Automatic document reimbursement method and system and electronic equipment

Abstract

The invention discloses an automatic document reimbursement method and system and electronic equipment, and the method comprises the steps: obtaining a delivered paper image, carrying out the enhancement preprocessing and preliminary filtering of the paper image, and obtaining a to-be-matched image; detecting the rectangular contours of the to-be-matched image by using an edge detection algorithm, and judging the reimbursement type of the to-be-matched image according to the number of the rectangular contours; extracting a recognition area in the to-be-matched image according to the reimbursement type, and extracting a first text from the recognition area; calling a reported electronic image corresponding to the delivered paper image in the reimbursement system, and extracting a second text from the reported electronic image; and matching the first text with the second text, and if the matching is successful, archiving and reimbursement are carried out on the currently delivered paper image. Classification and identification are carried out according to the bill types, whether the delivered paper images are matched with the electronic bills in the reimbursement system or not is judged, reimbursement can be successful if yes, the reimbursement efficiency is improved, and false alarms are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic document reimbursement processing, and in particular to an automatic document reimbursement method, system and electronic equipment. Background Art

[0002] The existing corporate reimbursement method generally adopts a combination of electronic information and manual methods. When applying for reimbursement, the person in charge needs to first fill in the electronic information through the reimbursement system, and then scan the paper materials and upload them to the reimbursement system. In the actual paper-electronic comparison business, the image files in the reimbursement system and the delivered paper files are matched only based on the key information of the invoice. Often, due to the person in charge's mistake in uploading the paper materials, they cannot be matched with the electronic image files in the reimbursement system, resulting in reimbursement errors. Existing, the auditor needs to manually compare each reimbursement document, which is inefficient. Summary of the invention

[0003] The purpose of the present invention is to provide a method, system and electronic device for automatic document reimbursement, which can classify and identify documents according to their types, and determine whether the delivered paper image matches the electronic form in the reimbursement system. If they match, reimbursement can be successfully made, thereby improving reimbursement efficiency and reducing false positives.

[0004] In one aspect, the present invention provides a method for automatic reimbursement of documents, which specifically comprises the following steps:

[0005] S1. Obtain the delivered paper image, perform enhancement preprocessing and preliminary filtering on the paper image, and obtain the image to be matched;

[0006] S2, using an edge detection algorithm to detect the rectangular contour of the image to be matched, and judging the reimbursement type of the image to be matched according to the number of rectangular contours;

[0007] S3, extracting a recognition area in the to-be-matched image according to the reimbursement type, and extracting a first text from the recognition area;

[0008] S4, calling the reported electronic image corresponding to the delivered paper image in the reimbursement system, and extracting the second text from the reported electronic image;

[0009] S5. Match the first text and the second text. If the match is successful, archive the currently delivered paper image and reimburse it.

[0010] In some specific implementations, the specific process of obtaining the image to be matched is:

[0011] S11, performing sharpening and grayscale processing on the paper image to obtain a first image;

[0012] S12, determining whether the first image is a blank image;

[0013] S13: If not, use an image segmentation method to remove the marked data on the first image to obtain an image to be matched.

[0014] In some specific implementations, the specific process of step S12 is:

[0015] S121, compressing the resolution of the first image, and uniformly compressing the width of the first image to fixed pixels to obtain a compressed image;

[0016] S122, calculating the gray value of each pixel in the compressed image, and calculating the variance of the pixel brightness value in the compressed image according to the gray value;

[0017] S123: Determine whether the variance is greater than a set threshold, and if so, determine that the image is not a blank image.

[0018] In some specific implementations, the reimbursement type includes a form document and a document attachment sheet, and the specific process of step S2 is:

[0019] S21, using the Canny edge detection algorithm and opencv to find the rectangular outline and image edge outline of the image to be matched, calculating the image size of the image to be matched according to the image edge outline, and identifying the text of the image to be matched;

[0020] S22, after finding the rectangular contours, screen and calculate the size of each rectangular contour, and select the largest rectangular contour from each rectangular contour;

[0021] S23, calculating the ratio of the size of the largest rectangular outline to the size of the image to be matched, and determining whether the ratio exceeds a preset threshold;

[0022] S24, if the ratio exceeds the first preset threshold, continue to determine whether the total number of rectangular contours exceeds the set value, if it exceeds the set value and the text contains "table", determine that the reimbursement type of the image to be matched is a table document;

[0023] S25. If the ratio exceeds the second preset threshold and the number of rectangular contours is less than the set value, and the text contains "paste", it is determined that the reimbursement type of the image to be matched is a document pasting form.

