Table handwritten information identification method and device and electronic equipment

Abstract

The invention discloses a table handwritten information identification method and device and electronic equipment. The method comprises the following steps: acquiring to-be-recognized table handwritten information; the table structure of the table handwritten information is determined, the table structure is corrected according to a preset table correction rule to obtain a target table structure, and the preset table correction rule is used for correcting abnormal table cells in the table structure based on the rule that adjacent table cell vertexes are shared, the same column of table cells are the same in column width, and the same row of table cells are the same in row height; determining a textbox where text content in the table handwritten information is located, and adjusting the textbox according to a relative position relationship between the textbox and cells in the target table structure to obtain a target textbox; and recognizing the text content in the target textbox to obtain a text recognition result. According to the method and the apparatus, the technical problems of low recognition efficiency of diversified handwritten table information, incapability of adapting to position changes of handwritten fonts inside and outside cells and instable character spacing among the cells in related technologies are solved.

Description

Technical Field

[0001] This application relates to the field of image recognition technology, and more specifically, to a method, apparatus, and electronic device for recognizing handwritten information in a form. Background Technology

[0002] In daily office work and data management, especially in situations where it is necessary to frequently collect and organize personal information, work records, or statistical data, handwritten forms have become a common tool for conveying information. Because original paper documents are prone to loss, take up a lot of space, and are inconvenient to search, information from original paper documents can be entered into electronic systems or spreadsheets to facilitate subsequent retrieval and efficient use of the paper form information.

[0003] However, in related technologies, the digitization of paper documents largely relies on manual recognition, which is inefficient, costly, and prone to errors. In recent years, the application of optical character recognition technology has greatly improved the automation level of paper information entry, but it still faces many challenges when processing handwritten form information.

[0004] For example, most table recognition technologies in related fields focus on parsing printed tables, identifying table boundaries and cell positions through preset table style templates or metadata. This method performs well when processing printed tables, but because it relies on a clearly defined structure, it falls short when dealing with handwritten tables. Handwritten tables lack the rules and consistency of printed tables; cell lines may be blurred or completely missing, and the writing position, size, and direction of text are more free and diverse, leading to significant errors in detecting cell boundaries, determining cell positions, and recognizing the text content within cells.

[0005] Furthermore, when dealing with the digitization of handwritten forms, the recognition methods in related technologies often struggle because they cannot accurately grasp the free-flowing nature of handwritten information. For example, text may extend beyond cell boundaries, be written on multiple lines within the same row, or have extremely small spacing between cells, causing content in different cells to be incorrectly identified as the same text block, resulting in the erroneous merging or separation of information. In addition, the diversity of handwriting styles brings additional recognition difficulties; printed text recognition algorithms often perform poorly in this situation, and may even fail to recognize text at all.

[0006] While some research has attempted to incorporate machine learning and deep learning techniques to improve handwritten form recognition, most methods still require large amounts of labeled data to train the model and are often limited in their generalization capabilities in practical applications, especially when encountering unfamiliar form styles or extreme handwriting situations. This not only increases development and maintenance costs but also limits the widespread application of recognition technologies.

[0007] There is currently no effective solution to the above problems. Summary of the Invention

[0008] This application provides a method, apparatus, and electronic device for recognizing handwritten information in tables, which at least solves the technical problems in the related art of low efficiency in recognizing diverse handwritten table information and inability to adapt to changes in the position of handwritten fonts inside and outside cells and unstable spacing between text in cells.

[0009] According to one aspect of the embodiments of this application, a method for recognizing handwritten information in a table is provided, comprising: acquiring handwritten information in a table to be recognized; determining the table structure of the handwritten information and correcting the table structure according to a preset table correction rule to obtain a target table structure, wherein the table structure is used to reflect the cell information in the handwritten information, and the preset table correction rule corrects abnormal cells in the table structure based on the rules that adjacent cells share vertices, cells in the same column have the same column width, and cells in the same row have the same row height; determining the text box containing the text content in the handwritten information and adjusting the text box according to the relative positional relationship between the text box and the cells in the target table structure to obtain a target text box; recognizing the text content in the target text box to obtain a text recognition result corresponding to the handwritten information in the table.

[0010] Optionally, determining the table structure of the handwritten information includes: recognizing the handwritten information using a table structure recognition model to obtain cell information of the handwritten information, wherein the cell information is a multidimensional array that includes at least the number of cells in the handwritten information, cell boundary coordinates, and cell row and column index coordinates; and determining the table structure based on the cell information.

[0011] Optionally, the table structure is corrected according to preset table correction rules to obtain the target table structure, including: determining the feature matrix corresponding to the table structure; determining the common vertex coordinate distance of adjacent cells in the feature matrix, the column width difference of cells in the same column, and the row height difference of cells in the same row according to preset table correction rules; determining abnormal cells in the feature matrix based on the common vertex distance, column width difference, and row height difference; correcting the boundary coordinates and row and column index coordinates of abnormal cells, and converting the corrected feature matrix into array form to obtain the target table structure.

[0012] Optionally, before determining the feature matrix corresponding to the table structure, the method further includes: determining the target number of columns and the average number of columns in the table structure, wherein the target number of columns is used to represent the number of columns in any row of cells in the table structure; determining the row corresponding to when the target number of columns is greater than the average number of columns as the handwritten part of the table, and correcting the cell row number of the table structure in the handwritten part of the table.

[0013] Optionally, before determining the text box containing the text content in the handwritten information of the table, the method further includes: recognizing the handwritten information of the table through a text recognition model to obtain the text content in the handwritten information of the table; determining the text box information corresponding to the text content, wherein the text box information is a multidimensional array that includes at least the number of text boxes and the coordinates of the text box boundaries.

