by a processor to execute a method for sorting elements of electronic maps of mineral exploration boreholes.

[0102] The above-described embodiments are merely preferred embodiments of the present invention, and the scope of protection of the present invention is not limited thereto. Any simple changes or equivalent substitutions of the technical solutions that can be obviously obtained by those skilled in the art within the scope of the technology disclosed in the present invention shall fall within the scope of protection of the present invention.

Claims

1. A method of sorting elements of an electronic mine exploration borehole map, characterized by, The method comprises the following steps: The table line in the mineral exploration drilling electronic map is identified, and the table is correspondingly split into multiple independent text image blocks and multiple independent lithology symbol image blocks based on the table line; The text in the multiple independent text image blocks is recognized using an optical character recognition algorithm to obtain text elements in the mineral exploration drilling electronic map; a lithology symbol image database is constructed according to the lithology symbols appearing in the mineral exploration drilling electronic map, images in the lithology symbol image database are input into a convolutional neural network, the convolutional neural network is trained, the multiple independent lithology symbol image blocks are input into the trained convolutional neural network, and lithology symbol recognition is performed to obtain different types of lithology symbol elements; The table line in the mineral exploration drilling electronic map is identified, and the table is correspondingly split into multiple independent text image blocks and multiple independent lithology symbol image blocks based on the table line, which comprises the following steps: The horizontal lines, vertical lines, number of rows and number of columns in the mineral exploration drilling electronic map are identified, a window is set according to a threshold value, and the horizontal lines and vertical lines of the cells are respectively screened out through the window sliding; The intersection points of the horizontal lines and vertical lines of the cells are merged, the coordinates of the intersection points in the mineral exploration drilling electronic map are calculated, the table structure of the mineral exploration drilling electronic map is restored according to the coordinates, and the table structure of the mineral exploration drilling electronic map is stored in the form of point coordinates and line coordinates; The mineral exploration drilling electronic map is split into multiple independent text image blocks and multiple independent lithology symbol image blocks according to the coordinates.

2. The method of claim 1, wherein, Before the table line in the mineral exploration drilling electronic map is identified, the mineral exploration drilling electronic map is preprocessed, which comprises the following steps: Red seal detection is performed on the mineral exploration drilling electronic map, pixels with a red channel greater than a threshold value are assigned a background value, and the red seal is eliminated; Binaryzation and morphological operation are performed on the mineral exploration drilling electronic map after the red seal is eliminated through a set threshold value, and noise and fuzziness in the mineral exploration drilling electronic map are eliminated; Page direction detection and rotation are performed on the mineral exploration drilling electronic map after the noise and fuzziness are eliminated, and the image is vertically projected, the histogram peak value of the projection image is analyzed, and the mineral exploration drilling electronic map with a direction other than the positive direction is adjusted; The table in the adjusted mineral exploration drilling electronic map is extracted, distorted tables are corrected, and other background areas are cropped.

3. The method of claim 2, wherein, The binaryzation and morphological operation performed on the mineral exploration drilling electronic map after the red seal is eliminated through a set threshold value specifically comprises the following steps: the color image is converted into a gray image, the gray image is binaryzated through a set threshold value, the image features are enhanced through dilation and erosion operation, and the noise and fuzziness in the mineral exploration drilling electronic map are eliminated.

4. The method of claim 1, wherein, After the different types of lithology symbol elements are obtained, the text elements are corrected according to the different types of lithology symbol elements, which comprises the following steps: The text element recognition confidence and the lithology symbol element confidence are obtained, and the correct value of the stratum is determined according to the text element recognition confidence and the lithology symbol element confidence. The identification content of the mineral exploration drilling electronic map is checked by artificial interpretation.

5. The method of claim 4, wherein, The correct value of the stratum is determined according to the text element identification confidence and the lithology symbol element confidence, specifically: whether the text element identification confidence is greater than a set threshold value is determined, if yes, the stratum is considered to be the stratum described in the text element; if the text element identification confidence is not greater than the set threshold value, whether the lithology symbol element confidence is greater than the set threshold value is determined, if yes, the stratum is considered to be the stratum corresponding to the lithology image classification result; if the lithology symbol element confidence is still not greater than the set threshold value, the text recognition result is used as a temporary output, and it is indicated that manual checking is needed during manual checking.

6. A mineral exploration borehole electronic map sheet element sorting system, characterised by, The method comprises the following steps: A table splitting module is used to identify table lines in the mineral exploration drilling electronic map, and the table is split into multiple independent text image blocks and multiple independent lithology symbol image blocks according to the content in the cells based on the table lines; An element sorting module is used to identify the text in the multiple independent text image blocks by using an optical character recognition algorithm to obtain text elements in the mineral exploration drilling electronic map; a lithology symbol image database is constructed according to the lithology symbols appearing in the mineral exploration drilling electronic map, images in the lithology symbol image database are input into a convolutional neural network, the convolutional neural network is trained, the multiple independent lithology symbol image blocks are input into the trained convolutional neural network, and lithology symbol recognition is performed to obtain different types of lithology symbol elements; The table lines in the mineral exploration drilling electronic map are identified, and the table is split into multiple independent text image blocks and lithology symbol image blocks according to the content in the cells based on the table lines, which comprises the following steps: Horizontal lines, vertical lines, row numbers and column numbers in the mineral exploration drilling electronic map are identified, a window is set according to a threshold value, and the horizontal lines and the vertical lines of the cells are respectively screened out by sliding the window; The intersection points of the horizontal lines and the vertical lines of the cells are merged, the coordinates of the intersection points in the mineral exploration drilling electronic map are calculated, the table structure of the mineral exploration drilling electronic map is restored according to the coordinates, and the table structure of the mineral exploration drilling electronic map is stored in the form of point coordinates and line coordinates; The mineral exploration drilling electronic map is split into multiple independent text image blocks and lithology symbol image blocks according to the coordinates.

7. A computer device, comprising: The computer readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute the mineral exploration drilling electronic map element sorting method according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute the mineral exploration drilling electronic map element sorting method according to any one of claims 1-5.

Images