[0024] In some specific implementations, the specific process of step S3 is:

[0025] For the image to be matched whose reimbursement type is a form document, the form area is extracted from the image to be matched as the recognition area;

[0026] Perform OCR text recognition on the recognition area to obtain a first text.

[0027] In some specific implementations, the specific process of step S3 is:

[0028] For the image to be matched whose reimbursement type is a receipt pasting form, the largest rectangular contour area except the edge contour of the image is extracted from the image to be matched as the recognition area;

[0029] Perform OCR text recognition on the recognition area to obtain a first text.

[0030] In some specific implementations, the specific matching process in step S5 includes:

[0031] S51, performing long word segmentation on the first text to obtain first word segmentation information, and performing long word segmentation on the second text to obtain second word segmentation information;

[0032] S52: Match the first word segmentation information with the second word segmentation information to determine whether the first word segmentation information includes all the content of the second word segmentation information. If so, determine that the first text and the second text are matched successfully.

[0033] In some specific implementations, the specific process of step S52 is:

[0034] Use the second word segmentation information to traverse the first word segmentation information, and calculate the similarity between each second word segmentation and the first word segmentation information, retain the word segmentation matching result with the highest similarity and store it in the word segmentation matching table, which includes the most similar word segmentation matched by each second word segmentation in the first word segmentation information and the corresponding similarity;

[0035] In the word segmentation matching table, the number of words is used as a weight, and the similarity is used as a reference to obtain the comprehensive similarity between the first word segmentation information and the second word segmentation information;

[0036] It is determined based on the comprehensive similarity whether the first text includes all the content of the second text. If so, it is determined that the first text and the second text are matched successfully.

[0037] In a second aspect, the present application includes a document automatic reimbursement system, including:

[0038] The image preprocessing module is used to obtain the delivered paper image, perform enhancement preprocessing and preliminary filtering on the paper image, and obtain the image to be matched;

[0039] An image type recognition module is used to detect the rectangular contour of the image to be matched using an edge detection algorithm, and determine the reimbursement type of the image to be matched according to the number of rectangular contours;

[0040] A text recognition module, used to extract a recognition area in the image to be matched according to the reimbursement type, and extract the first text from the recognition area;

[0041] The text matching module is used to call the reported electronic image corresponding to the delivered paper image in the reimbursement system, extract the second text from the reported electronic image, match the first text with the second text, and if the match is successful, archive the currently delivered paper image and reimburse it.

[0042] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor implements an automatic document reimbursement method as described in the first aspect when executing the computer program.

[0043] The concept of this application is:

[0044] Currently, since the reimbursement system is electronic and the person handling the document has a paper document, when handing in the paper material, the image is scanned by a machine and compared with the one in the reimbursement system. However, in the actual paper-electronic comparison business, the image file in the reimbursement system and the delivered paper document are matched only based on the key information of the invoice. There are often problems with the inconsistency between the image in the reimbursement system and the actual delivered image due to the pasting method or the omission of the person handling the document. For example, the source image in the reimbursement system is a document that has not been signed when submitted, but the document has been signed when the document is actually submitted; some non-A4 size text documents are pasted on the pasting sheet when delivered, and the delivered image has more text on the pasting sheet than the original image. In order to avoid the mismatch between the reported electronic impact of the reimbursement system and the delivered paper document, each document needs to be manually reviewed to ensure that the reimbursement system and the delivered paper material match. Manual review is inefficient and prone to errors.

