CN117034982A - Two-dimensional code identification method and device with automatic repair function - Google Patents
Two-dimensional code identification method and device with automatic repair function Download PDFInfo
- Publication number
- CN117034982A CN117034982A CN202311303893.2A CN202311303893A CN117034982A CN 117034982 A CN117034982 A CN 117034982A CN 202311303893 A CN202311303893 A CN 202311303893A CN 117034982 A CN117034982 A CN 117034982A
- Authority
- CN
- China
- Prior art keywords
- image
- dimensional code
- identified
- diagram
- module
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 230000008439 repair process Effects 0.000 title claims abstract description 26
- 238000010586 diagram Methods 0.000 claims abstract description 67
- 230000006870 function Effects 0.000 claims description 27
- 230000004927 fusion Effects 0.000 claims description 14
- 238000013256 Gubra-Amylin NASH model Methods 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 7
- 238000005215 recombination Methods 0.000 claims description 5
- 230000006798 recombination Effects 0.000 claims description 5
- 238000001228 spectrum Methods 0.000 claims description 4
- 230000003595 spectral effect Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 12
- 230000015556 catabolic process Effects 0.000 abstract description 4
- 238000006731 degradation reaction Methods 0.000 abstract description 4
- 230000009977 dual effect Effects 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000005461 lubrication Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000032258 transport Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
- G06K7/00—Methods or arrangements for sensing record carriers, e.g. for reading patterns
- G06K7/10—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation
- G06K7/14—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation using light without selection of wavelength, e.g. sensing reflected white light
- G06K7/1404—Methods for optical code recognition
- G06K7/146—Methods for optical code recognition the method including quality enhancement steps
- G06K7/1473—Methods for optical code recognition the method including quality enhancement steps error correction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
- G06K7/00—Methods or arrangements for sensing record carriers, e.g. for reading patterns
- G06K7/10—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation
- G06K7/14—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation using light without selection of wavelength, e.g. sensing reflected white light
- G06K7/1404—Methods for optical code recognition
- G06K7/1408—Methods for optical code recognition the method being specifically adapted for the type of code
- G06K7/1417—2D bar codes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/86—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using syntactic or structural representations of the image or video pattern, e.g. symbolic string recognition; using graph matching
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Electromagnetism (AREA)
- Toxicology (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The application relates to the field of image recognition, in particular to a two-dimensional code recognition method and device with an automatic repair function, which are used for judging whether materials on a conveyor belt reach a preset position or not and acquiring a two-dimensional code image to be recognized of the materials reaching the preset position; performing super-resolution reconstruction on the two-dimensional code image to be identified to obtain a noise-free and blur-free image and a high-resolution image; respectively inputting the noiseless and unambiguous image and the high-resolution image into a channel attention module and a double attention module to obtain a thermodynamic channel diagram and a branch output diagram; and fusing the thermodynamic channel diagram and the branch output diagram to obtain a fused feature diagram of the two-dimensional code image to be identified, and decoding the two-dimensional code in the fused feature diagram by adopting ZXing to obtain target information. The method has the advantages that the degradation phenomenon in the image restoration process can be avoided, the restored image is clear and not blurred, the restored two-dimensional code is clearly identified, and the information is conveniently and accurately extracted from the two-dimensional code.
Description
Technical Field
The application relates to the technical field of image recognition, in particular to a two-dimensional code recognition device with an automatic repair function.
Background
The global economy tends to be integrated nowadays, and industrial production is developing towards intellectualization and automation. In this context, in order to accurately record and trace each production link, stream and operation of a product, an efficient and accurate device is required to read and record product information. The device is two-dimension code identification equipment, and can read two-dimension codes on products and upload the information to a tracing system of a factory so as to realize the whole process tracing of each product.
