Method and device for recognizing and analyzing PDF text and picture based on OCR and storage medium

Abstract

The invention discloses a method and device for recognizing and analyzing PDF texts and pictures based on OCR and a storage medium, and particularly relates to the field of text and picture processing.The method for recognizing and analyzing PDF texts and pictures based on OCR comprises the following steps that S100, PDF documents are loaded, page analysis is conducted, page elements are extracted, and page layout information is processed; and step S200, for the extracted picture or the PDF page generated by scanning, performing high-precision OCR recognition by using a deep learning model, and converting characters in the image into editable texts. According to the method, efficient and accurate analysis of complex texts and picture contents in the PDF document is realized, the technical scheme not only improves the analysis precision and efficiency and enhances the flexibility and expandability of the system, but also promotes the development of digital and intelligent application of the document, and meanwhile, by saving resources and time cost, the method and the system have good application prospects. And remarkable economic benefits are brought to enterprises and individual users.

Description

Technical Field

[0001] The present invention relates to the field of text and image processing, and more specifically, to a method, device and storage medium for parsing PDF text and images based on OCR recognition. Background Art

[0002] Portable Document Format (PDF) is a widely used electronic file format developed by Adobe Systems in 1993. PDF documents have become a standard format for document exchange and storage due to their high compatibility across platforms and devices and their accurate retention of content formats (such as fonts, images, colors, etc.). PDF files can not only contain text content, but also embed multiple elements such as pictures, tables, forms, videos, and audios, making them widely used in publishing, education, corporate document management, and other fields.

[0003] Optical Character Recognition (OCR) technology is a technology that converts text in an image into a text form that can be edited and searched by a computer. Through image preprocessing, character segmentation, feature extraction, character recognition and post-processing, OCR technology converts text information in paper documents and pictures into digital text data, greatly improving the processing efficiency and accuracy of information. In recent years, with the development of deep learning technology, the recognition accuracy and scope of application of OCR technology have been significantly improved, and it can handle more complex and changeable text scenes.

[0004] Although the PDF format has many advantages, parsing its content, especially PDF documents containing mixed text and images, still faces many challenges:

[0005] Layout complexity: PDF documents can contain complex page layouts, such as multi-column layouts, tables, images wrapped around text, etc. These layouts increase the difficulty of extraction and parsing;

[0006] Font and encoding issues: PDF documents use a variety of fonts and encoding methods, and the processing of different language character sets requires specific support and conversion;

[0007] Image embedding: Images in PDF may be directly embedded as image objects or as part of the text background. Intelligent recognition technology is required to accurately distinguish and extract this information;

[0008] Security and permissions: Some PDF documents are password protected or have permission restrictions, which affects the feasibility and efficiency of parsing;

[0009] High-quality OCR requirements: For scanned PDFs or low-resolution images, OCR technology requires higher recognition accuracy and adaptability to reduce error rates.

[0010] There are some PDF parsing tools and libraries on the market, such as Apache PDFBox and iText, which mainly provide basic operations and content extraction functions for PDF documents. However, they still have the following shortcomings in processing complex mixed text and image content, especially in high-quality OCR recognition:

[0011] 1) Single function: Most tools focus on simple parsing and editing of PDF documents and lack intelligent recognition capabilities for complex page layouts;

[0012] 2) OCR recognition accuracy: Traditional OCR methods have low recognition accuracy when processing low-quality images or special fonts, and have limited support for multiple languages;

[0013] 3) Low integration: There is a lack of solutions that deeply integrate OCR recognition and PDF parsing, which means that users need to manually switch tools when processing PDFs containing a large number of images and complex texts, which is inefficient;

[0014] 4) Poor scalability: Existing tools often lack flexible configuration options and API interfaces, making it difficult to meet the customized requirements of specific application scenarios.

[0015] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0016] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a method, device and storage medium for parsing PDF text and images based on OCR recognition to solve the problems raised in the above-mentioned background technology.

