Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Method and device for extracting structured information of oversized image

A technology of structured information and extraction method, which is applied in the field of structured information extraction of super large images, can solve problems such as difficult to recognize text, increased difficulty of structured information extraction, no information structured extraction, etc., to expand the recognition area Effect

Pending Publication Date: 2022-05-27
上海云扩信息科技有限公司
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

For design drawings similar to advertisements, the rate can be in the range of 800–2000. In the prior art, it is difficult to recognize text in images with a rate value exceeding 1000.
[0004] On the other hand, the current text recognition technology simply recognizes the text information line by line, and does not perform structured extraction of the recognized information.
Especially for some complex and diverse documents, it is more difficult to extract structured information

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Method and device for extracting structured information of oversized image
  • Method and device for extracting structured information of oversized image
  • Method and device for extracting structured information of oversized image

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0053] In this embodiment, refer to the figure 1 to the attached Figure 4 , the technical problem to be solved by this embodiment is:

[0054] 1) The position of the text in the image with high rate value cannot be detected;

[0055] 2) The current text recognition technology simply recognizes text information line by line, and does not perform structured extraction of the recognized information.

[0056] Provided is a method for extracting structured information of super large images, the method for extracting structured information of super large images includes:

[0057] Locate the text area position through the neural network model;

[0058] Identify the ROI area in the text area;

[0059] Identify the text blocks in the ROI area and arrange the text blocks;

[0060] Define the tokens required to extract structured information, and arrange the tokens according to the arrangement rules;

[0061] Reduce the dimensionality of the identified text block into a one-dimens...

Embodiment 2

[0094] In this embodiment, refer to the figure 1 to the attached Figure 4 , the technical problem to be solved by this embodiment is:

[0095] 1) The position of the text in the image with high rate value cannot be detected;

[0096] 2) The current text recognition technology simply recognizes text information line by line, and does not perform structured extraction of the recognized information.

[0097] Provided is a method for extracting structured information of super large images, the method for extracting structured information of super large images includes:

[0098] Locate the text area position through the neural network model;

[0099] Identify the ROI area in the text area;

[0100] Identify the text blocks in the ROI area and arrange the text blocks;

[0101] Define the tokens required to extract structured information, and arrange the tokens according to the arrangement rules;

[0102] Reduce the dimensionality of the identified text block into a one-dimens...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a structured information extraction method and device for a super-large image. The structured information extraction method for the super-large image comprises the following steps: positioning a character region position through a neural network model; identifying an ROI (Region of Interest) area in the character area; identifying character blocks in the ROI region, and arranging the character blocks; the method comprises the following steps: defining tokens required for extracting structured information, and arranging the tokens according to an arrangement rule; the recognized character blocks are subjected to dimension reduction to form one-dimensional long sentences, the token array is traversed, and the specific positions of the tokens are found in the one-dimensional long sentences; according to the positioned tokens, respectively extracting corresponding structured information of the positioned tokens; the method has the advantages that structured information extraction can be carried out on the oversized image, and the extraction precision is high.

Description

technical field [0001] The invention relates to the field of image character recognition, in particular to a method and device for extracting structured information of super large images. Background technique [0002] The current OCR technology has been able to successfully extract the text content of the image. But for some specific images, such as advertisement design drawings, the image size is dozens of times that of normal text images. Using conventional OCR technology, the position of the text can hardly be detected, let alone subsequent recognition. [0003] To describe this phenomenon more clearly, we define rate = the longest side of the image / the height of a single character. For regular documents, the rate is in the range of about 50-200. For design drawings similar to advertisements, the rate can be in the range of 800-2000. In the prior art, it is difficult to recognize text in images with a rate value exceeding 1000. [0004] On the other hand, the curren...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06V10/25G06V10/22G06V30/148G06V10/774G06K9/62G06N3/02
CPCG06N3/02G06F18/214
Inventor 刘春刚史秋芳李佩钊侯乐
Owner 上海云扩信息科技有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com