The two-dimensional code recognition method adopted by the existing two-dimensional code recognition equipment is mainly divided into two types, namely image processing and template matching, wherein the method based on the image processing is to extract and decode the characteristic information of the two-dimensional code by processing and analyzing the image. Common algorithms include graying, binarizing, edge detection, positioning pattern recognition, etc. The method based on the template matching refers to matching the known template by utilizing a specific mode of the two-dimensional code, thereby realizing decoding. The two methods have high recognition speed and recognition accuracy when the two-dimensional code is clean and clear. However, in an industrial scene, two-dimensional code offset is common due to friction of a conveyor belt, lubrication oil stains of a processing machine, and the like. The two-dimensional code is stained, so that the existing identification method fails in identification, even if the two-dimensional code can be identified, the extracted information can be deviated, and the identification speed can be seriously reduced.
Disclosure of Invention
In order to solve the problems in the prior art, the application provides a two-dimensional code identification device with an automatic repair function.
According to a first aspect of the present application, the present application provides a two-dimensional code identification method with an automatic repair function, which is characterized by comprising:
judging whether the material on the conveyor belt reaches a preset position or not, and acquiring a two-dimensional code image to be identified of the material reaching the preset position;
performing super-resolution reconstruction on the two-dimensional code image to be identified to obtain a noise-free and blur-free image and a high-resolution image of the two-dimensional code image to be identified;
respectively inputting the noiseless and unambiguous image and the high-resolution image into a channel attention module and a double attention module to obtain a thermodynamic channel diagram and a branch output diagram of the two-dimensional code image to be identified;
and fusing the thermal channel diagram and the branch output diagram to obtain a fused feature diagram of the two-dimensional code image to be identified, and decoding the two-dimensional code in the fused feature diagram by ZXing to obtain target information.
Further, the step of determining whether the material on the conveyor belt reaches the preset position or not, and the step of obtaining the two-dimensional code image to be identified of the material reaching the preset position specifically includes:
the mechanical arm places the materials on a conveyor belt, and the conveyor belt conveys the materials to a preset position for image shooting;
and judging whether the material reaches a preset position by adopting an infrared sensor, and when the material reaches the preset position, carrying out image shooting on the material by using a camera to obtain a two-dimensional code image to be identified.
Further, the performing super-resolution reconstruction on the two-dimensional code image to be identified to obtain a noise-free and blur-free image and a high-resolution image of the two-dimensional code image to be identified specifically includes:
inputting the two-dimensional code image to be identified into a GAN model generator to obtain a noise-free and blur-free image of the two-dimensional code image to be identified;
and inputting the noiseless and fuzzy-free image into a high-frequency learning layer and an up-sampling layer to obtain a high-resolution image.
Further, the GAN model generator includes 2 channel attention modules, 4 residual modules, and 1 self-attention module;
the high frequency learning layer comprises 3 dense network blocks;
the upsampling layer includes 2 convolutional layers, 1 spectral normalization layer, 1 pixel rebinning layer, and LeakyRelu.
Further, the inputting the noiseless and unambiguous image and the high-resolution image into a channel attention module and a dual attention module respectively to obtain a thermal channel diagram and a branch output diagram of the two-dimensional code image to be identified specifically includes: taking the high-resolution image as a branch input convolution layer and a double-attention module to obtain a branch output image;
and inputting the noiseless and unambiguous image into a convolution layer and U-Net to obtain a thermodynamic diagram, and inputting the thermodynamic diagram into two channel attention modules to obtain a thermodynamic channel diagram.
Further, the fusing the thermal channel diagram and the branch output diagram to obtain the fused feature diagram of the two-dimensional code image to be identified specifically includes:
summing the thermodynamic channel diagram and the branch output diagram to obtain a fusion characteristic diagram;
and inputting the fusion feature map into a convolution layer to obtain a repaired image.
Further, when the GAN model generator is in the working process, the method specifically includes:
and sequentially inputting the two-dimensional code image to be identified into a first channel attention module, a first residual error module, a downsampling to a second residual error module, a downsampling to a self-attention module, an upsampling to a third residual error module, an upsampling to a fourth residual error module and a second channel attention module to obtain a noise-free and blur-free image of the two-dimensional code image to be identified.