[0014] Optionally, the text box is adjusted based on the relative positional relationship between the text box and the cells in the target table structure to obtain the target text box, including: determining the intersection between the target cell and the target text box, wherein the target cell is any cell in the handwritten information of the table, the target text box is any text box in the handwritten information of the table, and the intersection is used to reflect the degree of overlap between the target cell and the target text box; if the intersection is greater than a first preset threshold, it is determined that the target text box is located within the target cell; determining the number of text boxes within the target cell, if the number of text boxes is 0, the bounding box of the target cell is determined as the target text box within the target cell; if the number of text boxes is 1, the target text box within the target cell is retained; if the number of text boxes is greater than 1, the target text box within the target cell is adjusted according to the ordinate distance of the center point of the target text box.

[0015] Optionally, adjusting the target text boxes within the target cell based on the y-coordinate distance of the center points of the target text boxes includes: determining the difference in y-coordinates between the center points of the first target text box and the second target text box, wherein the first target text box and the second target text box are any two text boxes within the target cell; merging the first target text box and the second target text box if the difference in y-coordinates is less than a second preset threshold; and sorting the target text boxes and the second target text box according to the size of the y-coordinate values ​​of their center points if the difference in y-coordinates is greater than the second preset threshold, and retaining the sorted first target text box and the second target text box within the target cell.

[0016] Optionally, the method further includes: determining the difference in side length between the first target text box and the target cell; if the difference in side length is less than 0, not adjusting the first target text box; if the difference in side length is greater than 0 and the ratio of the absolute value of the difference in side length to the width of the target cell is less than a third preset threshold, not adjusting the first target text box; and if the difference in side length is greater than 0 and the ratio of the absolute value of the difference in side length to the width of the target cell is greater than or equal to the third preset threshold, determining the side length of the target cell as the side length of the first target text box.

[0017] Optionally, recognizing the text content in the target text box to obtain the text recognition result corresponding to the handwritten information in the table includes: recognizing the target text box using a first recognition algorithm and a second recognition algorithm respectively to obtain the first text content and the second text content, wherein the first recognition algorithm is used to recognize the handwritten information in the handwritten information in the table, and the second recognition algorithm is used to recognize the printed information in the handwritten information in the table; standardizing the first text content and fusing the standardized first text content, the second text content, and the target table structure to obtain the text recognition result corresponding to the handwritten information in the table; and saving the text recognition result as a table file.

[0018] According to another aspect of the embodiments of this application, a table handwritten information recognition device is also provided, comprising: an acquisition module for acquiring table handwritten information to be recognized; a correction module for determining the table structure of the table handwritten information and correcting the table structure according to preset table correction rules to obtain a target table structure, wherein the table structure is used to reflect the cell information in the table handwritten information, and the preset table correction rules correct abnormal cells in the table structure based on the rules that adjacent cells share vertices, cells in the same column have the same column width, and cells in the same row have the same row height; an adjustment module for determining the text box where the text content in the table handwritten information is located and adjusting the text box according to the relative positional relationship between the text box and the cells in the target table structure to obtain a target text box; and a recognition module for recognizing the text content in the target text box to obtain a text recognition result corresponding to the table handwritten information.

[0019] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory and a processor, wherein the memory is used to store program instructions; and the processor is connected to the memory and used to execute the above-described method for recognizing handwritten information in a table.

[0020] According to another aspect of the embodiments of this application, a non-volatile storage medium is also provided, the non-volatile storage medium including a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-described method for recognizing handwritten information in a form by running the computer program.

[0021] According to another aspect of the embodiments of this application, a computer program product is also provided, including computer instructions, which, when executed by a processor, implement the above-described method for recognizing handwritten information in a table.

[0022] In this embodiment, the following steps are taken: First, the handwritten information of the table to be identified is obtained. Then, the table structure of the handwritten information is determined, and the table structure is corrected according to preset table correction rules to obtain a target table structure. The table structure reflects the cell information in the handwritten information. The preset table correction rules correct abnormal cells in the table structure based on rules such as shared vertices between adjacent cells, identical column widths for cells in the same column, and identical row heights for cells in the same row. Next, the text box containing the text content in the handwritten information is determined, and the text box is adjusted according to its relative position to the cells in the target table structure to obtain a target text box. Finally, the text content in the target text box is identified to obtain the text recognition result corresponding to the handwritten information. This achieves the goal of accurately recognizing complex handwritten table information, thereby significantly improving the automation level of information entry and the accuracy of data processing. This solves the technical problems in related technologies, such as low efficiency in recognizing diverse handwritten table information and inability to adapt to changes in the position of handwritten characters inside and outside cells and unstable spacing between characters in cells. Attached Figure Description

[0023] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0024] Figure 1 This is a hardware structure diagram of a computer terminal for implementing a method for recognizing handwritten information in a form, according to an embodiment of this application.

[0025] Figure 2 This is a flowchart of a method for recognizing handwritten information in a form according to an embodiment of this application;

[0026] Figure 3 This is a flowchart of another method for recognizing handwritten information in a form according to an embodiment of this application;

[0027] Figure 4 This is a flowchart of an adaptive correction of a table structure according to an embodiment of this application;

[0028] Figure 5 This is a schematic diagram of cell information in a table structure according to an embodiment of this application;

[0029] Figure 6 This is a flowchart of a text position adaptive correction according to an embodiment of this application;

[0030] Figure 7 This is a structural diagram of a form handwritten information recognition device according to an embodiment of this application. Detailed Implementation

[0031] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0032] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0033] First, some nouns or terms that appear in the explanation of the embodiments of this application shall be interpreted as follows:

[0034] OCR (Optical Character Recognition) is a technology that can convert text in an image into an editable and searchable text format. OCR is widely used in various scenarios, such as ID card information reading, invoice recognition, and manuscript digitization, and can process printed or handwritten text.