[0017] To achieve the above object, the present invention provides the following technical solutions:

[0018] The method of parsing PDF text and images based on OCR recognition includes the following steps:

[0019] Step S100, loading a PDF document, parsing the page, extracting page elements, and processing page layout information;

[0020] Step S200, for the extracted image or the scanned PDF page, high-precision OCR recognition is performed using a deep learning model to convert the text in the image into editable text;

[0021] Step S300, based on the layout information of the PDF page and the OCR recognition result, intelligently distinguish text and images, reorganize the content, and restore the logical structure of the document;

[0022] Step S400, formatting the parsed text, such as correcting OCR recognition errors, adjusting text paragraphs, and identifying and extracting key information;

[0023] Step S500 provides multiple output format options and supports storing the analysis results in a specified storage medium.

[0024] In a preferred embodiment, step S100 specifically includes the following contents:

[0025] Use the PDF parsing library to load PDF documents and read page contents;

[0026] Traverse each page, extract the text, pictures and table elements in the page, and record their position, size and type attributes;

[0027] Apply deep learning models to identify page layouts and determine the layout structure of text blocks, image areas, and table frames;

[0028] The acquisition logic of the deep learning model is:

[0029] Collect a large number of annotated PDF page samples and convert them into a format suitable for deep learning model input;

[0030] Select a convolutional neural network-based architecture and build a deep learning model based on the selected architecture;

[0031] The labeled data is divided into training set, validation set and test set. The training set is used to train the model, and the test set is used to evaluate the trained model.

[0032] In a preferred embodiment, step S200 specifically includes the following contents:

[0033] Extract the image elements in the page or the scanned page as image files;

[0034] Grayscale, binarize and denoise the image, use the trained OCR recognition model to perform character recognition on the processed image, and output text information;

[0035] The acquisition logic of the OCR recognition model is:

[0036] Combined with long short-term memory network, an OCR recognition model is constructed.

[0037] In a preferred embodiment, step S300 specifically includes the following contents:

[0038] Distinguish text and image content based on layout information and OCR recognition results;

[0039] Reconstruct the logical structure of paragraphs, chapters, and tables of a document using natural language processing technology and page layout information;

[0040] Format the recognized text, adjust the font, size, color, and correct typos and punctuation.

[0041] In a preferred embodiment, step S400 specifically includes the following contents:

[0042] Automatically correct OCR recognition errors using contextual information and deep learning models;

[0043] Identify and extract structured data from tables and forms and format them;

[0044] Prepare output data according to user configuration.

[0045] In a preferred embodiment, step S500 specifically includes the following contents:

[0046] Provide multiple output format options for users to choose;

[0047] Supports storing analysis results in local disk, cloud storage and database.

[0048] In a preferred embodiment, an electronic device includes:

[0049] Processor and memory;

[0050] The processor is used to execute the steps of any of the above-mentioned methods for OCR recognition and parsing of PDF texts and images by calling the program or instructions stored in the memory.

[0051] In a preferred embodiment, a computer-readable storage medium stores a program or instruction, wherein the program or instruction enables a computer to execute the steps of any of the above-mentioned methods for OCR-based recognition and parsing of PDF text and images.

[0052] The technical effects and advantages of the method, device and storage medium for OCR recognition and parsing of PDF text and images of the present invention are as follows:

[0053] 1. By integrating advanced OCR technology, the parsing accuracy of complex text and image content in PDF documents is significantly improved, and the OCR recognition error rate is reduced. At the same time, intelligent page layout analysis and content reorganization technology make the parsing process more efficient and reduce the need for manual intervention;

[0054] 2. It provides a wealth of configuration options and API interfaces, allowing users to adjust parsing strategies according to specific needs and achieve customized development and services. In addition, by continuously updating the OCR model, it can continuously adapt to new document types and parsing requirements, maintaining good scalability;

[0055] 3. It helps to promote the digitization process of PDF documents and improve the efficiency of information processing and utilization. At the same time, the parsed structured data provides strong support for subsequent intelligent applications such as text analysis and data mining, and promotes the intelligent upgrade of document management;

[0056] 4. It greatly saves manpower and time costs and improves work efficiency. Especially when processing a large number of PDF documents, its automation and intelligent features are more prominent, bringing significant economic benefits to enterprises and individual users.

