A bottom map-based digital flowchart depiction method and device

Abstract

The application discloses a kind of based on bottom drawing digital flow chart delineation method and device, wherein, method includes the following steps: step S1, obtain flow chart picture file, the flow chart picture file is preprocessed and image enhancement, obtain bottom drawing;Step S2, the bottom drawing is carried out end-to-end graph element identification using deep learning model, and structured bottom drawing data is obtained;Step S3, the graph element of confidence higher than setting value is automatically drawn based on the bottom drawing data;Load the bottom drawing as drawing reference, user carries out flow chart delineation based on the bottom drawing;According to the part graph element that has been completed drawing, the part graph element that is not drawn is intelligently completed, and digital flow chart is obtained.The application improves the digital efficiency and quality of picture format flow chart.

Description

Technical Field

[0001] This invention relates to the field of digital flowchart technology, and in particular to a method and apparatus for drawing digital flowcharts based on a base map. Background Technology

[0002] Current nuclear power systems still contain a large number of flowcharts stored in image formats such as PDF, JPG, and PNG. These drawings cannot be directly associated with equipment, bound to data, or used in intelligent applications. Therefore, it is necessary to convert these image-based flowcharts into editable digital drawings. Currently, the digitization of these image-based flowcharts mainly relies on manual redrawing, which is inefficient and prone to errors. Furthermore, there is a lack of effective methods for handling special cases such as multi-page drawings, blurry drawings, and hand-drawn sketches. There is an urgent need for a method that can quickly, accurately, and intelligently digitize and reconstruct flowcharts to improve the efficiency and quality of image-based flowchart digitization. Summary of the Invention

[0003] In view of this, it is necessary to provide a method and apparatus for drawing digital flowcharts based on a base map, so as to effectively solve the technical problems of low efficiency and low quality in digitizing flowcharts in image format.

[0004] This invention provides a method for drawing digital flowcharts based on a base map, comprising the following steps: Step S1, obtaining a flowchart image file, preprocessing and enhancing the flowchart image file to obtain a base map; Step S2: Use a deep learning model to perform end-to-end primitive recognition on the base map to obtain structured base map data; Step S3: Automatically draw graphic elements with a confidence level higher than a set value based on the base map data; load the base map as a drawing reference, and the user traces the flowchart based on the base map; intelligently complete the undrawn graphic elements according to the already drawn graphic elements to obtain a digital flowchart.

[0005] Preferably, step S1 specifically comprises: The system receives flowchart image files uploaded by users, splits the flowchart image files, and converts them into single images of a unified format; it then uses an adaptive histogram equalization algorithm to enhance the image contrast; a deep learning denoising model is used to remove noise, blemishes, and creases from the image; the Hough transform is used to detect the tilt angle of the image, and automatic rotation correction is performed; a super-resolution reconstruction algorithm is applied to the image to improve its resolution; and finally, the image is converted into a grayscale image to obtain the base image.

[0006] Preferably, step S2 specifically comprises: The primitive recognition includes legend detection, pipeline extraction, text recognition, and connection point detection. Legend detection involves using an object detection algorithm to identify the position, type, and confidence level of various equipment legends in the base image. Pipeline extraction involves using U-net to perform pixel-level semantic segmentation on the image to obtain a pipeline binary mask image. Morphological closing operations are performed on the pipeline binary mask image to eliminate breaks. A thinning algorithm iteratively erodes the pipeline regions in the pipeline binary mask image into single-pixel-width skeleton lines to obtain pipeline primitives. Text recognition involves using an OCR engine to identify the position and content of text in the base image. Connection point detection involves detecting the connection points between the legend and the pipeline based on the legend position and the pipeline skeleton lines through geometric analysis, and establishing legend topological relationships based on these connection points. The results of primitive recognition are stored in structured data, which includes primitive type, coordinates, size, recognition confidence, and text content.