[0035] Adaptive table structure correction: This technique adjusts the cell detection results based on the actual layout and content of the table during the table structure recognition process, making them closer to the real structure. Adaptive correction can compensate for errors in cell detection algorithms, especially when processing handwritten tables, correcting cell size and position to ensure the accuracy and completeness of cell information.

[0036] Text position adaptive correction: A technique used in handwritten table recognition to adjust the relative position of text boxes to cells, ensuring that the text within each cell is accurately recognized and located. Text position adaptive correction can handle special cases such as handwritten text extending beyond the cell or line breaks, improving the accuracy and efficiency of text recognition.

[0037] To address the issue of poor accuracy in recognizing handwritten forms in related technologies, this application provides a method for recognizing handwritten information in forms. This method can be implemented in... Figure 1 The computer terminal shown is described below.

[0038] The handwritten information recognition method for forms provided in this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal for implementing a method for recognizing handwritten information in forms is shown. Figure 1 As shown, the computer terminal 10 may include one or more processors (shown as 102a, 102b, ..., 102n in the figure) (the processor may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission module 106 for communication functions connected via wired and / or wireless networks. In addition, it may also include: a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, and a BUS bus. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0039] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be implemented wholly or partially as software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be wholly or partially integrated into any other element in the computer terminal 10. As involved in the embodiments of this application, the data processing circuits serve as processor control (e.g., selection of a variable resistor termination path connected to an interface).

[0040] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the handwritten information recognition method in the form of this application embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the above-mentioned handwritten information recognition method in the form. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor, and these remote memories can be connected to the computer terminal 10 via a network. Examples of the above-mentioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0041] The transmission module 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission module 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission module 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0042] The display may be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10.

[0043] It should be noted here that, in some optional embodiments, the above... Figure 1 The computer terminal shown may include hardware elements (including circuitry), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware and software elements. It should be noted that... Figure 1 This is only one instance of a specific particular instance, and is intended to illustrate the types of components that may exist in the aforementioned computer terminal.

[0044] In the above operating environment, this application provides an embodiment of a method for recognizing handwritten information in a form. It should be noted that the steps shown in the flowchart in 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 flowchart, in some cases, the steps shown or described can be executed in a different order than that shown here.

[0045] Figure 2 This is a flowchart of a method for recognizing handwritten information in a form according to an embodiment of this application, such as... Figure 2As shown, the method includes the following steps:

[0046] Step S202: Obtain the handwritten information of the form to be recognized.

[0047] Step S204: Determine the table structure of the handwritten information in the table, and correct the table structure according to the preset table correction rules to obtain the target table structure. The table structure is used to reflect the cell information in the handwritten information in the table. The preset table correction rules correct abnormal cells in the table structure based on the rules that adjacent cells share vertices, cells in the same column have the same column width, and cells in the same row have the same row height.

[0048] Step S206: Determine the text box containing the text content in the handwritten information of the table, and adjust the text box according to the relative position between the text box and the cell in the target table structure to obtain the target text box.

[0049] Step S208: Recognize the text content in the target text box to obtain the text recognition result corresponding to the handwritten information in the table.

[0050] Through the above steps S202 to S208, the goal of accurately recognizing complex handwritten table information is achieved, thereby greatly improving the level of automation of information entry and the accuracy of data processing. This solves the technical problems in related technologies, such as low efficiency in recognizing diverse handwritten table information and inability to adapt to changes in the position of handwritten characters inside and outside cells and unstable spacing between characters in cells.

[0051] Figure 3 This is a flowchart of another method for recognizing handwritten information in a form according to an embodiment of this application, such as... Figure 3 As shown, the complete process of recognizing handwritten information on paper forms is presented in more detail. The entire process does not require preset formats, but rather uses adaptive correction to ensure the accuracy of table structure and text recognition in complex handwriting scenarios, ultimately achieving batch, efficient and low-cost digitization of paper forms.

[0052] The following combination Figure 3 Regarding the above Figure 2 Steps S202 to S208 are described in detail.

[0053] In step S202 above, the main responsibilities are to acquire input data and perform data preprocessing. The input data includes, but is not limited to, PDF scans or electronic images corresponding to the handwritten information in the form.

[0054] To ensure that all input data meets the requirements of subsequent image analysis algorithms, a standardized format conversion is necessary, such as converting it to a three-channel color image in RGB format. This is because the RGB color space contains richer color information, making it easier for the algorithm to accurately capture and distinguish table lines, text content, and other image features. Furthermore, preprocessed images—i.e., preprocessed handwritten table information—can be obtained through image rotation correction, image denoising, and image enhancement to improve the subsequent algorithm's ability to recognize, analyze, and understand the original image.

[0055] In step S204 above, the main responsibilities are table structure recognition and adaptive correction. The table structure reflects the basic structure of the handwritten table information, i.e., the position and size of the cells. Typically, deep learning or computer vision techniques can be used to predict cell boundaries and generate a preliminary table structure. However, due to the non-standardized nature of handwritten tables, cell detection errors may occur. Therefore, an adaptive correction technique can be used. This technique, based on preset table correction rules (such as shared vertices between adjacent cells, identical column widths for cells in the same column, and identical row heights for cells in the same row), corrects the initially recognized table structure to eliminate abnormal cells during the detection process and ensure the accuracy of the final table structure.

[0056] Optionally, determining the table structure of the handwritten information includes: recognizing the handwritten information using a table structure recognition model to obtain cell information of the handwritten information, wherein the cell information is a multidimensional array that includes at least the number of cells in the handwritten information, cell boundary coordinates, and cell row and column index coordinates; and determining the table structure based on the cell information.

