CN110321750A - Two-dimensional code identification method and system in a kind of picture - Google Patents
Two-dimensional code identification method and system in a kind of picture Download PDFInfo
- Publication number
- CN110321750A CN110321750A CN201910327101.2A CN201910327101A CN110321750A CN 110321750 A CN110321750 A CN 110321750A CN 201910327101 A CN201910327101 A CN 201910327101A CN 110321750 A CN110321750 A CN 110321750A
- Authority
- CN
- China
- Prior art keywords
- picture
- dimensional code
- image
- dimension code
- training
- 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.)
- Pending
Links
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
- G06K7/1408—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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Electromagnetism (AREA)
- Toxicology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses the two-dimensional code identification methods and system in a kind of picture, comprising: acquires the picture with two dimensional code and storage；Picture based on acquisition constructs training image data collection and test image data collection；Training image data collection is marked, outlines the two dimensional code in every picture with rectangle frame, and save the apex coordinate of rectangle frame；Mask R-CNN deep neural network model is constructed, for identification two dimensional code region in picture, and it is cut from original image；Input the apex coordinate of the rectangle frame of training image data collection picture and every picture mark, training Mask R-CNN deep neural network model；Two dimensional code picture to be identified is obtained, two dimensional code picture to be identified is inputted into trained Mask R-CNN deep neural network model, obtains the two-dimension code image in two dimensional code picture to be identified；The two-dimension code image obtained is scanned, identifies the relevant textual information in two-dimension code image；This method and system improve the discrimination of two dimensional code in picture.
Description
Technical field
The present invention relates to field of image processings, and in particular, to two-dimensional code identification method and system in a kind of picture.
Background technique
License picture such as food and beverage sevice licensing, food business licence etc. contains two dimensional code, can be by scanning license figure
Two dimensional code in piece accurately obtains the relevant textual information in license picture, this just proposes the scanning recognition rate of two dimensional code
High requirement.Due to differences such as the size of every license picture, fog-level, shading values, two dimensional code directly is carried out to original image and is swept
It is low to retouch discrimination, is unable to satisfy practical application request.
Summary of the invention
The present invention provides the two-dimensional code identification methods and system in a kind of picture, solve two in picture in the prior art
The not high problem of code scanning recognition rate is tieed up, the two dimensional code discrimination in picture is improved.
The two dimensional code that the present invention can farthest verify in photograph and picture extracts and successful scan parsing, accurately obtains
Obtain the relevant textual information in license picture.The present invention provide it is a kind of first to original picture carry out two-dimension code image extraction, it is then right
The two-dimension code image of extraction is repeatedly scaled and is rotated, and finally the related text in transformed two-dimension code image is extracted in scanning
The method of information.This method crawls a large amount of license pictures using crawler technology from certain network platform first, and by image data
It saves in the database.It randomly selects plurality of pictures from database manually to be marked, mask method is to utilize mark tool
The two dimensional code in every picture is outlined with rectangle frame, four apex coordinates of rectangle frame are the position for reflecting two dimensional code in picture
Information, the apex coordinate of the rectangle frame of every icon note is with the preservation of json data format.Then it constructs one and is used for target identification
With the Mask R-CNN deep neural network of segmentation, with the training image data collection picture training of the mark model, with the mould
Type extracts the two dimensional code in picture to be identified.Repeatedly scaling and rotation finally are done to the two-dimension code image of extraction, and swept
Parsing transformed two-dimension code image every time is retouched, once scanning successfully can be obtained text information.
Specifically, the technical solution of the present invention is as follows:
One aspect of the present invention provides the two-dimensional code identification method in a kind of picture, which comprises
Step 1: acquiring the picture with two dimensional code and storage；Picture based on acquisition constructs training image data collection and survey
Attempt sheet data collection；Training image data collection is marked, outlines the two dimensional code in every picture with rectangle frame, and save rectangle frame
Apex coordinate；
Step 2: Mask R-CNN deep neural network model is constructed, for identification two dimensional code region in picture, and
It is cut from original image；The apex coordinate of the rectangle frame of training image data collection picture and every picture mark is inputted,
Training Mask R-CNN deep neural network model；
Step 3: obtaining two dimensional code picture to be identified, it is deep that two dimensional code picture to be identified is inputted trained Mask R-CNN
Neural network model is spent, the two-dimension code image in two dimensional code picture to be identified is obtained；
Step 4: the two-dimension code image obtained in scanning step 3 identifies the relevant textual information in two-dimension code image.