[0007] Preferably, step S3 specifically comprises: Digital elements are automatically generated for elements with a confidence level higher than a set value. These digital elements are then placed on the flowchart after being modified or confirmed by the user. The loaded base image is a semi-transparent base image, which is displayed in the flowchart drawing interface for flowchart tracing; The intelligent completion includes pipeline completion, endpoint connection completion, parallel symmetry completion, and text completion; intelligent completion is performed on undrawn graphic elements to obtain a digital flowchart.

[0008] Preferably, the pipeline completion is as follows: A routing pattern library is established based on historical pipeline data to obtain different routing patterns. The routing vector sequences of some pipelines drawn by the user are recorded. The routing pattern with the highest matching degree to the routing vector sequence is retrieved from the routing pattern library as the prediction basis. Based on the matching routing pattern, the remaining pipeline paths are automatically predicted and completed. After user confirmation, these paths are placed on the flowchart. The endpoint connection completion is as follows: For drawn endpoints and elements to be connected, a path search algorithm is used in the rasterized space. With constraints of shortest path, fewest intersections, and avoidance of identified devices, orthogonal paths are searched as endpoint connections. After user confirmation, these paths are placed on the flowchart. The orthogonal path is depicted on the flowchart; and the direction of the orthogonal path is obtained based on connected component analysis. The identified primitive regions are removed, and the remaining arrow segments are binarized. Regions with more pixels in the connected component are identified as arrow segments, and regions with fewer pixels are identified as arrow tails, thereby completing the flow direction of the orthogonal path. The parallel symmetry completion is as follows: for equidistant parallel pipelines, based on the completed first pipeline, similar structures in adjacent regions are detected by image feature matching, and the remaining parallel relationships are automatically generated; for symmetrical structures, automatic mirror completion is performed after identifying the central axis of symmetry; the text completion is as follows: the identified text primitives are automatically displayed in the corresponding positions.

[0009] Preferably, step S3 further includes: The transparency of the base map is adjusted to 0% when checking the drawing progress; the base map and the digitized flowchart are displayed side-by-side for comparison and verification; after the digitized flowchart is drawn, a quality check is automatically performed based on the base map, including overlay comparison, difference highlighting, connection integrity check, and device code verification; the overlay comparison involves overlaying the digitized flowchart and the base map, and detecting inconsistencies using an image difference algorithm; the difference highlighting involves marking the inconsistencies; the connection integrity check involves detecting unconnected connection points and suspended pipeline endpoints in the digitized flowchart and marking them; the device code verification involves comparing the device codes on the digitized flowchart with the text primitive recognition results, and marking them when they are inconsistent.

[0010] Preferably, this also includes the digital stitching of multiple flowcharts and cross-page links: Each flowchart is digitized based on its own base map. The coordinates of the elements in multiple flowcharts are transformed to a unified coordinate system. One flowchart is used as the main map reference, and the elements in the other flowcharts are translated and aligned with the main map elements according to matching points. Jump symbols and connection marks in each flowchart are detected, and the positions of the jump symbols and connection marks and the target flowchart information are recorded. The two ends of the jump symbols or the pipeline endpoints matched by the connection marks are merged into a single connection point, and cross-drawing breakpoints are eliminated to form a continuous topology, resulting in a complete spliced ​​overall diagram. In the overall diagram, the positions of the jump symbols are marked, and a link relationship is established with the corresponding positions in the associated flowcharts.

[0011] Preferably, the method further includes acquiring multiple flowchart image files, queuing the multiple flowchart image files for processing, and counting the processing progress of each flowchart image file; if the multiple flowchart image files are in the same batch, the recognition parameters of the digitized flowchart are saved as a template, and the template is applied in batches to subsequent flowchart image files.

[0012] Preferably, it also includes recording user operation data during the drawing and correction process, optimizing the primitive recognition model based on the operation data, and training a dedicated model for primitive recognition for specific types of drawings.

[0013] The present invention also provides a basemap-based digital flowchart drawing apparatus, including a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements the basemap-based digital flowchart drawing method.