[0057] In this embodiment, a pre-trained table structure recognition model can be used to recognize the preprocessed handwritten table information in step S202, detecting the bounding box of each cell in the image and outputting an n×10 multidimensional array, i.e., cell information, including but not limited to: 1) the number of cells (cell box) n; 2) the cell boundary coordinates (x0, y0, x1, y1, x2, y2, x3, y3), i.e., the physical position coordinates of the four vertices of the cell (top left, top right, bottom right, bottom left); 3) the table row and column coordinates (r, c) obtained by sorting the top left corner coordinates (x0, y0) of the cells from top to bottom and from left to right. Subsequently, the skeleton of the table is constructed using the recognized cell information, including the distribution of cells, the detection of merged cells, and the determination of cell size, thereby obtaining a preliminary table structure of the handwritten table information.

[0058] Furthermore, the table structure is corrected according to preset table correction rules to obtain the target table structure, including: determining the feature matrix corresponding to the table structure; determining the common vertex coordinate distance of adjacent cells in the feature matrix, the column width difference of cells in the same column, and the row height difference of cells in the same row according to the preset table correction rules; determining abnormal cells in the feature matrix based on the common vertex distance, column width difference, and row height difference; correcting the boundary coordinates and row and column index coordinates of abnormal cells, and converting the corrected feature matrix into array form to obtain the target table structure.

[0059] In this embodiment, the table structure adaptive correction mechanism, from a statistical perspective, leverages the inherent relationships between cells and the advantages of matrix operations to efficiently and accurately correct errors in table structure recognition. The specific process is as follows: Figure 4 As shown, the analysis is as follows:

[0060] 1. Select the handwritten information section in the table.

[0061] In this embodiment of the application, the selection of handwritten / hand-filled information is based on the characteristics of the table structure, aiming to improve the accuracy and robustness of table recognition. This includes: determining the target number of columns and the average number of columns in the table structure, wherein the target number of columns is used to represent the number of columns in any row of cells in the table structure; determining the row corresponding to the target number of columns being greater than the average number of columns as the handwritten part of the table, and correcting the cell row number of the table structure in the handwritten part of the table.

[0062] Specifically, in the table recognition process, the number of columns in each row, c_cnt (i.e., the target number of columns mentioned above), can first be calculated, and its average value, cnt_mean (i.e., the average number of columns mentioned above), can be determined to better understand and recognize the true structure of the table. Tables, especially handwritten tables, may contain various row styles, such as header rows, data rows, or rows with merged cells. Calculating the number of columns in each row provides insight into the specific layout of the table, and then by calculating the average value cnt_mean, the typical structure of a data row can be estimated, i.e., how many columns it typically contains.

[0063] Secondly, in large tables, handwritten data is often filled in rows with many columns. Therefore, by comparing the number of cells in a row with the average number of columns, rows with more columns than `cnt_mean` can be selected as the handwritten information section. This selection effectively distinguishes the rows containing handwritten data from header rows or other special rows that may contain fewer columns, thus avoiding the misinterpretation of header rows as data rows in subsequent processing, which could lead to unnecessary errors or confusion.

[0064] Finally, the cell row number 'r' is updated to ensure that the logical position of each cell matches its actual position. During the recognition process, errors in the cell detection algorithm may lead to inaccurate cell position information, especially in row number recognition. By recalculating the row numbers in the rows containing the handwritten information, these potential errors can be corrected, ensuring that each cell is assigned to the correct row. This is crucial for subsequent cell correction, text position correction, and the final generation of the Excel spreadsheet, as it directly affects whether the cell content is correctly placed in its corresponding position within the spreadsheet.

[0065] 2. Reconstruct the cell matrix according to row and column coordinates.

[0066] Based on the table structure obtained from the initial detection, the number of rows (r_max) and columns (c_max) of the entire table are calculated. Then, a three-dimensional matrix A is initialized, which is the feature matrix corresponding to the table structure mentioned above. Its dimension is r_max×c_max×10, where the 10 dimensions represent the coordinates of the 8 vertices and the 2 row and column index coordinates of each cell.

[0067] 3. Verification and error detection for shared vertices, identical column widths, and identical row heights.

[0068] like Figure 5 As shown, in a correct table structure, table vertices are shared by adjacent cells. For example: r i-1 The coordinates of the bottom left corner of the row cell (x1, y1) and r i The top-left corner coordinates (x0, y0) of the row cells are the same; r i-1 The bottom right corner coordinates (x2, y2) of the row cell and r i The top-right corner coordinates (x3, y3) of the row cells are the same.

[0069] Therefore, pre-defined table correction rules can be used to identify abnormal / erroneous cells, such as whether adjacent cells share a vertex, whether the column width is the same, and whether the row height is the same. Specifically, in matrix A, the distance between the vertices of adjacent cells (d1), the difference between the width of the top and bottom edges of a cell (d2), and the difference between the height of the left and right edges of a cell (d3) are calculated. Then, the row and column coordinates of all erroneous cells can be obtained based on the set of indices of elements greater than 0 in d1, d2, and d3.

[0070] 4. Correction based on the position of adjacent cells.

[0071] Finally, based on the coordinate information of the surrounding correct cells, the vertex positions and row and column numbers of the abnormal cells marked as errors are dynamically corrected, thereby making the entire table structure more accurate and complete. The corrected cell information is then converted back into a unified n×10-dimensional array format to obtain the target table structure.

[0072] This series of operations completed the adaptive correction of the table structure, ensuring that the table structure closely matched the original handwritten table, and providing accurate cell boundary information for further text location detection and text content recognition.