[0045] This application obtains the delivered paper image. Since the paper image is uploaded by the person in charge, there may be problems such as unclear images. The paper image needs to be preliminarily processed and then the image area to be identified is detected from the paper image. Specifically, in order to improve the recognition efficiency, the text information of the corresponding area can be quickly identified according to the reimbursement type of the document. For example, for some form documents that need to be signed, it is only necessary to identify the text in the form area, and for the pasted documents, it is only necessary to identify the content of the pasted document, which also reduces the error recognition rate. Since the number of rectangles in the form is large, and the number of rectangles in the pasted document is small, it is possible to distinguish whether the delivered paper image is a form document or a pasted document by identifying whether there are multiple rectangular outlines in the image, and then classify and identify it to improve the recognition efficiency.

[0046] The present invention has the beneficial effects:

[0047] In order to improve recognition efficiency, the text information in the corresponding area can be quickly identified according to the reimbursement type of the document. For example, for some form documents that need to be signed, only the text in the form area needs to be identified, and for sticky documents, only the content of the pasted document needs to be identified, which also reduces the misrecognition rate. Since the form contains a large number of rectangles, while the number of rectangles in the sticky document is small, it is possible to distinguish whether the delivered paper image is a form document or a sticky document by identifying whether the image contains multiple rectangular outlines, and then classify and identify it to improve recognition efficiency.

[0048] Through fuzzy matching, as long as the delivered image contains all the text of the original image, it can be considered a match, which effectively solves the problem of the delivered image having more content than the original image. In order to prevent the situation where two pictures with high similarity in text composition but are not the same file are mistakenly matched together, natural language analysis and long word segmentation are used to divide the text information. The word segmentation results are longer, and the results are more accurate, which can avoid incorrect matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A flow chart of a method for automatic reimbursement of documents provided by an embodiment of the present invention;

[0050] Figure 2 A schematic diagram of table outline recognition provided by an embodiment of the present invention;

[0051] Figure 3 A schematic diagram of table recognition results provided by an embodiment of the present invention;

[0052] Figure 4 A schematic diagram of the identification area of ​​a table provided in an embodiment of the present invention;

[0053] Figure 5 A schematic diagram of identifying a sticker sheet provided in an embodiment of the present invention;

[0054] Figure 6 A schematic diagram of the recognition area of ​​a sticker provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] The relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

[0057] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0058] Additionally, descriptions of well-known structures, functions, and configurations may be omitted for clarity and conciseness.One of ordinary skill in the art will recognize that various changes and modifications may be made to the examples described herein without departing from the spirit and scope of the present disclosure.

[0059] Technologies, methods, and apparatus known to ordinary technicians in the relevant field may not be discussed in detail, but where appropriate, such technologies, methods, and apparatus should be considered part of the authorization specification.

[0060] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0061] Example 1

[0062] like Figure 2 As shown, on the one hand, the present invention provides a method for automatic reimbursement of documents, which specifically includes the following steps:

[0063] S1. Obtain the delivered paper image, perform enhancement preprocessing and preliminary filtering on the paper image, and obtain the image to be matched;

[0064] Since the delivered paper images are manually photographed and uploaded by the handlers, the uploaded images may have poor image quality and may be unclear, or may contain blank images uploaded by mistake, or may have watermarks or other marking data. Therefore, the specific process of obtaining the image to be matched can be as follows:

[0065] S11, sharpening and grayscale processing are performed on the paper image to obtain a first image; sharpening the image can slightly improve the image clarity; grayscale processing of the color image can improve the accuracy of OCR recognition and cutting:

[0066] S12, determining whether the first image is a blank image; filtering blank images before OCR recognition helps save GPU resources, reduce IO consumption of OCR services, and reduce network requests.

[0067] Method to filter blank images:

[0068] S121, compressing the resolution of the first image, and uniformly compressing the width of the first image to fixed pixels to obtain a compressed image;

[0069] S122, calculating the gray value of each pixel in the compressed image, and calculating the variance of the pixel brightness value in the compressed image according to the gray value;

[0070] S123: Determine whether the variance is greater than a set threshold, and if so, determine that the image is not a blank image.

