Key information extraction method and system for engineering drawing detailed statement

Abstract

The invention discloses a key information extraction method and system for an engineering drawing detailed statement, and the method comprises the steps: configuring rotation angles and corresponding perspective transformation strategies for different engineering drawing detailed statement image samples, constructing an automatic rotation correction model for correcting the direction deviation between the orientation of the target engineering drawing detail list image and the orientation of the standard engineering drawing detail list image sample; dividing different image samples into target drawings and non-target drawings, and constructing a non-target drawing filtering model for screening the target drawings; configuring a header region mark and a body region mark for different image samples respectively, and constructing a target detection model for identifying a key information region; and correcting the direction deviation between the direction of the current target image and the direction of the standard image sample by using each model, screening a target drawing, and further identifying a key information region to extract key information. According to the method, efficient conversion of the engineering drawing detailed statement from the unstructured image to the structured data is realized.

Description

Technical Field

[0001] This invention belongs to the field of image information recognition technology, and in particular relates to a method and system for extracting key information from engineering drawing detail tables. Background Technology

[0002] In modern equipment manufacturing, engineering drawing schedules serve as key process documents, carrying many important parameters.

[0003] Traditional information extraction methods rely heavily on manual image interpretation, which is inefficient and prone to errors, making it difficult to meet the high requirements of intelligent manufacturing for data-driven and process automation.

[0004] In recent years, although some studies have attempted to apply OCR or object detection technologies to automate the parsing of engineering drawing detail tables, they still face many challenges in practical applications. For example, engineering drawing detail tables generally suffer from diverse table styles, inconsistent layout structures, and mixed languages, which limits the generalization ability of the models; the models have low recognition accuracy for complex layouts and find it difficult to accurately restore the logical structure of the tables; most methods rely on direct calls to general OCR toolchains and lack customized processing mechanisms for the characteristics of process documents. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a method for extracting key information from engineering drawing detail tables, comprising: configuring rotation angles and corresponding perspective transformation strategies for different engineering drawing detail table image samples, constructing an automatic rotation correction model to correct the directional deviation between the orientation of the target engineering drawing detail table image and the orientation of the standard engineering drawing detail table image sample; dividing different image samples into target drawings and non-target drawings, constructing a non-target drawing filtering model to screen target drawings; configuring header area markers and body area markers for the different image samples, constructing a target detection model to identify key information regions; using each model to correct the directional deviation between the orientation of the current target image and the orientation of the standard image sample, and screening target drawings to further identify key information regions in order to extract key information.

[0006] Preferably, the key information includes text content and its spatial location information, wherein the step of extracting key information includes: using the PPOCRv5 model to perform multilingual text detection and recognition within the identified key information area, thereby obtaining the key information.

[0007] Preferably, after extracting key information, the key information extraction method further includes: using a post-processing mechanism based on regular expressions to perform semantic correction and format standardization on the extracted key information.

[0008] Preferably, the step of extracting key information further includes: constructing a table to accommodate the current text content, and then mapping the current text content to the corresponding cells in the table according to coordinate data that matches the current spatial location information; after completing the mapping, aligning the text content in the header area and the text content in the body area.

[0009] Preferably, the process of constructing a table to accommodate the current text content includes: using gradient analysis or the elbow rule to analyze the distribution density of rows and columns in the current text content in order to determine the number of rows and columns of the table.

[0010] Preferably, the step of configuring rotation angles and corresponding perspective transformation strategies for different engineering drawing detail table image samples includes: if the directional deviation between the orientation of the image sample and the orientation of the standard image sample is greater than or equal to 90°, then first determine the rotation angle that will allow the image sample to be rotated to satisfy the directional deviation being less than 90° and being an integer multiple of 90°, and then formulate the perspective transformation strategy for the new directional deviation obtained after rotating according to the determined rotation angle; or, if the directional deviation between the orientation of the image sample and the orientation of the standard image sample is less than 90°, then directly formulate the perspective transformation strategy for the current directional deviation.