[0057] The method, device and storage medium for OCR recognition and parsing of PDF text and images proposed in the present invention realize efficient and accurate parsing of complex text and image content in PDF documents by integrating advanced OCR technology and intelligent parsing algorithm. This technical solution not only improves the parsing accuracy and efficiency, enhances the flexibility and scalability of the system, but also promotes the development of document digitization and intelligent application. At the same time, it brings significant economic benefits to enterprises and individual users by saving resources and time costs. In the future, with the continuous advancement of technology and the expansion of application scenarios, the present invention is expected to play a greater role in document processing, data mining, information retrieval and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 The overall structure of the present invention Figure 1 .

[0059] Figure 2 The overall structure of the present invention Figure 2 .

[0060] Figure 3 It is a PDF preprocessing flow chart of the present invention.

[0061] Figure 4 It is the OCR recognition flow chart of the present invention.

[0062] Figure 5 It is a flow chart of content analysis and reorganization of the present invention. DETAILED DESCRIPTION

[0063] 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, not all of the embodiments. 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.

[0064] Example 1

[0065] Reference Figure 1 , Figure 2 , Figure 3 , Figure 4 and Figure 5 The present invention proposes a method for parsing PDF text and images based on OCR recognition, including the following steps:

[0066] Step S100, loading a PDF document, parsing the page, extracting page elements, and processing page layout information;

[0067] Step S200, for the extracted image or the scanned PDF page, high-precision OCR recognition is performed using a deep learning model to convert the text in the image into editable text;

[0068] Step S300, based on the layout information of the PDF page and the OCR recognition result, intelligently distinguish text and images, reorganize the content, and restore the logical structure of the document;

[0069] Step S400, formatting the parsed text, such as correcting OCR recognition errors, adjusting text paragraphs, and identifying and extracting key information;

[0070] Step S500 provides multiple output format options and supports storing the analysis results in a specified storage medium.

[0071] Step S100 specifically includes the following contents:

[0072] Use the PDF parsing library to load PDF documents and read page contents;

[0073] Traverse each page, extract the text, pictures and table elements in the page, and record their position, size and type attributes;

[0074] Apply deep learning models to identify page layouts and determine the layout structure of text blocks, image areas, and table frames;

[0075] The acquisition logic of the deep learning model is:

[0076] Collect a large number of annotated PDF page samples and convert them into a format suitable for deep learning model input;

[0077] Select a convolutional neural network-based architecture and build a deep learning model based on the selected architecture;

[0078] The labeled data is divided into training set, validation set and test set. The training set is used to train the model, and the test set is used to evaluate the trained model.

[0079] It should be added that it can automatically identify the structural features of PDF pages, such as paragraph division, image position, table borders, etc., to provide accurate layout information for subsequent content analysis. By training a large number of annotated PDF page samples, the model can learn the characteristics of different layout modes and achieve efficient and accurate layout recognition.

[0080] By collecting a large number of annotated PDF page samples for training, the model can learn the various feature differences of different elements such as text blocks, image areas, and table frames in page layout. For example, text blocks often have regular text arrangement, image areas have unique texture and color features, and table frames have regular row and column structures. Based on the powerful feature extraction capabilities of convolutional neural networks (CNN), the model can accurately identify these features, accurately distinguish different elements on the page, and reduce the occurrence of misjudgments.

[0081] Accurately determine the layout structure of each element, including its location, size and other attributes. This is critical for further processing of PDF documents (such as re-layout, extracting content from specific areas, etc.), ensuring that operations are based on accurate layout information and improving the accuracy and effectiveness of document processing.