Further, when the high-frequency learning layer is in the working process, the method specifically includes:
and inputting the noiseless and unambiguous images of the two-dimensional code image to be identified into a first dense network module, a second dense network module and a third dense network module in sequence to obtain a high-resolution intermediate image.
Further, when the upsampling layer is in the working process, the upsampling layer specifically includes:
and sequentially inputting the high-resolution intermediate image into a first convolution layer, a spectrum normalization layer, a pixel recombination layer, a LeakyRelu and a second convolution layer, and outputting to obtain the high-resolution image.
According to the second module of the application, the application provides a two-dimensional code identification device with an automatic repair function, which comprises:
the image acquisition module is used for judging whether the material on the conveyor belt reaches a preset position or not and acquiring a two-dimensional code image to be identified of the material reaching the preset position;
the image super-division module performs super-division reconstruction on the two-dimensional code image to be identified to obtain a noise-free and blur-free image and a high-resolution image of the two-dimensional code image to be identified;
the image restoration module is used for inputting the noiseless and fuzzy-free image and the high-resolution image into a channel attention module and a double attention module respectively to obtain a thermal channel image and a branch output image of the two-dimensional code image to be identified, and fusing the thermal channel image and the branch output image to obtain a fused feature image of the two-dimensional code image to be identified;
and the two-dimensional code identification module decodes the two-dimensional code in the fusion feature map by adopting ZXing and obtains target information.
The application relates to the field of image recognition, in particular to a two-dimensional code recognition method and device with an automatic repair function, which are used for judging whether materials on a conveyor belt reach a preset position or not and acquiring a two-dimensional code image to be recognized of the materials reaching the preset position; performing super-resolution reconstruction on the two-dimensional code image to be identified to obtain a noise-free and blur-free image and a high-resolution image; respectively inputting the noiseless and unambiguous image and the high-resolution image into a channel attention module and a double attention module to obtain a thermodynamic channel diagram and a branch output diagram; and fusing the thermodynamic channel diagram and the branch output diagram to obtain a fused feature diagram of the two-dimensional code image to be identified, and decoding the two-dimensional code in the fused feature diagram by adopting ZXing to obtain target information. The method has the advantages that the degradation phenomenon in the image restoration process can be avoided, the restored image is clear and not blurred, the restored two-dimensional code is clearly identified, and the information is conveniently and accurately extracted from the two-dimensional code.
Drawings
FIG. 1 is a workflow diagram of a two-dimensional code recognition method with an automatic repair function;
FIG. 2 is a block diagram of a two-dimensional code recognition device with an automatic repair function;
FIG. 3 is a schematic diagram of a generator of a two-dimensional code recognition device with an automatic repair function;
FIG. 4 is a schematic diagram of resolution reconstruction of a two-dimensional code recognition device with an automatic repair function;
fig. 5 is an image restoration schematic diagram of a two-dimensional code recognition device with an automatic restoration function.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
Example 1:
referring to fig. 1, the present application provides a two-dimensional code identification method with an automatic repair function, which is characterized by comprising:
step 101, judging whether a material on a conveyor belt reaches a preset position or not, and acquiring a two-dimensional code image to be identified of the material reaching the preset position;
step 201, performing super-resolution reconstruction on the two-dimensional code image to be identified to obtain a noise-free and blur-free image and a high-resolution image of the two-dimensional code image to be identified;
step 301, inputting the noiseless and unambiguous image and the high-resolution image into a channel attention module and a dual attention module respectively to obtain a thermal channel diagram and a branch output diagram of the two-dimensional code image to be identified;
and step 401, fusing the thermal channel diagram and the branch output diagram to obtain a fused feature diagram of the two-dimensional code image to be identified, and decoding the two-dimensional code in the fused feature diagram by adopting ZXing to obtain target information.