[0011] Preferably, the automatic rotation correction model, the non-target drawing filtering model, and the model are all obtained based on training on the corresponding preset models.

[0012] Preferably, the non-target drawings include, but are not limited to, drawing catalogs and borrowed materials.

[0013] This invention also provides a computer-readable storage medium comprising a series of instructions for performing method steps for extracting key information from engineering drawing schedules.

[0014] On the other hand, embodiments of the present invention also provide a key information extraction system for engineering drawing detail tables. The key information extraction system includes the following modules: a first model construction module, used to configure rotation angles and corresponding perspective transformation strategies for different engineering drawing detail table image samples, constructing an automatic rotation correction model to correct the directional deviation between the orientation of the target engineering drawing detail table image and the orientation of the standard engineering drawing detail table image sample; a second model construction module, used to divide different image samples into target drawings and non-target drawings, constructing a non-target drawing filtering model to screen target drawings; a third model construction module, used to configure header area markers and body area markers for the different image samples, constructing a target detection model to identify key information regions; and a key information extraction module, used to use each model to correct the directional deviation between the orientation of the current target image and the orientation of the standard image sample, screen target drawings, and further identify key information regions to extract key information.

[0015] Compared with the prior art, one or more embodiments of the above solutions may have the following advantages or beneficial effects.

[0016] This invention provides a method and system for extracting key information from engineering drawing detail tables. The method focuses on understanding and parsing semantically dense tabular documents in engineering drawings, such as engineering drawing detail tables. Based on multimodal artificial intelligence, it constructs an intelligent framework driven by visual perception, language understanding, and structural reasoning. It establishes an automatic rotation correction model to correct directional deviations between the orientation of the target engineering drawing detail table image and that of standard engineering drawing detail table image samples; a non-target drawing filtering model to screen target drawings; and a target detection model to identify key information regions. This accurately locates key information regions on the target image for subsequent key information extraction. This invention achieves end-to-end recognition from drawing pixels to process semantics, efficiently converting unstructured images of engineering drawing detail tables into structured data, ensuring the accuracy and reliability of the conversion results.

[0017] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description, claims, and drawings. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0019] Figure 1This is a flowchart illustrating the steps of a method for extracting key information from an engineering drawing detail table according to an embodiment of this application.

[0020] Figure 2 This is a block diagram of a key information extraction system for engineering drawing detail tables according to an embodiment of this application. Detailed Implementation

[0021] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings and examples, so that the process of how the present invention uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly. It should be noted that, as long as there is no conflict, the various embodiments and features in the various embodiments of the present invention can be combined with each other, and the resulting technical solutions are all within the protection scope of the present invention.

[0022] Furthermore, the steps illustrated in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Also, although a logical order is shown in the flowcharts, in some cases the steps shown or described may be performed in a different order than that shown here.

[0023] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments. Unless the context clearly indicates otherwise, the singular forms “a” and “an” as used herein are also intended to include the plural. It should also be understood that the terms “comprising” and / or “including” as used herein specify the presence of the stated features, integers, steps, operations, units, and / or components, without excluding the presence or addition of one or more other features, integers, steps, operations, units, components, and / or combinations thereof.

[0024] In modern equipment manufacturing, engineering drawing schedules serve as key process documents, carrying many important parameters.

[0025] Traditional information extraction methods rely heavily on manual image interpretation, which is inefficient and prone to errors, making it difficult to meet the high requirements of intelligent manufacturing for data-driven and process automation.

[0026] In recent years, although some studies have attempted to apply OCR or object detection technologies to automate the parsing of engineering drawing detail tables, they still face many challenges in practical applications. For example, engineering drawing detail tables generally suffer from diverse table styles, inconsistent layout structures, and mixed languages, which limits the generalization ability of the models; the models have low recognition accuracy for complex layouts and find it difficult to accurately restore the logical structure of the tables; most methods rely on direct calls to general OCR toolchains and lack customized processing mechanisms for the characteristics of process documents.

[0027] Therefore, to address the aforementioned problems, this invention proposes a method and system for extracting key information from engineering drawing detail tables. This key information extraction method focuses on understanding and parsing semantically dense tabular documents in engineering drawings, such as engineering drawing detail tables. Based on multimodal artificial intelligence, it constructs an intelligent framework driven by visual perception, language understanding, and structural reasoning. It establishes an automatic rotation correction model to correct the directional deviation between the orientation of the target engineering drawing detail table image and that of standard engineering drawing detail table image samples; a non-target drawing filtering model to screen target drawings; and a target detection model to identify key information regions. This accurately locates key information regions on the target image for subsequent key information extraction. This invention achieves end-to-end recognition from drawing pixels to process semantics, efficiently converting unstructured images of engineering drawing detail tables into structured data, ensuring the accuracy and reliability of the conversion results.

[0028] Example 1.

[0029] Figure 1 This is a flowchart illustrating the steps of a method for extracting key information from an engineering drawing schedule according to an embodiment of this application. The following refers to... Figure 1 This will explain the steps of this method.

[0030] like Figure 1 As shown, in step S110, rotation angles and corresponding perspective transformation strategies are configured for different engineering drawing detail table image samples to construct an automatic rotation correction model for correcting the directional deviation between the orientation of the target engineering drawing detail table image and the orientation of the standard engineering drawing detail table image sample.

[0031] In practical applications, engineering drawing detail sheet images are mostly electronic documents in formats such as PDF, JPG, and PNG obtained by scanning physical documents like paper. The different relative positions between the scanning device and the scanned object result in images with varying rotation and / or tilt angles. Therefore, this embodiment uses images without rotation or tilt angles as standard engineering drawing detail sheet image samples, and uses engineering drawing detail sheet images other than the target image as image samples. First, it analyzes the directional deviation between the orientation of the image sample and the standard image sample in the horizontal and vertical directions. Based on the analysis results, it configures rotation angles in both the horizontal and vertical directions for different image samples. Next, it analyzes the tilt or perspective distortion state of the image sample compared to the standard image sample after rotating according to the aforementioned rotation angles in the corresponding directions, and configures perspective transformation strategies for different image samples. Based on this, an automatic rotation correction model is constructed to correct the directional deviation between the target image orientation and the standard image sample orientation.

[0032] In the step of configuring rotation angles and corresponding perspective transformation strategies for image samples of different engineering drawing detail sheets, if the directional deviation between the orientation of the image sample and the orientation of the standard image sample is greater than or equal to 90°, then first determine the rotation angle that will allow the image sample to be rotated to satisfy the directional deviation being less than 90° and being an integer multiple of 90°, and then formulate a perspective transformation strategy for the new directional deviation obtained after rotating according to the determined rotation angle; or, if the directional deviation between the orientation of the image sample and the orientation of the standard image sample is less than 90°, then directly formulate a perspective transformation strategy for the current directional deviation.

[0033] In one specific embodiment of this application, when the directional deviation between the orientation of the image sample and the orientation of the standard image sample is greater than or equal to 90°, firstly, it is determined whether the rotation angle that would allow the image sample to be rotated clockwise to satisfy a directional deviation of less than 90° is 90°, 180°, or 270°. Subsequently, after rotating according to the determined rotation angle (i.e., 90°, 180°, or 270°), a new directional deviation is obtained. That is, the rotated image sample has a tilted state or perspective distortion state with a non-orthogonal angle (not an integer multiple of 90° but less than 90°) that matches the new directional deviation. Based on this, a perspective transformation strategy is formulated so that a perspective transformation algorithm can be used to geometrically correct the current tilted state or perspective distortion state.

[0034] Alternatively, when the directional deviation between the orientation of the image sample and the orientation of the standard image sample is less than 90°, a rotation angle of 0° is determined that allows the image sample to be rotated clockwise to satisfy a directional deviation of less than 90°. In other words, without rotation, the image sample will have a tilted or perspective distortion state with a non-orthogonal angle (not an integer multiple of 90° but less than 90°) that matches the current directional deviation. Based on this, a perspective transformation strategy is formulated to use a perspective transformation algorithm to geometrically correct the current tilted or perspective distortion state.