[0082] CNN has the ability to automatically extract features from image data (here, the converted format of PDF pages is treated as image data). Its convolution layer can capture multi-level features of text, pictures, tables, etc. from local to overall by sliding the image through convolution kernels of different sizes, such as the stroke shape of text characters, the edge contour of pictures, and the direction of table lines. This automatic feature extraction method is more efficient and comprehensive than the traditional method of manually designing features, and can dig out hidden features that are difficult for humans to intuitively define and extract, which helps to improve the accuracy of layout recognition. As the number of network layers increases, CNN can gradually abstract more advanced and semantic features, such as being able to summarize the features representing paragraphs and chapters from the features of many text blocks, so as to better understand the overall layout logic of the page and accurately divide different layout areas.

[0083] Step S200 specifically includes the following contents:

[0084] Extract the image elements in the page or the scanned page as image files;

[0085] Grayscale, binarize and denoise the image, use the trained OCR recognition model to perform character recognition on the processed image, and output text information;

[0086] The acquisition logic of the OCR recognition model is:

[0087] Combined with long short-term memory network, an OCR recognition model is constructed.

[0088] It should be added that the model not only has powerful character recognition capabilities, but can also effectively process texts with different fonts, sizes, colors, tilt angles and background complexities. Through transfer learning and data enhancement strategies, the model's recognition accuracy for low-quality images and special characters can be further improved.

[0089] For text recognition, there is a sequential relationship between characters. Long short-term memory (LSTM) networks can effectively process this sequence information. It can remember the information of previously recognized characters and use it for the recognition of subsequent characters. For example, when recognizing a word, LSTM can use the information of the previous characters to better infer the following characters, especially for some words with contextual associations, such as "although", the previous "al" can help the model better predict the following part.

[0090] In a document, the layout and format of text may be complex, including different fonts, font sizes, tilt angles, etc. LSTM can adapt to this complex text structure because it has the ability to process long sequences and can maintain good recognition performance under different text changes. For example, when recognizing handwritten text or documents with artistic fonts, LSTM can recognize based on the overall sequence characteristics of the characters, rather than relying solely on the shape of a single character.

[0091] Step S300 specifically includes the following contents:

[0092] Distinguish text and image content based on layout information and OCR recognition results;

[0093] Reconstruct the logical structure of paragraphs, chapters, and tables of a document using natural language processing technology and page layout information;

[0094] Format the recognized text, adjust the font, size, color, and correct typos and punctuation.

[0095] It should be added that the natural language processing (NLP) technology, such as semantic analysis and syntactic parsing, combined with the logical structure information of the PDF page, can be used to intelligently reorganize the text recognized by OCR to restore the original structure and semantics of the document. At the same time, through rule matching and machine learning algorithms, structured data such as tables and forms can be identified and extracted, and formatted for subsequent data analysis and processing.

[0096] Step S400 specifically includes the following contents:

[0097] Automatically correct OCR recognition errors using contextual information and deep learning models;

[0098] Identify and extract structured data from tables and forms and format them;

[0099] Prepare output data according to user configuration.

[0100] It should be added that the deep learning model can use rich contextual semantic information to correct OCR errors. Compared with traditional rule-based error correction methods, it can handle more complex semantic relationships and language changes. For example, for some words with similar spellings but different semantics, such as "rights" and "power", they can be accurately judged and corrected through contextual understanding. It can adapt to a variety of text types and language styles. Whether it is a formal business document, an academic paper, or a more casual social media text, the deep learning model can correct errors based on the learned language patterns, effectively reducing errors caused by differences in text style.

[0101] The automatic correction function can quickly process a large amount of OCR recognized text, greatly saving the time of manual inspection and error correction. Especially for large-scale document processing projects, it can significantly improve work efficiency. The deep learning model can continue to learn and optimize in the background. As the amount of processed text increases, its error correction ability will continue to increase, thereby further improving error correction efficiency and quality.

[0102] Automatic identification and processing can reduce data errors caused by human factors and ensure the consistency of table and form data. In the process of data integration and aggregation, high-quality data can avoid problems caused by inconsistent formats or data errors and improve the overall quality of data.