Further, it states in step 3 after obtaining the two-dimension code image in two dimensional code picture to be identified, further includes that step is more
Secondary scaling and rotation transformation handle two-dimension code image, and the two-dimension code image after then identifying each conversion process improves discrimination,
Then the two-dimension code image after conversion process is inputted into scanning element.
Further, conversion process two-dimension code image specifically includes: n times rotation processing original two-dimension code image, generates N number of
Former two-dimension code image and N number of postrotational two-dimension code image are carried out M amplification respectively, generate M by postrotational two-dimension code image
A picture of × (N+1), completes the conversion process of two-dimension code image, and wherein M and N is the positive integer more than or equal to 1.
Further, four apex coordinates of rectangle frame correspond to location information of the two dimensional code in picture.
Further, the apex coordinate of the rectangle frame of every picture mark is with the preservation of json data format.
Further, training image data collection is marked, specifically: it is outlined in every picture using mark tool with rectangle frame
Two dimensional code.
Further, the training process of Mask R-CNN deep neural network model includes:
Training data pretreatment；
Construct Mask R-CNN network structure；
RoIAlign uses bilinear interpolation, rather than is rounded quantization, completes Pixel-level alignment；
L is loss function, L=Lcls+Lbox+Lmask, wherein LclsFor error in classification, LboxFor detection error, LmaskFor segmentation
Error.
Further, picture is scanned the two-dimensional code specifically: scan two dimension using open source program bag zxing and Zbarlight
Code picture.
On the other hand, the present invention also provides the two dimensional code identifying system in a kind of license picture, the system comprises:
Acquisition and storage unit, for acquiring the picture with two dimensional code and storage；
Picture pretreatment unit constructs training image data collection and test image data collection for the picture based on acquisition；
Training image data collection is marked, outlines the two dimensional code in every picture with rectangle frame, and save the apex coordinate of rectangle frame；
Model construction and training unit input training picture number for constructing Mask R-CNN deep neural network model
According to the apex coordinate of collection picture and the rectangle frame of every picture mark, training Mask R-CNN deep neural network model；
Two dimensional code extraction unit trains two dimensional code picture input to be identified for obtaining two dimensional code picture to be identified
Mask R-CNN deep neural network model, obtain two dimensional code picture to be identified in two-dimension code image；
Scanning element identifies the phase in two-dimension code image for scanning the two-dimensional code the two-dimension code image of extraction unit acquisition
Close text information.
Wherein, the two dimensional code extraction unit is also used to after obtaining the two-dimension code image in two dimensional code picture to be identified,
Then two-dimension code image after conversion process is inputted scanning element by conversion process two-dimension code image；Conversion process two dimensional code figure
Piece specifically includes: n times rotation processing original two-dimension code image, N number of postrotational two-dimension code image is generated, by former two-dimension code image
M amplification is carried out respectively with N number of postrotational two-dimension code image, is generated M × (N+1) a picture, is completed the change of two-dimension code image
Change processing, wherein M and N is the positive integer more than or equal to 1.
One or more technical solution provided by the invention, has at least the following technical effects or advantages:
This method passes through deep neural network first and extracts two dimensional code region from former two dimensional code picture to be identified,
The noise region during two-dimensional code scanning can be greatly reduced.Secondly as two-dimensional code scanning process is to picture quality (as greatly
The information such as small, clarity, illumination, position) it is extremely sensitive, therefore repeatedly scaling and rotation process can be obviously improved two dimension for addition
The whole discrimination of code.
Detailed description of the invention
Attached drawing described herein is used to provide to further understand the embodiment of the present invention, constitutes of the invention one
Point, do not constitute the restriction to the embodiment of the present invention；
Fig. 1 is the structural schematic diagram of training pattern in the present invention；
Fig. 2 is a kind of flow diagram of the two-dimensional code identification method in the present invention in picture.
Specific embodiment
To better understand the objects, features and advantages of the present invention, with reference to the accompanying drawing and specific real
Applying mode, the present invention is further described in detail.It should be noted that in the case where not conflicting mutually, it is of the invention
Feature in embodiment and embodiment can be combined with each other.
In the following description, numerous specific details are set forth in order to facilitate a full understanding of the present invention, still, the present invention may be used also
Implemented with being different from the other modes being described herein in range using other, therefore, protection scope of the present invention is not by under
The limitation of specific embodiment disclosed in face.