Further, in the step 101, the method specifically includes:
the mechanical arm places the materials on a conveyor belt, and the conveyor belt conveys the materials to a preset position for image shooting;
and judging whether the material reaches a preset position by adopting an infrared sensor, and when the material reaches the preset position, carrying out image shooting on the material by using a camera to obtain a two-dimensional code image to be identified.
Further, the step 201 specifically includes:
inputting the two-dimensional code image to be identified into a GAN model generator to obtain a noise-free and blur-free image of the two-dimensional code image to be identified;
and inputting the noiseless and fuzzy-free image into a high-frequency learning layer and an up-sampling layer to obtain a high-resolution image.
Further, the GAN model generator includes 2 channel attention modules, 4 residual modules, and 1 self-attention module;
the high frequency learning layer comprises 3 dense network blocks;
the upsampling layer includes 2 convolutional layers, 1 spectral normalization layer, 1 pixel rebinning layer, and LeakyRelu.
Further, the step 301 specifically includes:
taking the high-resolution image as a branch input convolution layer and a double-attention module to obtain a branch output image;
and inputting the noiseless and unambiguous image into a convolution layer and U-Net to obtain a thermodynamic diagram, and inputting the thermodynamic diagram into two channel attention modules to obtain a thermodynamic channel diagram.
Further, in the step 401, the method specifically includes:
summing the thermodynamic channel diagram and the branch output diagram to obtain a fusion characteristic diagram;
and inputting the fusion feature map into a convolution layer to obtain a repaired image.
Further, when the GAN model generator is in the working process, the method specifically includes:
and sequentially inputting the two-dimensional code image to be identified into a first channel attention module, a first residual error module, a downsampling to a second residual error module, a downsampling to a self-attention module, an upsampling to a third residual error module, an upsampling to a fourth residual error module and a second channel attention module to obtain a noise-free and blur-free image of the two-dimensional code image to be identified.
Further, when the high-frequency learning layer is in the working process, the method specifically includes:
and inputting the noiseless and unambiguous images of the two-dimensional code image to be identified into a first dense network module, a second dense network module and a third dense network module in sequence to obtain a high-resolution intermediate image.
Further, when the upsampling layer is in the working process, the upsampling layer specifically includes:
and sequentially inputting the high-resolution intermediate image into a first convolution layer, a spectrum normalization layer, a pixel recombination layer, a LeakyRelu and a second convolution layer, and outputting to obtain the high-resolution image.
According to the embodiment, aiming at the situation that the degradation phenomenon occurs in the image restoration process, the restored and generated image is blurred and unclear, a high-resolution image is obtained according to the high-frequency learning layer and the up-sampling layer and is used as a branch input, the noiseless and blurry-free image generated by the generator is used as a main input, finally, the two inputs are overlapped to obtain a fusion feature map, and the fusion feature map is sent to the convolution layer to obtain the restored two-dimensional code image, so that the situations that the degradation phenomenon is avoided in the image restoration process, and the restored and generated image is clear and not blurred can be achieved.
Example 2
Referring to fig. 2, the present application provides a two-dimensional code recognition device with an automatic repair function, comprising:
the image acquisition module is used for judging whether the material on the conveyor belt reaches a preset position or not and acquiring a two-dimensional code image to be identified of the material reaching the preset position;
the image super-division module performs super-division reconstruction on the two-dimensional code image to be identified to obtain a noise-free and blur-free image and a high-resolution image of the two-dimensional code image to be identified;
the image restoration module is used for inputting the noiseless and fuzzy-free image and the high-resolution image into a channel attention module and a double attention module respectively to obtain a thermal channel image and a branch output image of the two-dimensional code image to be identified, and fusing the thermal channel image and the branch output image to obtain a fused feature image of the two-dimensional code image to be identified;
and the two-dimensional code identification module decodes the two-dimensional code in the fusion feature map by adopting ZXing and obtains target information.
