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

Damaged two-dimensional code recovery method of convolutional auto-encoder in combination with binary segmentation

A convolutional self-encoding and binary segmentation technology, which is applied in the field of damaged two-dimensional code restoration, achieves the effects of sufficient data set types, improved image quality, and strong model generalization ability

Active Publication Date: 2020-10-16
CHENGDU UNIVERSITY OF TECHNOLOGY +1
View PDF4 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0010] The purpose of the present invention is to provide a solution to the above problems, which can realize end-to-end repair of two-dimensional code images with blur, non-uniform illumination, noise, and simultaneous occurrence of the above problems, combined with binary segmentation convolutional self-encoding The recovery method of the damaged QR code of the device

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
  • Damaged two-dimensional code recovery method of convolutional auto-encoder in combination with binary segmentation
  • Damaged two-dimensional code recovery method of convolutional auto-encoder in combination with binary segmentation
  • Damaged two-dimensional code recovery method of convolutional auto-encoder in combination with binary segmentation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0059] Example 1: see Figure 1 to Figure 2 , A method for recovering damaged two-dimensional codes combined with binary segmentation convolutional autoencoder, including the following steps:

[0060] (1) Prepare the simulated data set and the real data set, together as the training data set, among which,

[0061] Simulation experiment data set: Simulation generates multiple clear QR code images with different data content and styles, and adds different degrees of blur, shading and / or additive noise to them to obtain multiple damaged QR code images;

[0062] Real collection of data sets: use the camera to shoot, keep the relative position of the QR code unchanged, take a clear image of the QR code, then adjust the camera parameters and control its jitter, shoot multiple images with varying degrees of blur and varying degrees of damage Dimension code image;

[0063] Use the clear QR code image as the label corresponding to the damaged QR code image, and the resolution of the damaged QR...

Embodiment 2

[0072] Example 2: see Figure 1 to Figure 5 To better explain the solution of the present invention, on the basis of Example 1, we add the following technical features:

[0073] The binary classification is calculated using the following formula

[0074]

[0075] Where x i Represents the feature representation of the channel i of the output tensor, n represents the two categories of black and white pixels, the value is 2, p(x i ) Is the output probability of the corresponding channel;

[0076] The cross entropy loss function is

[0077]

[0078] Where w(x i ) Is the label of the pixel of the damaged QR code image. The rest is the same as in Example 1.

Embodiment 3

[0079] Example 3: See Figure 1 to Figure 5 In order to better explain the solution of the present invention, on the basis of Example 1, we specifically describe as follows:

[0080] (1) Prepare the simulated data set and the real data set, together as the training data set, among which,

[0081] Simulation experiment data set: The simulation generates multiple clear QR code images with different data content and styles, and adds different degrees of blur, shading and / or additive noise to them to obtain multiple damaged QR code images; The parameter settings of the damaged QR code image are shown in Table 1 below:

[0082] Table 1: Parameter setting range of simulated damaged QR code image

[0083]

[0084] The brightness and dark pixel adjustment ratio parameters of the simulated QR code image are set to 0.8 to 0.9 and 0.1 to 0.2, respectively; the blur radius of the out-of-focus blur is set to 7 to 9; the pixel blur length of the motion blur is set to 18 to 22. Set the blur angle f...

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 damaged two-dimensional code recovery method of a convolutional auto-encoder in combination with binary segmentation. The method comprises the steps of preparing a training data set; constructing a deep convolutional self-encoding neural network, wherein the deep convolutional self-encoding neural network comprises an encoder, a decoder and a binary segmentation layer, wherein the decoder adopts an up-sampling part of a U-net network, and the binary segmentation layer is used for carrying out binary classification on each feature element point in a feature tensor output by the decoder according to black and white pixels; wherein the loss function adopts a cross entropy loss function; and finally training a model for image restoration. According to the method, a convolution auto-encoder, U-net, a binary segmentation layer and the like are organically combined, and finally end-to-end restoration can be carried out on fuzzy and non-uniform illumination, noise andtwo-dimensional code images with the problems at the same time.

Description

Technical field [0001] The invention relates to a method for restoring a two-dimensional code image, in particular to a method for restoring a damaged two-dimensional code combined with a convolutional autoencoder combined with binary segmentation. Background technique [0002] In real life, the QR code recognition process is susceptible to many factors. Camera motion blur, out-of-focus blur, uneven lighting and random noise will cause the problem that the QR code cannot be recognized and the image is damaged. The way is usually not a single occurrence. In dark areas, camera shake and blur, non-uniform illumination and noise factors often appear at the same time, which greatly reduces the user experience and recognition rate. Therefore, it is necessary to reduce noise and restore the QR code during the recognition process. , To achieve the recognition effect. The current methods for QR code recovery are: [0003] (1) Chen Kecheng and others proposed a damaged QR code recovery al...

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
Patent Type & Authority Applications(China)
IPC IPC(8): G06K7/14G06T5/00G06T7/10G06N3/04G06N3/08
CPCG06K7/1482G06K7/1417G06T7/10G06N3/088G06T2207/10004G06T2207/20081G06T2207/20084G06N3/045G06T5/73G06T5/77G06T5/70
Inventor 王向鹏林凡强强孙源
Owner CHENGDU UNIVERSITY OF TECHNOLOGY
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