[0073] In step S206 above, the main tasks are text position detection and adaptive text position correction. Text box detection also relies on advanced image recognition technology. However, the position, size, and orientation of handwritten text vary greatly, which may cause text boxes to cover multiple cells or fail to completely encompass the text within a cell. Therefore, the position of the text box can be adaptively adjusted based on its relative position to the cells in the target table structure, ensuring that the text within each cell is accurately framed and extracted, reducing the chance of information being incorrectly merged or omitted.

[0074] Optionally, text box detection includes: recognizing the handwritten information in the table using a text recognition model to obtain the text content in the handwritten information in the table; and determining the text box information corresponding to the text content, wherein the text box information is a multidimensional array that includes at least the number of text boxes and the coordinates of the text box boundaries.

[0075] In this embodiment of the application, the pre-trained text location detection model can be used to identify the pre-processed handwritten information in the table in step S202, detect the text box where the text content is located in the image, and output an m×8 multidimensional array, i.e., text box information, including but not limited to: 1) the number of text boxes m; 2) the boundary coordinates of the text boxes (x0, y0, x1, y1, x2, y2, x3, y3), i.e., the physical position coordinates of the four vertices of the text box, namely the upper left, upper right, lower right, and lower left.

[0076] Further, the text box is adjusted based on the relative positional relationship between the text box and the cells in the target table structure to obtain the target text box, including: determining the intersection between the target cell and the target text box, wherein the target cell is any cell in the handwritten information of the table, and the target text box is any text box in the handwritten information of the table, and the intersection is used to reflect the degree of overlap between the target cell and the target text box; if the intersection is greater than a first preset threshold, it is determined that the target text box is located within the target cell; determining the number of text boxes within the target cell, if the number of text boxes is 0, the bounding box of the target cell is determined as the target text box within the target cell; if the number of text boxes is 1, the target text box within the target cell is retained; if the number of text boxes is greater than 1, the target text box within the target cell is adjusted according to the ordinate distance of the center point of the target text box.

[0077] In this embodiment, the text box adjustment relies on the relative position of the text box to the cells in the target table structure. Through precise intersection calculation and center point ordinate analysis, it ensures that the handwritten information within each cell is accurately framed and recognized, avoiding information confusion or omission. The specific process is as follows: Figure 6 As shown, the analysis is as follows:

[0078] 1. Intersection calculation and location determination.

[0079] First, for any cell box i (the target cell), calculate its intersection with each of the m text boxes. The intersection reflects the degree of overlap between cell box i and text box i (the target text box).

[0080] If the intersection area exceeds the first preset threshold, then the text box i is considered to be located within the corresponding cellbox i.

[0081] 2. Analysis and adjustment of the number of text boxes.

[0082] Determine the number of text boxes whose intersection exceeds a first preset threshold, and perform the following processing based on the number of text boxes:

[0083] 1) When the number of detected text boxes in a cell is 0, it means that some information may have been missed in the text detection process. In this case, the bounding box of cell box i will be directly used as text box i for subsequent OCR recognition, and the missed text content will be supplemented.

[0084] 2) When the number of text boxes detected in a cell is 1, text box i is regarded as the entire carrier of the handwritten information in the cell, and no further adjustment is required. It is directly retained for subsequent processing.

[0085] 3) When the number of detected text boxes in a cell is greater than 1, it indicates the presence of multiple segments of handwritten text or text wrapping. In this case, it is necessary to further analyze the ordinates of the center points of all text boxes i within cell i and adjust the existence of the text boxes in the cell, including: determining the difference in ordinates of the center points of the first target text box and the second target text box, where the first target text box and the second target text box are any two text boxes within the target cell; merging the first target text box and the second target text box if the difference in ordinates is less than a second preset threshold; and sorting the target text boxes and the second target text box according to the size of their center point ordinates if the difference in ordinates is greater than the second preset threshold, and retaining the sorted first target text box and the second target text box within the target cell.

[0086] Specifically, two adjacent text boxes i are randomly selected, such as the first target text box and the second target text box. When the difference in the ordinates of the center points of the two text boxes is less than a second preset threshold, the two text boxes are considered to belong to the same line of handwritten information within the cell, possibly due to detection segmentation caused by inconsistencies in handwriting style. In this case, the two text boxes are directly merged to ensure that the handwritten information within the cell is completely recognized as a whole, avoiding information loss or incorrect merging due to detection errors. If the difference in the ordinates is greater than the second preset threshold, the handwritten information within the cell is considered to have undergone a line break in the vertical direction. In this case, the text boxes can be sorted according to the ordinate values ​​of their center points, and the sorted text box order is retained and recorded to ensure that the reading order of the handwritten information within the cell matches the writing order, avoiding information reversal and confusion, and ensuring the accuracy and standardization of the recognition results.

[0087] 3. Boundary correction and text box calibration.

[0088] For each text box i within cell box i, boundary correction is performed to ensure that the text is completely within the cell, preventing text from overflowing the cell boundaries. Specifically: the difference in side length between the first target text box and the target cell is determined. If the difference is less than 0, the first target text box is not adjusted; if the difference is greater than 0 and the ratio of the absolute value of the difference to the width of the target cell is less than a third preset threshold, the first target text box is not adjusted; if the difference is greater than 0 and the ratio of the absolute value of the difference to the width of the target cell is greater than or equal to the third preset threshold, the side length of the target cell is determined as the side length of the first target text box.

[0089] In this embodiment of the application, the left side length of text box i is used as an example for explanation. The difference between the left side length of text box i and the left side length of cell box i is calculated. If the difference is less than 0, the text is considered to be inside cell box i. If the difference is greater than 0, the text is considered to be outside cell box i.

[0090] 1) If the difference in side length is less than 0, no boundary adjustment is required. The left side of the original text box i is used as the left side of the corrected text box i to ensure the correct positioning of the handwritten information in the cell.