In this embodiment, the robot subunit is responsible for placing the material on the conveyer belt, and the conveyer belt transports the material to the preset position of image shooting subunit, and the image shooting subunit comprises camera and infrared inductor, and infrared inductor is used for judging whether the material reaches preset position, and the camera is responsible for carrying out image shooting.
The image super-division module is responsible for super-division reconstruction of the shot image and consists of a denoising and deblurring subunit and a resolution reconstruction subunit. The image super-division module firstly acquires the image acquired by the image acquisition module and sends the image to the denoising and deblurring subunit. Denoising and deblurring subunit is constructed based on GAN model and generatorWith reference to fig. 3, the generator includes: 2. the method comprises the steps of a channel attention module, 4 residual modules and 1 self-attention module, wherein downsampling is realized by convolution with the size of 3*3 step length of 2, and upsampling is realized by convolution with the size of 3*3 step length of 1 plus bilinear interpolation algorithm.
The model training process is expressed as:
;
where E represents the maximum likelihood estimation operation,for the image to be restored->Is group-trunk, +.>For judging device->Generator (s)/(s)>Representing training adjustment results of the determiner and generator. />The image generating device is specifically composed of 6 convolution layers and Relu, and is responsible for judging whether the image generated by the generator meets the standard.
The loss function is:;
wherein I and j respectively represent the row and column of the pixels in the image, W and H represent the number of rows and columns of the pixels in the image,group-trunk, which represents the ith row and jth column of pixels>Representing the image to be restored of the ith row and jth column pixels. The image to be restored is an input image, comprises noise and blurring, and can generate a noise-free and blurring-free image after being input into the generator.
After the image is subjected to noise removal and blurring removal, the image is sent to a resolution reconstruction subunit, and referring to fig. 4, the resolution reconstruction subunit consists of a high-frequency learning layer and an up-sampling layer;
wherein the high frequency learning layer is composed of 3 dense network blocks, and the up-sampling layer is composed of 2 convolution layers, a spectrum normalization layer, a pixel recombination layer and a LeakyRelu. The high-frequency learning layer acquires high-frequency information of the image through dense connection, wherein the high-frequency information mainly comprises edges, textures and the like, and the edge information is a two-dimensional codeIs very important. After the high-frequency learning layer is extracted, a high-frequency characteristic diagram is generatedAnd inputting the high-frequency characteristic image into an up-sampling layer, obtaining the high-resolution characteristic image by the up-sampling layer through a convolution layer and pixel recombination, and finally obtaining the high-resolution image through a convolution layer.
The image restoration module is responsible for restoring the defects of the two-dimensional code, and referring to fig. 5, the image restoration module consists of a U-Net, a channel attention module and a double-attention module;
the noiseless and unambiguous image output by the image super-division module is used as a main input, and the high-resolution image is used as a branch input.
The backbone consists of 2 convolutional layers, one U-Net, 2 channel attention modules. The branch is composed of a convolution layer and a dual-attention module. After the noiseless and unambiguous image is input into a main path, a thermodynamic diagram is obtained through a convolution layer and U-Net, the thermodynamic diagram is summed with the output of the branch path after passing through 2 channel attention modules to obtain a fusion characteristic diagram, and finally, a repaired image is obtained through the convolution layer.
The training data of the model is obtained by artificially covering the complete two-dimensional code, the uncovered image is used as the training data, and the loss function is that
;
Wherein,is group-trunk, +.>Is the output image after passing the model.
In the description of embodiments of the present application, the terms "first," "second," "third," "fourth" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first", "a second", "a third" and a fourth "may explicitly or implicitly include one or more such feature. In the description of the present application, unless otherwise indicated, the meaning of "a plurality" is two or more.
In the description of embodiments of the present application, the term "and/or" is merely an association relationship describing an association object, meaning that three relationships may exist, e.g., a and/or B, may represent: a exists alone, A and B exist together, and B exists alone. In addition, the character "/" herein generally indicates that the front and rear associated objects are an "or" relationship.