[0035] Furthermore, in step S120, different image samples are divided into target drawings and non-target drawings, and a non-target drawing filtering model is constructed to filter target drawings.

[0036] Specifically, this embodiment analyzes whether an image sample belongs to the target drawing or a non-target drawing, and then, based on the analysis results, classifies different image samples into target drawings and non-target drawings. Accordingly, a non-target drawing filtering model is constructed to filter target drawings.

[0037] In this embodiment of the application, non-target drawings include, but are not limited to: drawing catalogs and borrowed materials.

[0038] Furthermore, in step S130, header region markers and body region markers are configured for different image samples to construct a target detection model for identifying key information regions.

[0039] Specifically, this embodiment analyzes the header and body regions of image samples, and then, based on the analysis results, assigns header and body region labels to different image samples respectively. Based on this, a target detection model is constructed to identify key information regions.

[0040] Furthermore, in step S140, each model is used to correct the directional deviation between the orientation of the current target image and the orientation of the standard image sample, and the target drawing is screened to further identify key information areas in order to extract key information.

[0041] Specifically, this embodiment first uses an automatic rotation correction model to correct the directional deviation between the orientation of the current target image and the orientation of the standard image sample. Then, a non-target drawing filtering model is used to filter target drawings from the image after orientation deviation correction. Finally, a target detection model is used to identify key information regions on the filtered target drawings, awaiting subsequent key information extraction.

[0042] Key information includes text content and its spatial location information. In the step of extracting key information, the PPOCRv5 model is used to perform multilingual text detection and recognition within the identified key information area, thereby obtaining the key information.

[0043] In practical applications, engineering drawing detail sheets often contain multiple languages ​​such as Chinese, English, and German, and there are complex situations where different engineering drawing detail sheets have differences in fonts, font sizes, and layouts. The PPOCRv5 model has powerful multilingual support capabilities and can effectively handle text extraction needs in scenarios with mixed languages, low-quality images, and non-standard layouts. Therefore, this embodiment uses the PPOCRv5 model to perform multilingual text detection and recognition within the identified key information areas, locate the text content bounding boxes and obtain spatial coordinates (i.e., coordinate data matching the current spatial location information in the following text), and parse the text content. Furthermore, combined with pixel segmentation technology, the text content is divided into header area text content and body area text content, thereby forming a structured output that combines semantic information and spatial location. This structured output is the key information extraction result.

[0044] After extracting key information, this invention also employs a post-processing mechanism based on regular expressions to perform semantic correction and format standardization on the extracted key information. For example, for case-sensitivity errors in English fields, automatic correction is performed using predefined rule templates, improving the accuracy of key field extraction.

[0045] Because engineering drawing detail tables have various template styles, traditional information extraction methods struggle to effectively reconstruct structured features. In the key information extraction step, this embodiment also constructs a table to accommodate the current text content. Then, based on coordinate data matching the current spatial location information, the current text content is mapped to the corresponding cells within the table. This transforms the discrete text content into a table structure conforming to engineering drawing format specifications. This mapping method logically links previously scattered text information such as serial numbers, part drawing numbers, part names, associated drawing numbers, associated names, single-set quantity ratios, unit weight, substitute materials, length, width, and height. After mapping, the text content in the header area and the text content in the body area are accurately linked through column name matching and position alignment. This achieves alignment of the text content in the header area and the text content in the body area, avoiding semantic confusion caused by misalignment in traditional information extraction methods, and is particularly suitable for multilingual, multinational engineering drawings. Based on this, the extracted key information is structured and aligned using a predefined process flow BOM template, which can automatically generate output files (such as Excel format) that conform to process management specifications.