[0014] Compared with the prior art, the present invention has the following advantages: the present invention uses the processed image as a base map reference to help users quickly and accurately draw flowcharts; at the same time, it identifies the graphic elements on the base map to achieve automatic drawing, which greatly improves drawing efficiency and quality; finally, for the unrecognizable parts, it performs intelligent completion based on the user-drawn parts, further improving drawing efficiency. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart of an embodiment of a digital flowchart drawing method based on a base map provided by the present invention. Detailed Implementation

[0016] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0017] Example 1

[0018] Please see Figure 1 The digital flowchart drawing method based on a base map in this embodiment specifically includes the following steps: Step S1: Obtain the flowchart image file, preprocess and enhance the flowchart image file to obtain the base image; Step S2: Use a deep learning model to perform end-to-end primitive recognition on the base map to obtain structured base map data; Step S3: Automatically draw graphic elements with a confidence level higher than a set value based on the base map data; load the base map as a drawing reference, and the user traces the flowchart based on the base map; intelligently complete the undrawn graphic elements according to the already drawn graphic elements to obtain a digital flowchart.

[0019] This embodiment utilizes technologies such as assisted primitive recognition, automatic multi-image stitching, and intelligent tracing to achieve rapid, high-precision, and intelligent digital reconstruction of image flowcharts, significantly improving the efficiency and quality of image flowchart digitization.

[0020] The following provides a detailed explanation of each step.

[0021] Step S1: Image Preprocessing and Enhancement. Receive flowchart image files uploaded by the user; supported formats include PDF, JPG, PNG, TIFF, etc. Perform multi-dimensional preprocessing on the images, including format conversion and image enhancement. Format conversion: PDF files are split page by page and converted to high-resolution images; in this embodiment, the default is 300 DPI. Image enhancement: The adaptive histogram equalization algorithm CLAHE is used to enhance contrast and improve the readability of faded drawings; a deep learning denoising model, such as DnCNN, is used to remove noise, blemishes, and creases from scanned documents; the Hough transform is used to detect the image tilt angle and automatically rotate and correct it to ensure the frame is level; a super-resolution reconstruction algorithm, such as SRGAN, is applied to blurry images to improve resolution to a recognizable level; color space conversion: color images are converted to grayscale images for easier subsequent recognition and processing.

[0022] Step S2: Intelligent Primitive Recognition and Extraction. A deep learning model is used to perform end-to-end primitive recognition on the preprocessed image, including legend detection, pipeline extraction, text recognition, and connection point detection. Legend detection uses object detection algorithms, such as YOLOv8, to identify the location, type, and confidence level of various equipment legends in the base image. Equipment legends include those for valves, pumps, instruments, heat exchangers, etc. Pipeline extraction uses U-Net for pixel-level semantic segmentation of the image, outputting a binary mask image of the pipeline. After morphological closing operations to eliminate breaks, the Zhang-Suen thinning algorithm iteratively erodes the pipeline area into a skeleton line with a single pixel width. The skeleton pixels are traversed to identify endpoints and bifurcation points, and each segment is traced and converted into multi-segment line vector coordinates. Text recognition uses an OCR engine, such as PaddleOCR, to recognize equipment codes, pipe numbers, and annotation text in the drawing, and records the text position and content. Connection point detection, based on the legend position and pipeline skeleton, uses geometric analysis to detect the connection points between the legend and the pipeline, initially constructing topological relationships. The recognition results are stored in structured data, including primitive type, coordinates, size, recognition confidence, and text content.