Although embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the application, the scope of which is defined in the appended claims and their equivalents.
Claims (10)
1. The two-dimensional code identification method with the automatic repair function is characterized by comprising the following steps of:
judging whether the material on the conveyor belt reaches a preset position or not, and acquiring a two-dimensional code image to be identified of the material reaching the preset position;
performing super-resolution reconstruction on the two-dimensional code image to be identified to obtain a noise-free and blur-free image and a high-resolution image of the two-dimensional code image to be identified;
respectively inputting the noiseless and unambiguous image and the high-resolution image into a channel attention module and a double attention module to obtain a thermodynamic channel diagram and a branch output diagram of the two-dimensional code image to be identified;
and fusing the thermal channel diagram and the branch output diagram to obtain a fused feature diagram of the two-dimensional code image to be identified, and decoding the two-dimensional code in the fused feature diagram by ZXing to obtain target information.
2. The two-dimensional code identification method with an automatic repair function according to claim 1, comprising the steps of:
judging whether the material on the conveyor belt reaches a preset position, and acquiring a two-dimensional code image to be identified of the material reaching the preset position specifically comprises the following steps:
the mechanical arm places the materials on a conveyor belt, and the conveyor belt conveys the materials to a preset position for image shooting;
and judging whether the material reaches a preset position by adopting an infrared sensor, and when the material reaches the preset position, carrying out image shooting on the material by using a camera to obtain a two-dimensional code image to be identified.
3. The two-dimensional code identification method with an automatic repair function according to claim 1, comprising the steps of:
performing super-resolution reconstruction on the two-dimensional code image to be identified to obtain a noise-free and blur-free image and a high-resolution image of the two-dimensional code image to be identified, wherein the super-resolution reconstruction comprises the following steps of:
inputting the two-dimensional code image to be identified into a GAN model generator to obtain a noise-free and blur-free image of the two-dimensional code image to be identified;
and inputting the noiseless and fuzzy-free image into a high-frequency learning layer and an up-sampling layer to obtain a high-resolution image.
4. The two-dimensional code identification method with an automatic repair function according to claim 3, comprising:
the GAN model generator comprises 2 channel attention modules, 4 residual modules, and 1 self-attention module;
the high frequency learning layer comprises 3 dense network blocks;
the upsampling layer includes 2 convolutional layers, 1 spectral normalization layer, 1 pixel rebinning layer, and LeakyRelu.
5. The two-dimensional code identification method with an automatic repair function according to claim 1, comprising the steps of:
the step of inputting the noiseless and fuzzy-free image and the high-resolution image into a channel attention module and a double attention module respectively to obtain a thermodynamic channel diagram and a branch output diagram of the two-dimensional code image to be identified, which specifically comprises the following steps: taking the high-resolution image as a branch input convolution layer and a double-attention module to obtain a branch output image;
and inputting the noiseless and unambiguous image into a convolution layer and U-Net to obtain a thermodynamic diagram, and inputting the thermodynamic diagram into two channel attention modules to obtain a thermodynamic channel diagram.
6. The two-dimensional code identification method with an automatic repair function according to claim 5, comprising:
the fusion of the thermodynamic channel diagram and the branch output diagram to obtain a fusion feature diagram of the two-dimensional code image to be identified specifically comprises the following steps:
summing the thermodynamic channel diagram and the branch output diagram to obtain a fusion characteristic diagram;
and inputting the fusion feature map into a convolution layer to obtain a repaired image.
7. The two-dimensional code identification method with the automatic repair function according to claim 4, wherein the GAN model generator specifically comprises:
and sequentially inputting the two-dimensional code image to be identified into a first channel attention module, a first residual error module, a downsampling to a second residual error module, a downsampling to a self-attention module, an upsampling to a third residual error module, an upsampling to a fourth residual error module and a second channel attention module to obtain a noise-free and blur-free image of the two-dimensional code image to be identified.