[0071] Compress the image resolution to improve processing efficiency. Compress the image width to 300 pixels; calculate the grayscale value of each pixel; calculate the variance of the pixel brightness value in the image. If the variance is small, it means that the pixel brightness in the image does not change much, and it may be a blank image; if the variance is large, it means that there is a significant brightness change in the image, and it may contain some content. The threshold can be customized according to the actual situation.

[0072] S13: If not, use an image segmentation method to remove the marked data on the first image to obtain an image to be matched.

[0073] Images can carry data and code snippets. To reduce the risk of server attacks, the data markers on the first image can be filtered out and a new identical image can be redrawn to remove the data carried by the image.

[0074] S2, using an edge detection algorithm to detect the rectangular contour of the image to be matched, and judging the reimbursement type of the image to be matched according to the number of rectangular contours;

[0075] From a business perspective, it is reasonable to add signatures to documents or paste bills onto A4 paper. At this time, simple text similarity matching is often too different to pass the match. The optimization algorithm of fuzzy matching the source image and the delivery image can solve this problem. The following is a brief description of the optimization algorithm: The reimbursement types include form documents and document paste sheets. The specific process of step S2 is:

[0076] S21, using the Canny edge detection algorithm and opencv to find the rectangular outline and image edge outline of the image to be matched, calculating the image size of the image to be matched according to the image edge outline, and identifying the text of the image to be matched;

[0077] S22, after finding the rectangular contours, screen and calculate the size of each rectangular contour, and select the largest rectangular contour from each rectangular contour;

[0078] S23, calculating the ratio of the size of the largest rectangular outline to the size of the image to be matched, and determining whether the ratio exceeds a preset threshold;

[0079] S24, if the ratio exceeds the first preset threshold, continue to determine whether the total number of rectangular contours exceeds the set value, if it exceeds the set value and the text contains "table", determine that the reimbursement type of the image to be matched is a table document;

[0080] S25. If the ratio exceeds the second preset threshold and the number of rectangular contours is less than the set value, and the text contains "paste", it is determined that the reimbursement type of the image to be matched is a document pasting form.

[0081] S3, extracting a recognition area in the to-be-matched image according to the reimbursement type, and extracting a first text from the recognition area;

[0082] (1) For an image to be matched whose reimbursement type is a form document, a table area is extracted from the image to be matched as a recognition area; and OCR text recognition is performed on the recognition area to obtain a first text.

[0083] (2) For the image to be matched whose reimbursement type is a receipt pasting form, extract the largest rectangular contour area except the edge contour of the image from the image to be matched as the recognition area; perform OCR text recognition on the recognition area to obtain the first text. For pasted receipts, cut off the receipt area and only match the receipt area.

[0084] S4, calling the reported electronic image corresponding to the delivered paper image in the reimbursement system, and extracting the second text from the reported electronic image;

[0085] S5. Match the first text and the second text. If the match is successful, archive the currently delivered paper image and reimburse it.

[0086] The specific matching process in step S5 includes:

[0087] S51, performing long word segmentation on the first text to obtain first word segmentation information, and performing long word segmentation on the second text to obtain second word segmentation information;

[0088] S52: Match the first word segmentation information with the second word segmentation information to determine whether the first word segmentation information includes all the content of the second word segmentation information. If so, determine that the first text and the second text are matched successfully.

[0089] In some specific implementations, the specific process of step S52 is:

[0090] Use the second word segmentation information to traverse the first word segmentation information, and calculate the similarity between each second word segmentation and the first word segmentation information, retain the word segmentation matching result with the highest similarity and store it in the word segmentation matching table, which includes the most similar word segmentation matched by each second word segmentation in the first word segmentation information and the corresponding similarity;

[0091] In the word segmentation matching table, the number of words is used as a weight, and the similarity is used as a reference to obtain the comprehensive similarity between the first word segmentation information and the second word segmentation information;

[0092] It is determined based on the comprehensive similarity whether the first text includes all the content of the second text. If so, it is determined that the first text and the second text are matched successfully.