[0023] Step S3: AI-assisted tracing and semi-automatic completion. The pre-processed image is loaded as a semi-transparent background. In this embodiment, the default background transparency is 40%, which is displayed in the drawing interface. The system provides various assisted tracing functions based on the recognition results, including one-click generation, semi-automatic tracing, and intelligent completion. One-click generation: For high-confidence recognized elements, such as those with a confidence level > 90%, the system automatically generates digital elements, which the user can directly place after confirmation. Semi-automatic tracing: For elements recognized with medium confidence, the system displays suggested positions and types with dashed boxes, which the user can drag to confirm or manually correct. Intelligent path completion: After the user traces part of the pipeline, the system automatically predicts and completes the remaining path based on the routing pattern of the traced portion, displaying suggestions as gray dashed lines. The system generates the final path after user confirmation. Routing pattern learning: The system records the routing vector sequence of the traced pipeline segments, establishing a routing pattern library. For incomplete sections, it retrieves the routing pattern with the highest matching degree from the library as the prediction basis. Endpoint connectivity analysis: For the traced endpoint A and the endpoint B of the device or pipeline to be connected, an improved A-plane path search algorithm is used in the rasterized space, with constraints of shortest path, fewest intersections, and avoidance of identified devices, to search for orthogonal paths. Predicted path generation: The search results are displayed as gray dashed lines. If multiple equally optimal paths exist, the predicted path is then generated. The system provides a side-by-side display for user selection; direction inference, drawing on connected component analysis, removes identified primitive regions and binarizes the remaining arrow segments. It detects regions with more pixels in the connected component as the arrowhead (flowing towards the target) and regions with fewer pixels as the arrowtail (flowing towards the starting point), automatically determining the flow direction of the completed pipeline; parallel and symmetrical completion: for equidistant parallel pipelines, after the user completes the first one, the system detects similar structures in adjacent regions through image feature matching and automatically suggests generating the remaining parallel pipelines; for symmetrical structures, it automatically mirrors and completes the structure after identifying the central axis of symmetry; automatic text filling: OCR-recognized text is automatically displayed in the corresponding position, and the user can click to confirm, at which point the system automatically fills in primitive attributes, such as device codes.

[0024] Specifically, this embodiment also includes automatic multi-drawing stitching and cross-page linking. For multi-page drawings, such as large systems split into multiple PDF flowcharts, the system automatically identifies and stitches them together. Boundary recognition detects jump characters (OPCs) or connection markers on the edges of the drawings, recording their positions and target drawing information. Feature matching uses image feature matching algorithms, such as SIFT and ORB, to find matching areas in adjacent drawings, such as pipeline endpoints and equipment outlines, for drawings without explicit jump characters, such as hand-drawn drawings. Coordinate normalization unifies the coordinates of elements in multiple drawings to the same coordinate system, using the main drawing, such as the first page, as the reference, and aligning other drawings based on matching points. Connection point merging merges the two ends of jump characters or matching pipeline endpoints into a single connection point, eliminating cross-drawing breakpoints and forming a continuous topology. Cross-page jump preview displays a special marker at the original jump character position in the stitched overall drawing; clicking it quickly locates the corresponding position in the associated drawing.

[0025] Specifically, this embodiment also includes base map guidance and quality verification. During the tracing process, the base map is always used as a reference, supporting interactive features such as transparency adjustment and split-screen comparison. For transparency adjustment, users can slide a slider at any time to adjust the base map transparency from 0% to 100%. At 0%, the base map is completely hidden, facilitating progress monitoring. For split-screen comparison, it supports left-right or top-bottom split-screen displays, simultaneously showing the digital drawing and the original base map for easy manual verification. After tracing is complete, the system automatically performs quality verification, including overlay comparison, difference highlighting, connection integrity checks, and device code verification. Overlay comparison overlays the digital drawing with the original base map, using image difference algorithms to detect inconsistencies such as missing elements, positional offsets, and connection errors. Difference highlighting marks inconsistencies with red borders or semi-transparent red overlays, allowing users to confirm or ignore them. Connection integrity checks automatically detect unconnected connection points and suspended pipeline endpoints, highlighting them with yellow warning signs. Device code verification compares the device codes on the digital drawing with the OCR recognition results, highlighting inconsistencies.

