Character detection method and device based on single characters and inter-character connection relationship prediction

A connection relationship, text detection technology, used in character and pattern recognition, image data processing, instruments, etc.

Pending Publication Date: 2020-10-20
北京思图场景数据科技服务有限公司
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, when using this method, it is necessary to use pr

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
  • Character detection method and device based on single characters and inter-character connection relationship prediction
  • Character detection method and device based on single characters and inter-character connection relationship prediction
  • Character detection method and device based on single characters and inter-character connection relationship prediction

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0051] In the first aspect, the present embodiment provides a text detection method based on single character and connection relationship prediction between text, including the following steps:

[0052] Training neural network model; Carry out feature extraction by described neural network model, obtain feature map, described feature map comprises character position score map Text-map and connection relationship score map Link-map between characters, to Text-map and Link-map Carry out post-processing, the step of described post-processing is as follows:

[0053] Convert the Text-map into a binary image Bm1, set the threshold λ1 for the Text-map,

[0054] Convert the Link-map into a binary image Bm2, set the threshold λ2 for the Link-map,

[0055] Initialize the binary map Bm1 to 0, and when the feature map value of the corresponding position is greater than the threshold λ1, the position value on the binary map Bm1 is 1;

[0056] Initialize the binary map Bm2 to 0, and when ...

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 character detection method and device based on single characters and inter-character connection relationship prediction. The method comprises the following steps of: traininga neural network model; performing feature extraction through the neural network model to obtain a feature map which comprises a character position score map Text-map and an inter-character connection relationship score map Link-map; carrying out post-processing on the Text-map and the Link-map; and calculating the minimum bounding rectangle of each connected domain to realize detection of the position of a character from a picture. The convolution kernel of the prediction model based on the character level can be forced to pay more attention to the character-level features, the character-level features learned by the model can be effectively associated at the post-processing branch, long text detection is achieved, and missing detection of characters is avoided.

Description

technical field [0001] The invention relates to the field of character detection, and specifically relates to a character detection method based on prediction of a single character and connection relationship between characters. Background technique [0002] Before the outbreak of neural network technology, text detection was usually based on manually set feature extraction methods, such as regional feature extraction MSER, programming framework SWT (Standard Widget Toolkit), etc. as basic components. In recent years, with the development of neural network technology, more and more new methods based on neural network have become popular. Some new text detection methods have been designed successively, but in general these methods are essentially improved based on general target detection or general instance segmentation methods, such as detecting text in natural images (SingleShotDetection, SSD), FasterR- CNN, Fully Convolutional Network (FCN). We can divide it into two ma...

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): G06T7/187G06T7/136G06T7/11G06T7/00G06K9/34
CPCG06T7/187G06T7/11G06T7/0002G06T7/136G06T2207/20081G06T2207/20084G06V30/153G06V30/287
Inventor 吕馗汪明浩
Owner 北京思图场景数据科技服务有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap