OCR Recognition and Correction Method for Bill Format

Abstract

The present invention proposes a method for correcting OCR recognition of bill formats, including collecting bill information; establishing a word group model by OCR scanning the text information in the picture data; deleting the text data in the picture, retaining the bill format picture, comparing the bill format picture with the standard format model, and marking the error areas; regularizing the marked areas of all bills, obtaining the bill format error rate, recording the marked areas with an error rate exceeding the threshold, and repeating the screening of the recorded areas when collecting bill information next time. Bills are very common in daily life. They are converted from original paper bills, and their main purpose is to facilitate storage and transportation. However, if the bills do not have the correct format and data, it will be very difficult to realize their functions. For most users, they will scan the original paper bills and perform digital processing (traditional image recognition technology).

Description

Technical Field

[0001] The present invention belongs to the field of OCR recognition, and particularly relates to a method for correcting OCR recognition of bill formats. Background Art

[0002] Currently, sometimes we need to convert paper bills into electronic formats. For example, when issuing invoices to merchants, we need to convert paper bills into electronic formats. Then, the common situation is that I need to convert paper bills into electronic documents and then perform OCR recognition. Generally, there are two methods: 1. Directly convert the picture into text form and then perform OCR recognition; 2. First recognize the text from the image and then perform OCR recognition. Both methods have their own advantages and disadvantages: 1. After converting the picture into text, when we perform OCR recognition, errors will be directly recognized. The reason is that the picture format is not considered. Such a situation has been encountered in practice. 2. When converting into text content after performing OCR recognition first, the error rate is relatively high because the picture format does not meet the requirements. Currently, there are two common methods: 1. First convert the picture into text content; 2. First use the form matching method to determine whether the drawing meets the requirements; the first method is more suitable for most bill format problems (including certificates and checks). The second method requires more data support, but currently, there is no good way to perform complete recognition and data correction.

[0003] Therefore, there is an urgent need for a method for correcting OCR recognition of bill formats. Summary of the Invention

[0004] The present invention provides a method for correcting OCR recognition of bill formats, which solves the problem that there is no complete and feasible error correction method for errors that easily occur when bills are digitized in the prior art, and marks and corrects the possible errors in bill digitization.

[0005] The technical solution of the present invention is implemented as follows: A method for correcting OCR recognition of bill formats includes:

[0006] Collect bill information;

[0007] Build a word group model by OCR scanning the text information in the picture data;

[0008] Delete the text data in the picture, retain the bill format picture, compare the bill format picture with the standard format model, and mark the error areas;

[0009] Regularize all the marked areas of the bills, obtain the bill format error rate, record the marked areas with an error rate exceeding the threshold, and repeat the screening of the recorded areas when collecting bill information next time.

[0010] Bills are very common in daily life. They are converted from original paper bills, and their main purpose is to facilitate storage and transportation. However, if the bills do not have the correct format and data, it will be very difficult to realize their functions. For most users, they will scan the original paper bills and then perform digital processing (traditional image recognition technology). Now, OCR technology can convert them into formats that can be used in various application programs, such as Excel, Word, PDF, TIFF, etc.

[0011] This application document first scans the original bill through OCR technology and then performs digital processing of the image. After scanning the bill, the image is first converted into a binary data format, and then the data is further processed. During the processing, the character features in the original bill will be analyzed, such as identifying characters. Then the character data will be compared with the characters in the original bill. If they are inconsistent, the format of the bill will be determined according to the characters we have identified. If they do not match, the specific number of characters of each bill and the position of each byte will be identified through OCR technology. After further processing of the format, it can be converted into a format that can be used in various application programs through OCR software.

[0012] As a preferred embodiment, the collection of bill information includes the following steps:

[0013] Determine the type of bill to be collected;

[0014] Retrieve the standard data model corresponding to the type of bill; pre-determine the corresponding relationships of the input fields, basic templates, and input templates;

[0015] After collecting the bill surface information, classify and summarize it.

[0016] As a preferred embodiment, the word and phrase model analyzes the word and phrase model input by the user using the bill text database, obtains the matching probability of the character distribution in the word and phrase model, screens the text in the word and phrase model according to the match, determines whether the text characters in the word and phrase model are misspelled words, and determines the correctness rate of the front and back word groups according to the matching probability.

[0017] As a preferred embodiment, before deleting the picture text data, extract the text in the picture, then remove the text in the original picture, and then perform format comparison. When performing format comparison, split the bill format picture into structures and compare them one by one according to the split structures.

[0018] As a preferred embodiment, when regularizing the marked area of the bill, classify it according to the bill type, sort it according to the bill issue date, and then perform a secondary sort according to the amount numbers in the bill.

[0019] After adopting the above technical solution, the beneficial effects of the present invention are as follows: improving the efficiency of bill processing, reducing the bill production cost, and having a significant improvement in processing efficiency compared with the manual method. Through the image recognition software integrating OCR technology, it can perform OCR recognition and correction on bills in different formats such as pictures, texts, and tables. It can perform recognition and correction on various formats of bills and can accurately recognize various characters in the bills.