8. The two-dimensional code identification method with the automatic repair function according to claim 4, wherein the high-frequency learning layer specifically comprises:
and inputting the noiseless and unambiguous images of the two-dimensional code image to be identified into a first dense network module, a second dense network module and a third dense network module in sequence to obtain a high-resolution intermediate image.
9. The two-dimensional code identification method with the automatic repair function according to claim 8, wherein the up-sampling layer specifically comprises:
and sequentially inputting the high-resolution intermediate image into a first convolution layer, a spectrum normalization layer, a pixel recombination layer, a LeakyRelu and a second convolution layer, and outputting to obtain the high-resolution image.
10. Two-dimensional code recognition device with automatic repair function, characterized by comprising:
the image acquisition module is used for judging whether the material on the conveyor belt reaches a preset position or not and acquiring a two-dimensional code image to be identified of the material reaching the preset position;
the image super-division module performs super-division reconstruction on the two-dimensional code image to be identified to obtain a noise-free and blur-free image and a high-resolution image of the two-dimensional code image to be identified;
the image restoration module is used for inputting the noiseless and fuzzy-free image and the high-resolution image into a channel attention module and a double attention module respectively to obtain a thermal channel image and a branch output image of the two-dimensional code image to be identified, and fusing the thermal channel image and the branch output image to obtain a fused feature image of the two-dimensional code image to be identified;
the two-dimensional code identification module decodes the two-dimensional code in the fusion feature map by adopting ZXing and obtains target information;
the two-dimensional code identification device with the automatic repair function is used for executing the two-dimensional code identification method with the automatic repair function according to any one of claims 2-9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311303893.2A CN117034982B (en) | 2023-10-10 | 2023-10-10 | Two-dimensional code identification method and device with automatic repair function |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311303893.2A CN117034982B (en) | 2023-10-10 | 2023-10-10 | Two-dimensional code identification method and device with automatic repair function |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117034982A true CN117034982A (en) | 2023-11-10 |
CN117034982B CN117034982B (en) | 2023-12-29 |
Family
ID=88624918
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311303893.2A Active CN117034982B (en) | 2023-10-10 | 2023-10-10 | Two-dimensional code identification method and device with automatic repair function |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117034982B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN118070826A (en) * | 2024-04-19 | 2024-05-24 | 宁波安得智联科技有限公司 | Two-dimensional code repairing method, device, terminal equipment and storage medium |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112884657A (en) * | 2021-05-06 | 2021-06-01 | 中南大学 | Face super-resolution reconstruction method and system |
US20210326656A1 (en) * | 2020-04-15 | 2021-10-21 | Adobe Inc. | Panoptic segmentation |
CN115471703A (en) * | 2022-09-21 | 2022-12-13 | 正泰集团研发中心(上海)有限公司 | Two-dimensional code detection method, model training method, device, equipment and storage medium |
CN115486068A (en) * | 2020-12-29 | 2022-12-16 | 腾讯美国有限责任公司 | Method and apparatus for inter-frame prediction based on deep neural network in video coding |
CN115829876A (en) * | 2022-12-15 | 2023-03-21 | 南京邮电大学 | Real degraded image blind restoration method based on cross attention mechanism |
CN115908753A (en) * | 2023-01-06 | 2023-04-04 | 粤港澳大湾区数字经济研究院(福田) | Whole body human mesh surface reconstruction method and related device |
CN115953494A (en) * | 2023-03-09 | 2023-04-11 | 南京航空航天大学 | Multi-task high-quality CT image reconstruction method based on low dose and super-resolution |
CN116703725A (en) * | 2023-06-01 | 2023-09-05 | 华东理工大学 | Method for realizing super resolution for real world text image by double branch network for sensing multiple characteristics |