[0093] 1) Perform long word segmentation on the text results of the source image and the delivered image (the word segmentation granularity is larger, and the word segmentation results are longer);

[0094] 2) The source image segmentation results are matched with the delivered image segmentation one by one, and a segmentation matching table (the most similar segmentation matched by each source image segmentation in the delivered image, and a table of similarities) is calculated;

[0095] 3) In the word segmentation matching table, the number of words is used as the weight and the similarity is used as the reference to obtain a comprehensive similarity.

[0096] By matching in this way, as long as the delivered image contains all the text of the original image, it can be considered a match, which effectively solves the situation where the delivered image has more content than the original image. In order to prevent the situation where two pictures with high similarity in text composition but are not the same file are mistakenly matched together, the algorithm uses natural language analysis to analyze long words. The word segmentation results are longer, and the results are more accurate, which can avoid false matches.

[0097] It is understandable that if Figure 2-Figure 3 As shown, through Canny edge detection, opencv is used to find contours. For example, Canny edge detection and Hough transform can be used to extract straight lines in the image, and cv2.line function can be used to draw straight lines. When the image contains a larger (for example, more than 35% of the size of the image to be matched) rectangular contour and multiple small rectangles, and there is "check-in" or "signature" text in the image, it is a form that needs to be signed. When the image to be identified is a form document, the source image of the reimbursement system is a document that has not been signed when it is submitted, but the document has been signed when it is actually submitted. For form documents, the weight of the signature area should be reduced before matching for the subsequent parts that need to be signed, and the text in the same column of the cell where "signature" / "signature" is located is extracted from the table. The text in the same column of the cell where "signature" / "signature" is located (the same as the x-axis of the keyword block) does not participate in the matching or reduces the matching requirements. As shown Figure 4 As shown, only the text information in the column area where the signature is located is recognized.

[0098] like Figure 5As shown in the figure, when a larger rectangular outline is identified, and it is not a table shape (a large rectangle is divided by a small rectangle), and there is a color difference between the rectangular area and the outer area, or the outer area contains text such as "sticker", it is a sticky ticket. Since a sticky ticket actually occupies most of the area of ​​the sticky ticket, in addition to the edge of the image to be matched, only the edge of the sticky document has the largest rectangular area. Therefore, the area where the bill is located can be directly identified and intercepted as the identification area, and then the text in the bill area can be identified for matching.

[0099] Example 2

[0100] This embodiment provides a document automatic reimbursement system, including:

[0101] The image preprocessing module is used to obtain the delivered paper image, perform enhancement preprocessing and preliminary filtering on the paper image, and obtain the image to be matched;

[0102] An image type recognition module is used to detect the rectangular contour of the image to be matched using an edge detection algorithm, and determine the reimbursement type of the image to be matched according to the number of rectangular contours;

[0103] A text recognition module, used to extract a recognition area in the image to be matched according to the reimbursement type, and extract the first text from the recognition area;

[0104] The text matching module is used to call the reported electronic image corresponding to the delivered paper image in the reimbursement system, extract the second text from the reported electronic image, match the first text with the second text, and if the match is successful, archive the currently delivered paper image and reimburse it.

[0105] Example 3

[0106] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that when the processor executes the computer program, an automatic document reimbursement method described in the first aspect is implemented.

[0107] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. According to the technical essence of the present invention, within the spirit and principles of the present invention, any simple modification, equivalent replacement and improvement made to the above embodiment still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for automatic reimbursement of documents, characterized in that: The specific steps include: S1. Obtain the delivered paper image, perform enhancement preprocessing and preliminary filtering on the paper image, and obtain the image to be matched; S2, using an edge detection algorithm to detect the rectangular contour of the image to be matched, and judging the reimbursement type of the image to be matched according to the number of rectangular contours; S3, extracting a recognition area in the to-be-matched image according to the reimbursement type, and extracting a first text from the recognition area; S4, calling the reported electronic image corresponding to the delivered paper image in the reimbursement system, and extracting the second text from the reported electronic image; S5. Match the first text and the second text. If the match is successful, archive the currently delivered paper image and reimburse it.

2. The method for automatic reimbursement of documents according to claim 1, characterized in that: The specific process of obtaining the image to be matched is: S11, performing sharpening and grayscale processing on the paper image to obtain a first image; S12, determining whether the first image is a blank image; S13: If not, use an image segmentation method to remove the marked data on the first image to obtain an image to be matched.