[0020] It can perform recognition and correction on different bills. When performing OCR recognition and correction on different printed fonts or handwritten fonts, it has good effects. Compared with the manual method, it can automatically perform operations such as picture format conversion, image skew correction, and text extraction. In terms of processing speed, the efficiency is also significantly improved compared with the manual method. For example, when performing OCR recognition on a blank commercial bill, it takes about 30 seconds. In addition, the software also supports operations such as text extraction and character recognition of PDF file formats, and can export the recognition results to a word document. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0022] Figure 1 It is the method flow chart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0024] Embodiment:

[0025] As Figure 1 shown, the method for OCR recognition and correction of bill formats includes:

[0026] Collect bill information;

[0027] Build a word and phrase model by OCR scanning the text information in the picture data;

[0028] Delete the text data in the picture, retain the bill format picture, compare the bill format picture with the standard format model, and mark the error areas;

[0029] Regularize the marked areas of all bills, obtain the bill format error rate, record the marked areas with an error rate exceeding the threshold, and repeat the screening of the recorded areas when collecting bill information next time.

[0030] The steps for collecting bill information are as follows:

[0031] Determine the types of bills to be collected;

[0032] Retrieve the standard data model corresponding to the types of bills; predetermine the corresponding relationships among the input fields, the basic template, and the input template;

[0033] Classify and summarize the information on the bill surface after collection.

[0034] In terms of data entry, the advantages of OCR technology are very obvious. The traditional manual entry method requires users to classify and organize the bills by themselves and then convert the bills into text files. Now, only need to install OCR software in the computer, then input the format of the bills in the OCR software and classify, organize and edit them as required, and finally convert the original data into text files.

[0035] The word and phrase model analyzes the word and phrase model input by the user using the bill text database, obtains the matching probability of the character distribution in the word and phrase model, screens the text in the word and phrase model according to the matching, judges whether the text characters in the word and phrase model are misspelled words, and determines the accuracy rate of the previous and subsequent phrases according to the matching probability. Through data input, first scan the paper document on the bill with a scanner, and then use OCR software to preprocess the original image, remove various interferences, and improve the recognition efficiency; then perform segmentation and extraction. Subsequently, perform character segmentation, and segment the text on the bill into character strings with characters. Use the characters on the bill when performing character recognition.

[0036] The adopted text character recognition can complete the recognition and segmentation of different types of texts, and splice the words on the bill through text splicing. This step is different from the first step. It is necessary to first segment the single-image text and then perform text splicing. Most importantly, it is Chinese character recognition. Using OCR technology can quickly and accurately recognize Chinese characters and punctuation marks;

[0037] In actual use, save the file to a document and then edit it in various formats such as Excel, Word, PDF, TIFF, etc. If you are not satisfied with the converted document, you can perform the conversion. To meet the needs of different customers, OCR technology can also be used in various fields, such as the automotive industry. For banks or other financial institutions, scan the bills and upload them to the background, and then use OCR technology to convert them into formats that can be used in various applications. The results obtained through OCR technology are usually very accurate and can be saved in various formats and used by other users.

[0038] Before deleting the text data in the picture, extract the text in the picture, then remove the text in the original picture, and then perform format comparison. When performing format comparison, split the bill format picture into structures and compare them one by one according to the split structures. When regularizing the marked area of the bill, classify it according to the bill type, sort it according to the bill issue date, and then perform a secondary sort according to the amount figures in the bill.

[0039] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for OCR recognition and correction of bill formats, characterized in that: include: Collect bill information; Scan the text information in the image data through OCR and build a word group model; Delete the text data in the image, keep the bill format image, compare the bill format image with the standard format model, and mark the error area; Regularize the marked areas of all bills, obtain the bill format error rate, record the marked areas where the error rate exceeds the threshold, and repeat the screening of the recorded areas when collecting bill information next time.

2. The method for OCR recognition and correction of bill formats according to claim 1, characterized in that: The collecting of bill information includes the following steps: Determine the type of bills to be collected; Retrieve the standard data model for the corresponding bill type; predetermine the corresponding relationship between the input field, basic template, and input template; Collect the information on the bill and classify it.

3. The method for OCR recognition and correction of bill formats according to claim 1, characterized in that: The word and phrase model uses the bill text database to analyze the word and phrase model input by the user, obtains the matching probability of the character distribution in the word and phrase model, filters the text in the word and phrase model based on the matching, determines whether the text characters in the word and phrase model are typos, and determines the accuracy of the previous and next phrases based on the matching probability.

4. The method for OCR recognition and correction of bill formats according to claim 1, characterized in that: Before deleting the text data in the image, the text in the image is extracted, and then the text in the original image is removed, and then a format comparison is performed. When performing the format comparison, the bill format image is structurally split and compared one by one according to the split structure.

5. The method for OCR recognition and correction of bill formats according to claim 1, characterized in that: When regularizing the bill marking area, first classify according to the bill type, sort according to the bill issuance date, and then perform a secondary sort according to the amount in the bill.

Images