CN116797541A (en) * | 2023-05-14 | 2023-09-22 | 哈尔滨理工大学 | Transformer-based lung CT image super-resolution reconstruction method |
-
2023
- 2023-10-10 CN CN202311303893.2A patent/CN117034982B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20210326656A1 (en) * | 2020-04-15 | 2021-10-21 | Adobe Inc. | Panoptic segmentation |
CN115486068A (en) * | 2020-12-29 | 2022-12-16 | 腾讯美国有限责任公司 | Method and apparatus for inter-frame prediction based on deep neural network in video coding |
CN112884657A (en) * | 2021-05-06 | 2021-06-01 | 中南大学 | Face super-resolution reconstruction method and system |
CN115471703A (en) * | 2022-09-21 | 2022-12-13 | 正泰集团研发中心(上海)有限公司 | Two-dimensional code detection method, model training method, device, equipment and storage medium |
CN115829876A (en) * | 2022-12-15 | 2023-03-21 | 南京邮电大学 | Real degraded image blind restoration method based on cross attention mechanism |
CN115908753A (en) * | 2023-01-06 | 2023-04-04 | 粤港澳大湾区数字经济研究院(福田) | Whole body human mesh surface reconstruction method and related device |
CN115953494A (en) * | 2023-03-09 | 2023-04-11 | 南京航空航天大学 | Multi-task high-quality CT image reconstruction method based on low dose and super-resolution |
CN116797541A (en) * | 2023-05-14 | 2023-09-22 | 哈尔滨理工大学 | Transformer-based lung CT image super-resolution reconstruction method |
CN116703725A (en) * | 2023-06-01 | 2023-09-05 | 华东理工大学 | Method for realizing super resolution for real world text image by double branch network for sensing multiple characteristics |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN118070826A (en) * | 2024-04-19 | 2024-05-24 | 宁波安得智联科技有限公司 | Two-dimensional code repairing method, device, terminal equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN117034982B (en) | 2023-12-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110163953B (en) | Three-dimensional face reconstruction method and device, storage medium and electronic device | |
CN110008956B (en) | Invoice key information positioning method, invoice key information positioning device, computer equipment and storage medium | |
JP4887374B2 (en) | A method for obtaining scattered parallax field in stereo vision | |
CN117034982B (en) | Two-dimensional code identification method and device with automatic repair function | |
US9626568B2 (en) | Use of spatially structured light for dynamic three dimensional reconstruction and reality augmentation | |
CN112233067A (en) | Hot rolled steel coil end face quality detection method and system | |
CN111972700B (en) | Cigarette appearance detection method and device, equipment, system and medium thereof | |
Dhaya | CCTV surveillance for unprecedented violence and traffic monitoring | |
US11651581B2 (en) | System and method for correspondence map determination | |
US11348349B2 (en) | Training data increment method, electronic apparatus and computer-readable medium | |
JP7170605B2 (en) | Defect inspection device, defect inspection method, and program | |
CN112396073A (en) | Model training method and device based on binocular images and data processing equipment | |
CN116168025B (en) | Oil curtain type fried peanut production system | |
CN113066088A (en) | Detection method, detection device and storage medium in industrial detection | |
CN113657528A (en) | Image feature point extraction method and device, computer terminal and storage medium | |
Han et al. | Convolutional neural network-based infrared image super resolution under low light environment | |
Attard et al. | Image mosaicing of tunnel wall images using high level features | |
CN111178405A (en) | Similar object identification method fusing multiple neural networks | |
US9305235B1 (en) | System and method for identifying and locating instances of a shape under large variations in linear degrees of freedom and/or stroke widths | |
CN113191469A (en) | Logistics management method, system, server and storage medium based on two-dimension code | |
JPH04130587A (en) | Three-dimensional picture evaluation device | |
CN116452556A (en) | Metal diaphragm surface defect detection method based on cross feature fusion | |
CN115063813B (en) | Training method and training device of alignment model aiming at character distortion | |
KR102221317B1 (en) | Inspection apparatus and method based on deep learning | |
EP3879495A1 (en) | Image generation for training a neural network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |