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

Natural scene text position detection method based on image segmentation

A natural scene and image segmentation technology, applied in the field of target detection, can solve problems such as poor results

Pending Publication Date: 2019-07-26
TIANJIN UNIV
View PDF5 Cites 6 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Applying scanned text algorithms to this type of image yields terrible results

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
  • Natural scene text position detection method based on image segmentation
  • Natural scene text position detection method based on image segmentation
  • Natural scene text position detection method based on image segmentation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0016] In order to make the technical solution of the present invention more clear, the specific implementation manners of the present invention will be further described below in conjunction with the accompanying drawings.

[0017] Step 1: The present invention uses the data set of the first task (challenge4 Task1) of the fourth challenge in the picture ICDAR2015 as a training set and a test set, including a total of 1500 pictures. 1000 images are used for training and 500 images are used for testing. The images are captured by Google Glass (Google Class). Random shooting is adopted, and the image size is uniformly 1280*760. The lens did not focus on the text content during the shooting. After the shooting was completed, the pictures containing text were selected from the randomly captured images and marked. Therefore, the text position of the ICDAR2015 dataset is relatively random, the content is blurred, and the text direction is uncertain. Text annotation is also based o...

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 relates to a natural scene text position detection method based on image segmentation. The method comprises the following steps: selecting a data set; selecting and marking images containing characters, and making a training set and a test set; data enhancement: 1, random rotation: randomly rotating the picture at a probability of 0.25; 2, random cutting; 3, random color disturbance;constructing a training neural network, wherein the VGG16 is used as a basic network; a full connection layer of the network is changed into a convolutional layer, the output of Pooling5 is kept unchanged, Fc6 is changed into a 7 * 7 * 512 convolutional layer, Fc7 and Fc8 are changed into a convolutional layer by using a 1 * 1 convolutional kernel, pixel-by-pixel prediction is carried out on thetext according to a segmentation idea, and finally, a prediction result of each pixel point is obtained; designing a loss function; and carrying out post-processing on the training result.

Description

technical field [0001] The invention belongs to the technical field of target detection, and relates to a method for detecting text positions in natural scene images based on deep learning technology. Background technique [0002] Text detection algorithm is a branch of Optical Character Recognition (OCR) field. The original OCR technology is to scan this text, which is characterized by high resolution, neat and regular arrangement of text, simple background, and a large proportion of text area in the overall image. Scanned text recognition rate reached 97.38% [1] . With the widespread popularization of digital cameras, the algorithm of scanning text cannot meet the needs of the society gradually. The quality of the detected natural image is degraded, the background is complex, the text direction is uncertain and the proportion of the overall image is small, such as recognizing road signs contained in street view photos or characters contained in street nameplates. Apply...

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): G06K9/00G06K9/20G06K9/32G06K9/34G06K9/62G06N3/04
CPCG06V30/40G06V10/242G06V10/22G06V30/153G06N3/045G06F18/214
Inventor 侯春萍杨阳徐金辰夏晗
Owner TIANJIN UNIV
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