[0091] 2) If the difference is greater than 0 and the absolute value of the difference / cell width is less than the third preset threshold, it is considered that the text content is written out of the cell, but the excess part is small, and the whole is still handwritten information inside the cell. At this time, there is no need to adjust the boundary. The left side of the original text box i is directly used as the left side of the corrected text box i, so as to retain as much information as possible without sacrificing the integrity of the information.

[0092] 3) If the difference is greater than 0 and the absolute value of the difference / cell width is greater than or equal to the third preset threshold, it is considered that the text content is written out of the cell and the excess part is too large. At this time, boundary adjustment is required. The left side of cell box i is used as the left side of the corrected text box i to ensure that the information in each cell is independent and accurate, and to avoid erroneously merging the information of different cells together.

[0093] Similar checks and corrections are made to the right, top, and bottom edges of the text box to further ensure that every part of the handwritten information in the text box can be accurately mapped to the correct cell, even if the handwritten text is not evenly distributed or has phenomena such as line breaks and out-of-bounds errors.

[0094] In step S208 above, the main task is to perform OCR text recognition, that is, to recognize the text content in the target text box, thereby obtaining the text recognition result of the handwritten information in the table, and then to standardize the text recognition result and save it to an Excel file.

[0095] Optionally, recognizing the text content in the target text box to obtain the text recognition result corresponding to the handwritten information in the table includes: recognizing the target text box using a first recognition algorithm and a second recognition algorithm respectively to obtain the first text content and the second text content, wherein the first recognition algorithm is used to recognize the handwritten information in the handwritten information in the table, and the second recognition algorithm is used to recognize the printed information in the handwritten information in the table; standardizing the first text content and fusing the standardized first text content, the second text content, and the target table structure to obtain the text recognition result corresponding to the handwritten information in the table; and saving the text recognition result as a table file.

[0096] In this embodiment of the application, during the recognition and post-processing of text content, the original RGB format three-channel color image first needs to be precisely cropped based on the corrected target text box, and then an OCR text recognition algorithm is applied to parse the handwritten text in the image. The specific analysis is as follows:

[0097] 1. A dual recognition algorithm is used for text recognition.

[0098] To accommodate the diversity of handwriting styles, the recognition process defaults to using the first recognition algorithm. This algorithm is specifically designed for recognizing handwriting information and can more accurately understand and interpret the diversity of handwriting styles, providing a high recognition rate even in cases of irregular handwriting or blurred characters. For the table header, a second recognition algorithm can be configured as needed. This second algorithm can specifically recognize printed information, primarily used to process printed titles or fixed information in tables, ensuring the accurate recognition of this structured information.

[0099] By combining two different OCR recognition algorithms, the system comprehensively covers the recognition needs of handwritten information in tables, effectively recognizing and processing both handwritten and printed information.

[0100] 2. Standardization and integration of text content.

[0101] The first text content (handwritten information) and the second text content (printed information) obtained from the identification are subjected to a series of normalization processes to ensure the uniformity and standardization of the information format.

[0102] For example, in the phone number column, only numbers are retained, and numbers shorter than 11 digits are padded with underscores ("_") to ensure the integrity of the number and to indicate potential recognition errors. In the name column, numbers and special characters are removed to ensure the purity of the name information. The date column supports conversions to various date formats, such as YYYY / MM / DD, YYYY.MM.DD, and YYYYMMDD, to ensure the consistency and accuracy of the date information.

[0103] 3. Result saving and output.

[0104] Finally, the standardized content is integrated with the target table structure to ensure that information is accurately filled into the corresponding cells, avoiding issues such as misaligned rows and columns, and errors in information merging or splitting. The integrated information will be saved in a spreadsheet file (such as an Excel file) for easy information management and use later.

[0105] This application innovatively combines table structure adaptive correction and text position adaptive correction technologies to achieve accurate recognition and electronic conversion of diverse handwritten tables. By intelligently analyzing the structural features of the table, even in cases of missed or false cell detections or complex and varied handwritten information, the position of cells and text boxes can be automatically adjusted and corrected to ensure accurate information entry. Furthermore, this application introduces a dual-mode OCR recognition algorithm, selecting the most suitable recognition strategy based on the text type (handwritten or printed), and is equipped with a sophisticated post-processing workflow. This standardizes common information such as phone numbers, names, and dates, greatly improving the accuracy and readability of the recognition results and reducing the workload of manual proofreading.

[0106] Overall, this application not only significantly improves the efficiency and robustness of handwritten form information recognition, but also reduces application costs and enhances its scalability and practicality through its modular design and flexible configuration, providing a comprehensive and efficient solution to the digitization challenges of handwritten forms in real-world scenarios.

[0107] According to embodiments of this application, a form handwritten information recognition device is provided. It should be noted that the form handwritten information recognition device of this application can be used to execute the form handwritten information recognition method provided in the embodiments of this application. The form handwritten information recognition device provided in the embodiments of this application will be described below.

[0108] Figure 7 This is a structural diagram of a handwritten information recognition device for forms according to an embodiment of this application. As shown in Figure X, the device includes:

[0109] The acquisition module 70 is used to acquire the handwritten information of the table to be recognized;

[0110] The correction module 72 is used to determine the table structure of the handwritten information in the table and correct the table structure according to the preset table correction rules to obtain the target table structure. The table structure is used to reflect the cell information in the handwritten information in the table. The preset table correction rules are based on the rules that adjacent cells share the same vertex, cells in the same column have the same column width, and cells in the same row have the same row height to correct abnormal cells in the table structure.

[0111] The adjustment module 74 is used to determine the text box containing the text content in the handwritten information of the table, and adjust the text box according to the relative positional relationship between the text box and the cell in the target table structure to obtain the target text box;

[0112] The recognition module 76 is used to recognize the text content in the target text box and obtain the text recognition result corresponding to the handwritten information in the table.