3. The automatic reimbursement method according to claim 2, characterized in that: The specific process of step S12 is: S121, compressing the resolution of the first image, and uniformly compressing the width of the first image to fixed pixels to obtain a compressed image; S122, calculating the gray value of each pixel in the compressed image, and calculating the variance of the pixel brightness value in the compressed image according to the gray value; S123: Determine whether the variance is greater than a set threshold, and if so, determine that the image is not a blank image.

4. The method for automatic reimbursement of documents according to claim 1, characterized in that: The reimbursement types include form documents and document attachment sheets. The specific process of step S2 is as follows: S21, using the Canny edge detection algorithm and opencv to find the rectangular outline and image edge outline of the image to be matched, calculating the image size of the image to be matched according to the image edge outline, and identifying the text of the image to be matched; S22, after finding the rectangular contours, screen and calculate the size of each rectangular contour, and select the largest rectangular contour from each rectangular contour; S23, calculating the ratio of the size of the largest rectangular outline to the size of the image to be matched, and determining whether the ratio exceeds a preset threshold; S24, if the ratio exceeds the first preset threshold, continue to determine whether the total number of rectangular contours exceeds the set value, if it exceeds the set value and the text contains "table", determine that the reimbursement type of the image to be matched is a table document; S25. If the ratio exceeds the second preset threshold value and the number of rectangular contours is less than the set value, and the text contains "paste", it is determined that the reimbursement type of the image to be matched is a document pasting form.

5. The automatic reimbursement method according to claim 4, characterized in that: The specific process of step S3 is: For the image to be matched whose reimbursement type is a form document, the form area is extracted from the image to be matched as the recognition area; Perform OCR text recognition on the recognition area to obtain a first text.

6. The method for automatic reimbursement of documents according to claim 4, characterized in that: The specific process of step S3 is: For the image to be matched whose reimbursement type is a receipt pasting form, the largest rectangular contour area except the edge contour of the image is extracted from the image to be matched as the recognition area; Perform OCR text recognition on the recognition area to obtain a first text.

7. The method for automatic reimbursement of documents according to claim 1, characterized in that: The specific matching process in step S5 includes: S51, performing long word segmentation on the first text to obtain first word segmentation information, and performing long word segmentation on the second text to obtain second word segmentation information; S52: Match the first word segmentation information with the second word segmentation information to determine whether the first word segmentation information includes all the content of the second word segmentation information. If so, determine that the first text and the second text are matched successfully.

8. The method for automatic reimbursement of documents according to claim 1, characterized in that: The specific process of step S52 is: Use the second word segmentation information to traverse the first word segmentation information, and calculate the similarity between each second word segmentation and the first word segmentation information, retain the word segmentation matching result with the highest similarity and store it in the word segmentation matching table, which includes the most similar word segmentation matched by each second word segmentation in the first word segmentation information and the corresponding similarity; In the word segmentation matching table, the number of words is used as a weight, and the similarity is used as a reference to obtain the comprehensive similarity between the first word segmentation information and the second word segmentation information; It is determined based on the comprehensive similarity whether the first text includes all the content of the second text. If so, it is determined that the first text and the second text are matched successfully.

9. A document automatic reimbursement system, characterized in that: include: The image preprocessing module is used to obtain the delivered paper image, perform enhancement preprocessing and preliminary filtering on the paper image, and obtain the image to be matched; An image type recognition module is used to detect the rectangular contour of the image to be matched using an edge detection algorithm, and determine the reimbursement type of the image to be matched according to the number of rectangular contours; A text recognition module, used to extract a recognition area in the image to be matched according to the reimbursement type, and extract the first text from the recognition area; The text matching module is used to call the reported electronic image corresponding to the delivered paper image in the reimbursement system, extract the second text from the reported electronic image, match the first text with the second text, and if the match is successful, archive the currently delivered paper image and reimburse it.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: When the processor executes the computer program, the method for automatic reimbursement of documents as described in any one of claims 1 to 8 is implemented.

Images