[0026] Specifically, this embodiment also includes batch processing and template reuse. It supports batch uploading of multiple images, with the system automatically queuing them for processing. Queue management displays the processing progress of each drawing, including preprocessing, recognition, tracing, and pending verification. Batch stitching automatically identifies the relationships between multiple drawings from the same system and performs batch stitching. Template reuse allows multiple drawings with consistent frame and legend styles to have their recognition parameters (legend template, line type rules) saved as templates for batch application to subsequent drawings if the frame and legend styles are consistent.

[0027] Specifically, this embodiment also includes manual feedback and model optimization. User actions during the tracing and correction process are recorded for continuous optimization of the recognition model. Correction records, including manually corrected elements such as adjusting positions, modifying types, and supplementing omissions, are recorded and used as training data. Incremental model training involves periodically collecting correction data to incrementally train the recognition model, improving accuracy. Personalized models can be trained for specific power plants or drawing styles, adapting to special legends and annotation habits.

[0028] Compared with existing technologies, this embodiment has the following beneficial effects: Significantly improved efficiency: the auxiliary element recognition and intelligent completion functions greatly shorten the digitization time of a single drawing, and the efficiency of batch processing is significantly improved; Guaranteed accuracy: the combination of intelligent recognition and manual correction results in extremely high digitization accuracy, meeting the high safety requirements of the nuclear power field; Seamless cross-page splicing: automatic recognition and jump matching of features eliminates cross-drawing breakpoints, forming a continuous topology and avoiding manual splicing errors; Activation of existing assets: tens of thousands of existing drawings from various nuclear power plants can be quickly converted into structured digital assets through this method; Continuous evolution: the human feedback loop continuously optimizes the recognition model, becoming more accurate with use and adapting to various drawing styles; Low barrier to entry: even personnel unfamiliar with drawing standards can quickly complete digitization work with AI assistance.

[0029] The following section provides a more detailed explanation of this embodiment, using the digitization of RCV system drawings for nuclear power plants as an example.

[0030] Scenario 1: Digitization of a single high-definition PDF.

[0031] The user uploads "RCV_PID_Rev5.pdf", and the system automatically splits it page by page, with page 3 being the main image. Image enhancement is performed during preprocessing, including contrast improvement and noise reduction, and tilt detection is performed. Deep learning model recognition identifies 23 valves, 5 pumps, and 18 instruments, with an average confidence level of 95% for valves, 92% for pumps, and 88% for instruments. Pipeline extraction generates a vector skeleton, identifying 47 connection points. OCR identifies 25 equipment codes with a confidence level >90%, including "RCV001VP" and "RCV002PO". The system automatically generates high-confidence primitives (>90%), displayed with green borders. Medium-confidence primitives are displayed with yellow dashed borders; these are generated after the user drags and confirms their positions. During the user's tracing process, the system detects a pipeline interruption and automatically displays a suggested connection path with gray dashed lines; this is generated after user confirmation. After tracing, quality checks reveal a valve position that deviates from the original image by 2mm, highlighted in red. The user manually adjusts this and re-verifies, and the tracing passes. The entire process took 25 minutes to generate a structured digital drawing containing all equipment, pipelines, and topology relationships.

[0032] Scenario 2: Assembling multiple pages of drawings.

[0033] The user uploaded three PDF drawings to the RCV system: RCV-101, RCV-102, and RCV-103. The system detected jump marks between the drawings. After recognition, the system automatically stitched the three drawings into a single master drawing based on the jump mark information. Connection markers were displayed at the jump mark locations; clicking on them highlighted the corresponding entry point. The user checked the stitching effect, and after confirming it was correct, saved the master drawing while also keeping independent versions of each sub-drawing.