[0113] Through the acquisition module, correction module, adjustment module and recognition module in the above-mentioned handwritten information recognition device, the goal of accurately recognizing complex handwritten table information is achieved, thereby realizing the technical effect of greatly improving the automation level of information entry and the accuracy of data processing. In turn, it solves the technical problems in related technologies such as low recognition efficiency of diverse handwritten table information and inability to adapt to changes in the position of handwritten fonts inside and outside cells and unstable spacing between text in cells.

[0114] In the table handwritten information recognition device provided in this application embodiment, the correction module is further used to recognize the table handwritten information through the table structure recognition model to obtain the cell information of the table handwritten information, wherein the cell information is a multidimensional array that includes at least the number of cells in the table handwritten information, the cell boundary coordinates and the cell row and column number coordinates; and to determine the table structure based on the cell information.

[0115] In the table handwritten information recognition device provided in this application embodiment, the correction module is further used to determine the feature matrix corresponding to the table structure; determine the common vertex coordinate distance of adjacent cells in the feature matrix, the column width difference of cells in the same column, and the row height difference of cells in the same row according to the preset table correction rules; determine abnormal cells in the feature matrix according to the common vertex distance, column width difference, and row height difference; correct the boundary coordinates and row and column index coordinates of the abnormal cells, and convert the corrected feature matrix into an array form to obtain the target table structure.

[0116] In the table handwriting information recognition device provided in this application embodiment, the correction module is further used to determine the target number of columns and the average number of columns in the table structure, wherein the target number of columns is used to represent the number of columns in any row of cells in the table structure; the row corresponding to the target number of columns being greater than the average number of columns is determined as the handwriting part of the table, and the cell row number of the table structure is corrected in the handwriting part of the table.

[0117] In the form handwritten information recognition device provided in this application embodiment, the adjustment module is further used to recognize the form handwritten information through a text recognition model to obtain the text content in the form handwritten information; and to determine the text box information corresponding to the text content, wherein the text box information is a multidimensional array including at least the number of text boxes and the coordinates of the text box boundaries.

[0118] In the form handwritten information recognition device provided in this application embodiment, the adjustment module is further used to determine the intersection between the target cell and the target text box, wherein the target cell is any cell in the form handwritten information, the target text box is any text box in the form handwritten information, and the intersection is used to reflect the degree of overlap between the target cell and the target text box; when the intersection is greater than a first preset threshold, it is determined that the target text box is located inside the target cell; the number of text boxes in the target cell is determined, and when the number of text boxes is 0, the bounding box of the target cell is determined as the target text box inside the target cell; when the number of text boxes is 1, the target text box inside the target cell is retained; when the number of text boxes is greater than 1, the target text box inside the target cell is adjusted according to the ordinate distance of the center point of the target text box.

[0119] In the form handwritten information recognition device provided in this application embodiment, the adjustment module is further used to determine the difference in the vertical coordinates of the center points of the first target text box and the second target text box, wherein the first target text box and the second target text box are any two text boxes within the target cell; if the difference in the vertical coordinates is less than a second preset threshold, the first target text box and the second target text box are merged; if the difference in the vertical coordinates is greater than the second preset threshold, the target text box and the second target text box are sorted according to the size of the vertical coordinates of their center points, and the sorted first target text box and the second target text box are retained within the target cell.

[0120] In the form handwritten information recognition device provided in this application embodiment, the adjustment module is further used to determine the difference in side length between the first target text box and the target cell. If the difference in side length is less than 0, the first target text box is not adjusted; if the difference in side length is greater than 0 and the ratio of the absolute value of the difference in side length to the width of the target cell is less than a third preset threshold, the first target text box is not adjusted; if the difference in side length is greater than 0 and the ratio of the absolute value of the difference in side length to the width of the target cell is greater than or equal to the third preset threshold, the side length of the target cell is determined as the side length of the first target text box.

[0121] In the form handwritten information recognition device provided in this application embodiment, the recognition module is further used to recognize the target text box using a first recognition algorithm and a second recognition algorithm respectively, to obtain a first text content and a second text content. The first recognition algorithm is used to recognize the handwritten information in the form handwritten information, and the second recognition algorithm is used to recognize the printed information in the form handwritten information. The first text content is normalized, and the normalized first text content, the second text content, and the target table structure are fused to obtain the text recognition result corresponding to the form handwritten information. The text recognition result is saved as a table file.

[0122] This application also provides an electronic device, including: a memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the above-described method for recognizing handwritten information in a table.

[0123] It should be noted that the aforementioned electronic equipment is used to perform Figure 2 The handwritten information recognition method shown in the table is also applicable to this electronic device, and will not be repeated here.

[0124] This application also provides a non-volatile storage medium, which includes a stored computer program, wherein the device containing the non-volatile storage medium executes the above-described method for recognizing handwritten information in a form by running the computer program.

[0125] It should be noted that the aforementioned non-volatile storage media is used for execution. Figure 2 The handwritten information recognition method shown above applies to this non-volatile storage medium as well, and will not be repeated here.

[0126] This application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the above-described method for recognizing handwritten information in a table.

[0127] It should be noted that the above-mentioned computer program product is used to execute Figure 2 The handwritten information recognition method shown in the table is also applicable to this computer program product, and will not be repeated here.

[0128] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0129] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0130] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0131] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0132] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0133] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0134] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for recognizing handwritten information in a form, characterized in that, include: Obtain the handwritten information of the form to be recognized; The table structure of the handwritten information in the table is determined, and the table structure is corrected according to the preset table correction rules to obtain the target table structure. The table structure is used to reflect the cell information in the handwritten information in the table. The preset table correction rules are based on the rules that adjacent cells share vertices, cells in the same column have the same column width, and cells in the same row have the same row height to correct abnormal cells in the table structure. Identify the text box containing the text content in the handwritten information of the table, and adjust the text box according to the relative positional relationship between the text box and the cell in the target table structure to obtain the target text box; The text content in the target text box is identified to obtain the text recognition result corresponding to the handwritten information in the table.