[0103] Step S500 specifically includes the following contents:

[0104] Provide multiple output format options for users to choose;

[0105] Supports storing analysis results in local disk, cloud storage and database.

[0106] It should be added that it provides a wealth of configuration options, allowing users to adjust OCR recognition parameters, output format, parsing strategy, etc. according to specific needs. At the same time, it opens the API interface to support third-party application integration and realize customized development and services.

[0107] An electronic device, comprising:

[0108] Processor and memory;

[0109] The processor is used to execute the steps of any of the above-mentioned methods for OCR recognition and parsing of PDF texts and images by calling the program or instructions stored in the memory.

[0110] A computer-readable storage medium stores a program or instruction, wherein the program or instruction enables a computer to execute the steps of any of the above-mentioned methods for parsing PDF text and images based on OCR recognition.

[0111] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0112] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0113] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0114] In the several embodiments provided in the present application, it should be understood that the disclosed methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.

[0115] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0116] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0117] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0118] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0119] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for parsing PDF text and images based on OCR recognition, characterized in that: The steps include: Step S100, loading a PDF document, parsing the page, extracting page elements, and processing page layout information; Step S200, for the extracted image or the scanned PDF page, high-precision OCR recognition is performed using a deep learning model to convert the text in the image into editable text; Step S300, based on the layout information of the PDF page and the OCR recognition result, intelligently distinguish text and images, reorganize the content, and restore the logical structure of the document; Step S400, formatting the parsed text, such as correcting OCR recognition errors, adjusting text paragraphs, and identifying and extracting key information; Step S500 provides multiple output format options and supports storing the analysis results in a specified storage medium.

2. The method for analyzing PDF text and images based on OCR recognition according to claim 1, characterized in that: Step S100 specifically includes the following contents: Use the PDF parsing library to load PDF documents and read page contents; Traverse each page, extract the text, pictures and table elements in the page, and record their position, size and type attributes; Apply deep learning models to identify page layouts and determine the layout structure of text blocks, image areas, and table frames; The acquisition logic of the deep learning model is: Collect a large number of annotated PDF page samples and convert them into a format suitable for deep learning model input; Select a convolutional neural network-based architecture and build a deep learning model based on the selected architecture; The labeled data is divided into training set, validation set and test set. The training set is used to train the model, and the test set is used to evaluate the trained model.

3. The method for analyzing PDF text and images based on OCR recognition according to claim 2 is characterized in that: Step S200 specifically includes the following contents: Extract the image elements in the page or the scanned page as image files; Grayscale, binarize and denoise the image, use the trained OCR recognition model to perform character recognition on the processed image, and output text information; The acquisition logic of the OCR recognition model is: Combined with long short-term memory network, an OCR recognition model is constructed.

4. The method for analyzing PDF text and images based on OCR recognition according to claim 3 is characterized in that: Step S300 specifically includes the following contents: Distinguish text and image content based on layout information and OCR recognition results; Reconstruct the logical structure of paragraphs, chapters, and tables of a document using natural language processing technology and page layout information; Format the recognized text, adjust the font, size, color, and correct typos and punctuation.

5. The method for analyzing PDF text and images based on OCR recognition according to claim 4 is characterized in that: Step S400 specifically includes the following contents: Automatically correct OCR recognition errors using contextual information and deep learning models; Identify and extract structured data from tables and forms and format them; Prepare output data according to user configuration.

6. The method for analyzing PDF text and images based on OCR recognition according to claim 5, characterized in that: Step S500 specifically includes the following contents: Provide multiple output format options for users to choose; Supports storing analysis results in local disk, cloud storage and database.

7. An electronic device, characterized in that: The electronic device comprises: Processor and memory; The processor is used to execute the steps of the method for parsing PDF text and images based on OCR recognition as described in any one of claims 1 to 6 by calling the program or instruction stored in the memory.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program or instruction, and the program or instruction enables a computer to execute the steps of the method for parsing PDF text and images based on OCR recognition as described in any one of claims 1 to 6.

Images