[0034] Scenario 3: Processing old scanned documents The user uploaded a scanned copy of an RCP system blueprint from 1995. The scan was low-resolution, wrinkled, and partially faded. In the preprocessing stage, super-resolution reconstruction was applied, increasing the resolution to 300 DPI; a denoising model removed wrinkles and shadows; and contrast enhancement restored faded lines. The recognition model had low confidence, averaging 70%, so the system started in semi-automatic tracing mode with 30% background transparency. The user manually traced the main pipelines, and the system automatically predicted and completed the branch pipelines based on the traced portions. The user confirmed the completion. Text recognition failed; the user manually entered the device code, and the system recorded the correction data for subsequent optimization.

[0035] Scenario 4: Batch processing and template application.

[0036] An administrator needs to digitize 30 drawings from a certain system. After batch uploading, the system automatically queues them for processing. Once the first drawing is processed, the administrator saves the recognition parameters as a template, which includes legend mappings, line type rules, etc. Subsequent drawings automatically apply the template, improving recognition accuracy. After batch processing is complete, the system generates a summary report listing the processing time, automatic recognition rate, and number of elements requiring manual intervention for each drawing.

[0037] Example 2

[0038] This embodiment provides a basemap-based digital flowchart drawing device, including a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, it implements the basemap-based digital flowchart drawing method described in Embodiment 1.

[0039] The basemap-based digital flowchart drawing device provided in this embodiment is used to implement the basemap-based digital flowchart drawing method. Therefore, the basemap-based digital flowchart drawing device also possesses the technical effects of the basemap-based digital flowchart drawing method, which will not be described in detail here.

[0040] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of the present invention.

Claims

1. A method for depicting digital flowcharts based on a base map, characterized in that, Includes the following steps: Step S1: Obtain the flowchart image file, preprocess and enhance the flowchart image file to obtain the base image; Step S2: Use a deep learning model to perform end-to-end primitive recognition on the base map to obtain structured base map data; Step S3: Automatically draw graphic elements with a confidence level higher than a set value based on the base map data; load the base map as a drawing reference, and the user traces the flowchart based on the base map; intelligently complete the undrawn graphic elements according to the already drawn graphic elements to obtain a digital flowchart.

2. The method for depicting digital flowcharts based on a base map according to claim 1, characterized in that, Step S1 specifically involves: The system receives flowchart image files uploaded by users, splits the flowchart image files, and converts them into single images of a unified format; it then uses an adaptive histogram equalization algorithm to enhance the image contrast; a deep learning denoising model is used to remove noise, blemishes, and creases from the image; the Hough transform is used to detect the tilt angle of the image, and automatic rotation correction is performed; a super-resolution reconstruction algorithm is applied to the image to improve its resolution; and finally, the image is converted into a grayscale image to obtain the base image.

3. The method for depicting digital flowcharts based on a base map according to claim 1, characterized in that, Step S2 specifically involves: The primitive recognition includes legend detection, pipeline extraction, text recognition, and connection point detection. Legend detection involves using an object detection algorithm to identify the position, type, and confidence level of various equipment legends in the base image. Pipeline extraction involves using U-net to perform pixel-level semantic segmentation on the image to obtain a pipeline binary mask image. Morphological closing operations are performed on the pipeline binary mask image to eliminate breaks. A thinning algorithm iteratively erodes the pipeline regions in the pipeline binary mask image into single-pixel-width skeleton lines to obtain pipeline primitives. Text recognition involves using an OCR engine to identify the position and content of text in the base image. Connection point detection involves detecting the connection points between the legend and the pipeline based on the legend position and the pipeline skeleton lines through geometric analysis, and establishing legend topological relationships based on these connection points. The results of primitive recognition are stored in structured data, which includes primitive type, coordinates, size, recognition confidence, and text content.

4. The method for depicting digital flowcharts based on a base map according to claim 1, characterized in that, Step S3 specifically involves: Digital elements are automatically generated for elements with a confidence level higher than a set value. These digital elements are then placed on the flowchart after being modified or confirmed by the user. The loaded base image is a semi-transparent base image, which is displayed in the flowchart drawing interface for flowchart tracing; The intelligent completion includes pipeline completion, endpoint connection completion, parallel symmetry completion, and text completion; intelligent completion is performed on undrawn graphic elements to obtain a digital flowchart.