2. The method according to claim 1, characterized in that, Determining the table structure of the handwritten information in the table includes: The handwritten information of the table is identified by a table structure recognition model to obtain the cell information of the handwritten information of the table. The cell information is a multidimensional array that includes at least the number of cells in the handwritten information of the table, the cell boundary coordinates, and the cell row and column index coordinates. The table structure is determined based on the cell information.

3. The method according to claim 2, characterized in that, The table structure is corrected according to preset table correction rules to obtain the target table structure, including: Determine the feature matrix corresponding to the table structure; Based on the preset table correction rules, the common vertex coordinate distance between adjacent cells in the feature matrix, the column width difference between cells in the same column, and the row height difference between cells in the same row are determined respectively. Abnormal cells in the feature matrix are determined based on the shared vertex distance, the column width difference, and the row height difference; Correct the boundary coordinates and row and column index coordinates of the abnormal cells, and convert the corrected feature matrix into an array to obtain the target table structure.

4. The method according to claim 3, characterized in that, Before determining the feature matrix corresponding to the table structure, the method further includes: Determine the target number of columns and the average number of columns in the table structure, wherein the target number of columns is used to represent the number of columns in any row of cells in the table structure; The row corresponding to the target number of columns being greater than the average number of columns is identified as the handwritten part of the table, and the cell row number of the table structure is corrected in the handwritten part of the table.

5. The method according to claim 1, characterized in that, Before determining the text box containing the handwritten information in the table, the method further includes: The handwritten information in the table is identified using a text recognition model to obtain the text content within the handwritten information. Determine the text box information corresponding to the text content, wherein the text box information is a multidimensional array that includes at least the number of text boxes and the coordinates of the text box boundaries.

6. The method according to claim 5, characterized in that, The text box is adjusted according to its relative position to the cells in the target table structure to obtain the target text box, including: Determine the intersection between the target cell and the target text box, wherein the target cell is any cell in the handwritten information of the table, and the target text box is any text box in the handwritten information of the table, and the intersection is used to reflect the degree of overlap between the target cell and the target text box; If the intersection is greater than a first preset threshold, it is determined that the target text box is located within the target cell; Determine the number of text boxes within the target cell. If the number of text boxes is 0, define the bounding box of the target cell as the target text box within the target cell. If the number of text boxes is 1, the target text box within the target cell is retained; If the number of text boxes is greater than 1, the target text box in the target cell is adjusted according to the ordinate distance of the center point of the target text box.

7. The method according to claim 6, characterized in that, Adjusting the target text box within the target cell based on the y-coordinate distance of the center point of the target text box includes: Determine the difference in the ordinate of the center points of the first target text box and the second target text box, wherein the first target text box and the second target text box are any two text boxes within the target cell; If the difference in the vertical coordinate is less than a second preset threshold, the first target text box and the second target text box are merged. If the difference in the vertical coordinates is greater than the second preset threshold, the target text box and the second target text box are sorted according to the size of the vertical coordinate value of the center point, and the sorted first target text box and second target text box are retained in the target cell.

8. The method according to claim 7, characterized in that, The method further includes: Determine the difference in side length between the first target text box and the target cell. If the difference in side length is less than 0, do not adjust the first target text box. If the difference in side length is greater than 0 and the ratio of the absolute value of the difference in side length to the width of the target cell is less than a third preset threshold, the first target text box will not be adjusted. If the difference in side length is greater than 0 and the ratio of the absolute value of the difference in side length to the width of the target cell is greater than or equal to the third preset threshold, the side length of the target cell is determined as the side length of the first target text box.

9. The method according to claim 1, characterized in that, Recognizing the text content in the target text box to obtain the text recognition result corresponding to the handwritten information in the table includes: The target text box is identified by a first recognition algorithm and a second recognition algorithm respectively to obtain a first text content and a second text content. The first recognition algorithm is used to identify handwritten information in the handwritten information of the table, and the second recognition algorithm is used to identify printed information in the handwritten information of the table. The first text content is normalized, and the normalized first text content, the second text content, and the target table structure are fused to obtain the text recognition result corresponding to the handwritten information in the table; Save the text recognition results as a table file.

10. A device for recognizing handwritten information in a form, characterized in that, include: The acquisition module is used to acquire the handwritten information of the form to be recognized; The correction module is used to determine the table structure of the handwritten information in the table, and correct the table structure according to the preset table correction rules to obtain the target table structure. The table structure is used to reflect the cell information in the handwritten information in the table. The preset table correction rules are based on the rules that adjacent cells share vertices, cells in the same column have the same column width, and cells in the same row have the same row height to correct abnormal cells in the table structure. The adjustment module is used to determine the text box containing the text content in the handwritten information of the table, and adjust the text box according to the relative positional relationship between the text box and the cell in the target table structure to obtain the target text box; The recognition module is used to recognize the text content in the target text box and obtain the text recognition result corresponding to the handwritten information in the table.

11. An electronic device, characterized in that, include: A memory and a processor, wherein the memory is used to store program instructions; The processor, connected to the memory, is used to execute the table handwritten information recognition method according to any one of claims 1 to 9.

12. A non-volatile storage medium, characterized in that, The non-volatile storage medium includes a stored computer program, wherein the device containing the non-volatile storage medium executes the table handwritten information recognition method according to any one of claims 1 to 9 by running the computer program.

13. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the table handwritten information recognition method according to any one of claims 1 to 9.

Images