The present invention realizes the principle of example segmentation, training pattern using deep neural network on picture, and performance model mentions
It takes two-dimension code image and saves, then two-dimension code image is repeatedly scaled and rotated, the two dimensional code after last scan conversion
Picture, relevant text information can successfully be extracted by once scanning.Referring to FIG. 2, entire two-dimension code image is extracted and scanning
Resolving is as follows:
Step 1: license picture collection and storage.It is swashed using crawler system in the network platform and takes certification photo picture data, with
80%, 20% ratio is divided into trained image data collection and test image data collection, and is stored in cloud server；People
Work marks training set, i.e., is outlined in every picture using mark tool (VIA, VGG Image Annotator) with rectangle frame
Two dimensional code, four apex coordinates of rectangle frame are the location information for reflecting two dimensional code in picture, the rectangle of every picture mark
The apex coordinate of frame is with the preservation of json data format；
Step 2: constructing and train Mask R-CNN deep neural network.Construct Mask R-CNN deep neural network mould
Type, inputs training image data collection picture and corresponding json label data trains the model, the structural schematic diagram of training pattern
As shown in Figure 1.
RoIAlign uses bilinear interpolation, rather than is rounded quantization, completes Pixel-level alignment；
L is loss function, L=Lcls+Lbox+Lmask, wherein LclsFor error in classification, LboxFor detection error, LmaskFor segmentation
Error；
Step 3: test Mask R-CNN model.Image data collection picture will be tested, exports result picture after input model
And save, the picture of preservation is checked, statistics two dimensional code recovery rate u is 99.4%, illustrates the deep neural network that the training generates
Model is effective.
Recovery rateWherein N is the number for testing picture, NsFor the picture number for successfully extracting two-dimension code image.
Step 4: transformation two-dimension code image.After model successfully extracts two-dimension code image, so that it may be scanned to it, to mention
Discrimination is risen, needs to convert it.For balance discrimination and sweep time, 3 two-dimension code images are first rotated, rotate angle
Be [90 °, 180 °, 270 °] respectively, then each rotation amplified 7 times respectively, amplification factor be [1.1,1.2,1.3,1.4,
1.5,1.6,1.7], transformed two-dimension code image will save every time, in addition original two dimensional code picture is total to (3+1) × 7=28
Picture is used for follow up scan.
Step 5: scanning parsing two-dimension code image.It is scanned using open source program bag zxing and Zbarlight 28 two above
Code picture is tieed up, once scans and successfully terminates scanning.Relevant text information is extracted after scanning successfully resolved.
Picture in this method can be the license picture picture with two dimensional code that is also possible to other, and the present invention is to processing
Object picture without limit, this method be suitable for the picture with two dimensional code.License picture can be by crawler from the network platform
It obtains and (such as crawls food business licence from take-away platform), or obtained from other channels.
This method or this system two dimensional code on successful scan license picture and can accurately obtain license two dimensional code figure
Relevant textual information in piece is most important to subsequent text information processing.
It is verified, 3214 food business licences are scanned using the following method respectively, count discrimination: 1) with wechat or
Alipay scans original image, discrimination 51%；2) original image, discrimination 26% are scanned with program bag zxing and ZbarLight；
3) two-dimension code image is first extracted using deep neural network Mask R-CNN, then scans (not converting), discrimination 43%；
4) two-dimension code image is first extracted using deep neural network Mask R-CNN, 6 rotations is first carried out to the two-dimension code image of extraction,
Then 7 amplifications are executed to postrotational picture every time, finally the two-dimension code image after scan conversion, discrimination are respectively
63%.Can be seen that by above-mentioned experiment with this method i.e. method 4) discrimination can be substantially improved.
Although preferred embodiments of the present invention have been described, it is created once a person skilled in the art knows basic
Property concept, then additional changes and modifications may be made to these embodiments.So it includes excellent that the following claims are intended to be interpreted as
It selects embodiment and falls into all change and modification of the scope of the invention.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
Mind and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to include these modifications and variations.
Claims (9)
1. the two-dimensional code identification method in a kind of picture, which is characterized in that the described method includes:
Step 1: acquiring the picture with two dimensional code and storage；Picture based on acquisition constructs training image data collection and test chart
Sheet data collection；Training image data collection is marked, outlines the two dimensional code in every picture with rectangle frame, and save the vertex of rectangle frame
Coordinate；
Step 2: building Mask R-CNN deep neural network model, two dimensional code region in picture for identification, and by its
It is cut from original image；Input the apex coordinate of the rectangle frame of training image data collection picture and every picture mark, training
Mask R-CNN deep neural network model；
Step 3: obtaining two dimensional code picture to be identified, two dimensional code picture to be identified is inputted into trained Mask R-CNN depth mind
Through network model, the two-dimension code image in two dimensional code picture to be identified is obtained；
Step 4: the two-dimension code image obtained in scanning step 3 identifies the relevant textual information in two-dimension code image.