5. The method for depicting digital flowcharts based on a base map according to claim 4, characterized in that, The pipeline completion process involves: establishing a routing pattern library based on historical pipeline data to obtain different routing patterns; recording the routing vector sequences of some pipelines drawn by the user; retrieving the routing pattern with the highest matching degree from the routing pattern library as the prediction basis; automatically predicting and completing the remaining pipeline paths based on the matching routing pattern; and placing them on the flowchart after user confirmation. The endpoint connection completion process involves: for drawn endpoints and elements to be connected, using a path search algorithm in the rasterized space, with constraints of shortest path, fewest intersections, and avoidance of identified devices, searching for orthogonal paths as endpoint connections; and drawing the orthogonal paths on the flowchart after user confirmation. The direction of the orthogonal path is obtained based on connected component analysis. The identified primitive regions are removed, and the remaining arrow segments are binarized. Regions with more pixels in the connected component are identified as arrow segments, and regions with fewer pixels are identified as arrow tails, thereby completing the flow direction of the orthogonal path. The parallel symmetry completion is as follows: for equidistant parallel pipelines, based on the completed first pipeline, similar structures in adjacent regions are detected by image feature matching, and the remaining parallel relationships are automatically generated; for symmetrical structures, automatic mirror completion is performed after identifying the central axis of symmetry; the text completion is as follows: the identified text primitives are automatically displayed in the corresponding positions.

6. The method for depicting digital flowcharts based on a base map according to claim 1, characterized in that, Step S3 further includes: The transparency of the base map is adjusted to 0% when checking the drawing progress; the base map and the digitized flowchart are displayed side-by-side for comparison and verification; after the digitized flowchart is drawn, a quality check is automatically performed based on the base map, including overlay comparison, difference highlighting, connection integrity check, and device code verification; the overlay comparison involves overlaying the digitized flowchart and the base map, and detecting inconsistencies using an image difference algorithm; the difference highlighting involves marking the inconsistencies; the connection integrity check involves detecting unconnected connection points and suspended pipeline endpoints in the digitized flowchart and marking them; the device code verification involves comparing the device codes on the digitized flowchart with the text primitive recognition results, and marking them when they are inconsistent.

7. The method for depicting digital flowcharts based on a base map according to claim 1, characterized in that, It also includes the digital stitching of multiple flowcharts and cross-page links: The flowcharts are digitized based on their respective base maps. The coordinates of the elements in the multiple flowcharts are transformed to a unified coordinate system. One flowchart is used as the base map, and the elements in the other flowcharts are translated and aligned with the elements in the base map according to the matching points. Detect jump characters and connection marks in each of the flowcharts, and record the positions of the jump characters and connection marks and the target flowchart information; The two ends of the jump symbol or the pipeline endpoints matched by the connection mark are merged into a single connection point, and cross-drawing breakpoints are eliminated to form a continuous topology, resulting in a complete spliced ​​overall drawing. In the overall drawing, the position of the jump symbol is marked, and a link relationship is established with the corresponding position in the associated flowchart.

8. The method for depicting digital flowcharts based on a base map according to claim 1, characterized in that, It also includes acquiring multiple flowchart image files, queuing and processing the multiple flowchart image files, and calculating the processing progress of each flowchart image file; If multiple flowchart image files are in the same batch, the recognition parameters of the digitized flowcharts are saved as templates, and the templates are applied in batches to subsequent flowchart image files.

9. The method for depicting digital flowcharts based on a base map according to claim 1, characterized in that, It also includes recording user operation data during the drawing and correction process, optimizing the primitive recognition model based on the operation data, and training a dedicated model for primitive recognition for specific types of drawings.

10. A digital flowchart drawing device based on a base map, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, implements the basemap-based digital flowchart drawing method as described in any one of claims 1-9.

Images