[0046] In constructing a table to accommodate the current text content, gradient analysis is used to calculate the density distribution characteristics of the text bounding box in the vertical and horizontal directions, determining the row and column boundary points to obtain the number of rows and columns of the table. Alternatively, the elbow method is used to evaluate the segmentation effect under different row and column number conditions through clustering algorithms, selecting the optimal solution and analyzing the density distribution characteristics of the current text content in the vertical and horizontal directions to determine the number of rows and columns of the table. Therefore, this invention lays a solid foundation for achieving accurate alignment and two-dimensional matrix reproduction of text content and table row and column structures, effectively solving the problem of key information extraction errors caused by differences in table structure.

[0047] The automatic rotation correction model, the non-target drawing filtering model, and the model are all obtained by training on the corresponding preset models, making the extraction of key information more efficient and accurate.

[0048] Example 2.

[0049] This invention also provides a computer-readable storage medium storing at least one instruction that is loaded and executed by a processor to perform the operation of extracting key information from an engineering drawing schedule as described in the above embodiments. For example, the computer-readable storage medium may be a ROM (Read Only Memory), RAM (Random Access Memory), CD-ROM (Compact Disc-Read Only Memory), magnetic tape, floppy disk, or optical data storage device, etc.

[0050] Example 3.

[0051] Based on the key information extraction method for engineering drawing detail tables described in Embodiment 1 above, the present invention also provides a key information extraction system for engineering drawing detail tables (hereinafter referred to as "key information extraction system").

[0052] Figure 2 This is a block diagram of a key information extraction system for engineering drawing detail tables, according to an embodiment of this application. Figure 2 As shown, the key information extraction system of the present invention includes: a first model construction module 21, a second model construction module 22, a third model construction module 23, and a key information extraction module 24. Specifically, the first model construction module 21 is implemented according to the method described in step S110 above, configured to configure rotation angles and corresponding perspective transformation strategies for different engineering drawing detail table image samples, and construct an automatic rotation correction model to correct the directional deviation between the orientation of the target engineering drawing detail table image and the orientation of the standard engineering drawing detail table image sample; the second model construction module 22 is implemented according to the method described in step S120 above, configured to divide different image samples into target drawings and non-target drawings, and construct a non-target drawing filtering model to screen target drawings; the third model construction module 23 is implemented according to the method described in step S130 above, configured to configure header area markers and body area markers for different image samples, and construct a target detection model to identify key information areas; the key information extraction module 24 is implemented according to the method described in step S140 above, configured to use each model to correct the directional deviation between the orientation of the current target image and the orientation of the standard image sample, and screen target drawings to further identify key information areas in order to extract key information.

[0053] This invention proposes a method and system for extracting key information from engineering drawing detail tables. This key information extraction method focuses on understanding and parsing semantically dense tabular documents in engineering drawings, such as engineering drawing detail tables. Based on multimodal artificial intelligence, it constructs an intelligent framework driven by visual perception, language understanding, and structural reasoning. It establishes an automatic rotation correction model to correct directional deviations between the orientation of the target engineering drawing detail table image and that of standard engineering drawing detail table image samples; a non-target drawing filtering model to screen target drawings; and a target detection model to identify key information regions. This accurately locates key information regions on the target image for subsequent key information extraction. This invention achieves end-to-end recognition from drawing pixels to process semantics, efficiently converting unstructured images of engineering drawing detail tables into structured data, ensuring the accuracy and reliability of the conversion results.

[0054] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0055] In the description of this invention, unless otherwise stated, "a plurality of" means two or more; the terms "upper," "lower," "left," "right," "inner," "outer," "front end," "rear end," "head," "tail," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the 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, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0056] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "connected" and "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 based on the specific circumstances.

[0057] It should be understood that the embodiments disclosed herein are not limited to the specific structures, processing steps, or materials disclosed herein, but should be extended to equivalent substitutions of these features as understood by those skilled in the art. It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting.

[0058] The phrase "an embodiment" or "an embodiment" used in this specification means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the invention. Therefore, the phrase "an embodiment" or "an embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment.

[0059] While the embodiments disclosed in this invention are as described above, the content is merely for the purpose of facilitating understanding of the invention and is not intended to limit the invention. Any person skilled in the art to which this invention pertains may make any modifications and changes in form and detail of the implementation without departing from the spirit and scope disclosed herein; however, the scope of patent protection of this invention shall still be determined by the scope defined in the appended claims.

Claims

1. A method for extracting key information from engineering drawing detail tables, characterized in that, include: For different engineering drawing detail table image samples, configure rotation angles and corresponding perspective transformation strategies respectively, and construct an automatic rotation correction model to correct the directional deviation between the orientation of the target engineering drawing detail table image and the orientation of the standard engineering drawing detail table image sample; Different image samples are divided into target drawings and non-target drawings, and a non-target drawing filtering model is constructed to filter target drawings. Configure header region labels and body region labels for the different image samples respectively, and construct a target detection model to identify key information regions; Using various models, the orientation deviation between the current target image orientation and the standard image sample orientation is corrected, and the target drawings are screened to further identify key information areas in order to extract key information.

2. The key information extraction method according to claim 1, characterized in that, The key information includes text content and its spatial location information, wherein the step of extracting key information includes: The PPOCRv5 model is used to perform multilingual text detection and recognition within the identified key information areas, thereby obtaining the key information.

3. The key information extraction method according to claim 2, characterized in that, After extracting the key information, the key information extraction method further includes: A post-processing mechanism based on regular expressions is used to perform semantic correction and format standardization on the extracted key information.

4. The key information extraction method according to claim 2 or 3, characterized in that, The steps for extracting key information also include: A table is constructed to hold the current text content, and then the current text content is mapped to the corresponding cells in the table according to the coordinate data that matches the current spatial location information. After mapping is complete, align the text content of the table header area with the text content of the table body area.

5. The key information extraction method according to claim 4, characterized in that, The process of constructing a table to hold the current text content includes: Gradient analysis or the elbow rule is used to analyze the distribution density of rows and columns in the current text content to determine the number of rows and columns of the table.

6. The key information extraction method according to any one of claims 1 to 5, characterized in that, The steps of configuring rotation angles and corresponding perspective transformation strategies for different engineering drawing detail sheet image samples include: If the directional deviation between the orientation of the image sample and the orientation of the standard image sample is greater than or equal to 90°, then first determine the rotation angle that will allow the image sample to be rotated to satisfy the directional deviation being less than 90° and being an integer multiple of 90°. Then, for the new directional deviation obtained after rotating according to the determined rotation angle, formulate the perspective transformation strategy; or... If the directional deviation between the orientation of the image sample and the orientation of the standard image sample is less than 90°, then the perspective transformation strategy is formulated directly for the current directional deviation.

7. The key information extraction method according to any one of claims 1 to 6, characterized in that, The automatic rotation correction model, the non-target drawing filtering model, and the model are all obtained by training on the corresponding preset models.

8. The key information extraction method according to any one of claims 1 to 7, characterized in that, The non-target drawings include, but are not limited to: drawing catalogs and borrowed materials.

9. A computer-readable storage medium, characterized in that, It includes a series of instructions for performing the steps of the method for extracting key information from a drawing schedule as described in any one of claims 1 to 8.

10. A system for extracting key information from engineering drawing detail tables, characterized in that, The key information extraction system includes the following modules: The first model construction module is used to configure rotation angles and corresponding perspective transformation strategies for different engineering drawing detail table image samples, and to construct an automatic rotation correction model to correct the directional deviation between the orientation of the target engineering drawing detail table image and the orientation of the standard engineering drawing detail table image sample. The second model building module is used to divide different image samples into target drawings and non-target drawings, and to build a non-target drawing filtering model for filtering target drawings. The third model construction module is used to configure header region labels and body region labels for the different image samples respectively, and to construct a target detection model for identifying key information regions. The key information extraction module is used to use various models to correct the directional deviation between the orientation of the current target image and the orientation of the standard image sample, and to screen the target drawing to further identify key information areas in order to extract key information.

Images