2. the two-dimensional code identification method in a kind of picture according to claim 1, which is characterized in that in the step 3
It further include that step repeatedly scales and rotation transformation processing two dimensional code figure after obtaining the two-dimension code image in two dimensional code picture to be identified
Piece, the two-dimension code image after then identifying each conversion process.
3. the two-dimensional code identification method in a kind of picture according to claim 2, which is characterized in that conversion process two dimensional code
Two-dimension code image in picture to be identified specifically includes: n times rotation processing original two-dimension code image, generates N number of postrotational two dimension
Former two-dimension code image and N number of postrotational two-dimension code image are carried out M amplification respectively, generate M × (N+1) a figure by code picture
Piece completes the conversion process of two-dimension code image, and wherein M and N is the positive integer more than or equal to 1.
4. the two-dimensional code identification method in a kind of picture according to claim 1, which is characterized in that four tops of rectangle frame
Point coordinate pair answers location information of the two dimensional code in picture.
5. the two-dimensional code identification method in a kind of picture according to claim 1, which is characterized in that every picture mark
The apex coordinate of rectangle frame is with the preservation of json data format.
6. the two-dimensional code identification method in a kind of picture according to claim 1, which is characterized in that mark training picture number
According to collection, specifically: the two dimensional code in every picture is outlined with rectangle frame using mark tool.
7. the two-dimensional code identification method in a kind of picture according to claim 1, which is characterized in that Mask R-CNN depth
The training process of neural network model includes:
Training data pretreatment；
Construct Mask R-CNN network structure；
RoIAlign uses bilinear interpolation, rather than is rounded quantization, completes Pixel-level alignment；
L is loss function, L=Lcls+Lbox+Lmask, wherein LclsFor error in classification, LboxFor detection error, LmaskIt is missed for segmentation
Difference.
8. the two dimensional code identifying system in a kind of picture, which is characterized in that the system comprises:
Acquisition and storage unit, for acquiring the picture with two dimensional code and storage；
Picture pretreatment unit constructs training image data collection and test image data collection for the picture based on acquisition；Mark
Training image data collection, outlines the two dimensional code in every picture with rectangle frame, and save the apex coordinate of rectangle frame；
Model construction and training unit input training image data collection for constructing Mask R-CNN deep neural network model
The apex coordinate of the rectangle frame of picture and every picture mark, training Mask R-CNN deep neural network model；
Two dimensional code extraction unit, for two dimensional code picture to be identified to be inputted trained Mask R-CNN deep neural network mould
Type obtains the two-dimension code image in two dimensional code picture to be identified；
Scanning element identifies the correlation text in two-dimension code image for scanning the two-dimensional code the two-dimension code image of extraction unit acquisition
This information.
9. the two dimensional code identifying system in a kind of picture according to claim 8, which is characterized in that the two dimensional code extracts
Unit is also used to after obtaining the two-dimension code image in two dimensional code picture to be identified, conversion process two-dimension code image, then will be become
Two-dimension code image input scanning element of changing that treated；Conversion process two-dimension code image, specifically includes: n times rotation processing original two
Code picture is tieed up, N number of postrotational two-dimension code image is generated, former two-dimension code image and N number of postrotational two-dimension code image are distinguished
M amplification is carried out, M × (N+1) a picture is generated, completes the conversion process of two-dimension code image, wherein M and N is more than or equal to 1
Positive integer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910327101.2A CN110321750A (en) | 2019-04-23 | 2019-04-23 | Two-dimensional code identification method and system in a kind of picture |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910327101.2A CN110321750A (en) | 2019-04-23 | 2019-04-23 | Two-dimensional code identification method and system in a kind of picture |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110321750A true CN110321750A (en) | 2019-10-11 |
Family
ID=68112943
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910327101.2A Pending CN110321750A (en) | 2019-04-23 | 2019-04-23 | Two-dimensional code identification method and system in a kind of picture |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110321750A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111160501A (en) * | 2019-12-25 | 2020-05-15 | 北京亮亮视野科技有限公司 | Construction method and device of two-dimensional code training data set |
| CN111161227A (en) * | 2019-12-20 | 2020-05-15 | 成都数之联科技有限公司 | Target positioning method and system based on deep neural network |
| CN111612012A (en) * | 2020-05-25 | 2020-09-01 | 信雅达***工程股份有限公司 | Health code identification method and device |
| CN112036205A (en) * | 2020-08-03 | 2020-12-04 | 浙江大华技术股份有限公司 | Two-dimensional code identification method, electronic device and storage medium |
| CN113408682A (en) * | 2021-06-29 | 2021-09-17 | 中航航空服务保障(天津)有限公司 | ResNet network-based tool identification method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9892301B1 (en) * | 2015-03-05 | 2018-02-13 | Digimarc Corporation | Localization of machine-readable indicia in digital capture systems |
| CN109409159A (en) * | 2018-10-11 | 2019-03-01 | 上海亿保健康管理有限公司 | A kind of fuzzy two-dimensional code detection method and device |
-
2019
- 2019-04-23 CN CN201910327101.2A patent/CN110321750A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9892301B1 (en) * | 2015-03-05 | 2018-02-13 | Digimarc Corporation | Localization of machine-readable indicia in digital capture systems |
| CN109409159A (en) * | 2018-10-11 | 2019-03-01 | 上海亿保健康管理有限公司 | A kind of fuzzy two-dimensional code detection method and device |
Non-Patent Citations (1)
| Title |
|---|
| 蔡若君等: "基于深度学习的二维码定位与检测技术", 《现代计算机(专业版)》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111161227A (en) * | 2019-12-20 | 2020-05-15 | 成都数之联科技有限公司 | Target positioning method and system based on deep neural network |
| CN111160501A (en) * | 2019-12-25 | 2020-05-15 | 北京亮亮视野科技有限公司 | Construction method and device of two-dimensional code training data set |
| CN111612012A (en) * | 2020-05-25 | 2020-09-01 | 信雅达***工程股份有限公司 | Health code identification method and device |
| CN112036205A (en) * | 2020-08-03 | 2020-12-04 | 浙江大华技术股份有限公司 | Two-dimensional code identification method, electronic device and storage medium |
| CN113408682A (en) * | 2021-06-29 | 2021-09-17 | 中航航空服务保障(天津)有限公司 | ResNet network-based tool identification method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110321750A (en) | Two-dimensional code identification method and system in a kind of picture | |
| US11818303B2 (en) | Content-based object detection, 3D reconstruction, and data extraction from digital images | |
| CN109948510B (en) | Document image instance segmentation method and device | |
| US10713528B2 (en) | System for determining alignment of a user-marked document and method thereof | |
| CN102790841B (en) | Method of detecting and correcting digital images of books in the book spine area | |
| US8917935B2 (en) | Detecting text using stroke width based text detection | |
| US8780131B2 (en) | Systems and methods for text-based personalization of images | |
| US5892854A (en) | Automatic image registration using binary moments | |
| CN102750555B (en) | Expression identification device applied to instant messaging tool | |
| CN103975342A (en) | Systems and methods for mobile image capture and processing | |
| CN101460937A (en) | Model- based dewarping method and apparatus | |
| Brown et al. | Restoring 2D content from distorted documents | |
| Zhao et al. | Deep learning-based forgery attack on document images | |
| US20080084413A1 (en) | Method for generating multi-resolution three-dimensional model | |
| JP6188052B2 (en) | Information system and server | |
| CN112949455B (en) | Value-added tax invoice recognition system and method | |
| CN113592735A (en) | Text page image restoration method and system, electronic equipment and computer readable medium | |
| CN102930251A (en) | Two-dimensional collection data recording and discriminating device and method | |
| CN114119949A (en) | Method and system for generating enhanced text synthetic image | |
| CN102737240A (en) | Method of analyzing digital document images | |
| CN110390228A (en) | The recognition methods of traffic sign picture, device and storage medium neural network based | |
| KR102562170B1 (en) | Method for providing deep learning based paper book digitizing service | |
| CN113221897B (en) | Image correction method, image text recognition method, identity verification method and device | |
| CN110766001B (en) | Bank card number positioning and end-to-end identification method based on CNN and RNN | |
| Su et al. | Automatic seal imprint verification systems using edge difference |
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 | ||
| CB02 | Change of applicant information |
Address after: 610000 No. 270, floor 2, No. 8, Jinxiu street, Wuhou District, Chengdu, Sichuan Applicant after: Chengdu shuzhilian Technology Co.,Ltd. Address before: 610000 No.2, 4th floor, building 1, Jule Road intersection, West 1st section of 1st ring road, Wuhou District, Chengdu City, Sichuan Province Applicant before: CHENGDU SHUZHILIAN TECHNOLOGY Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191011 |
|
| RJ01 | Rejection of invention patent application after publication |