CN109740395A - The two dimensional code localization method and system that deep learning is combined with SHAPE DETECTION - Google Patents
The two dimensional code localization method and system that deep learning is combined with SHAPE DETECTION Download PDFInfo
- Publication number
- CN109740395A CN109740395A CN201811539959.7A CN201811539959A CN109740395A CN 109740395 A CN109740395 A CN 109740395A CN 201811539959 A CN201811539959 A CN 201811539959A CN 109740395 A CN109740395 A CN 109740395A
- Authority
- CN
- China
- Prior art keywords
- dimensional code
- candidate
- image
- code
- deep learning
- 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
Abstract
The two dimensional code localization method and system combined the invention discloses a kind of deep learning with SHAPE DETECTION;This method includes sample image acquisition module, sample image labeling module, deep learning module, test image acquisition module, two dimensional code candidate region extraction module, two dimensional code candidate region pinpoint module, two dimensional code chip extraction module;A large amount of Datamatrix code sample images are acquired under complex light environment, then sample mark manually is carried out to it, deep learning is carried out to the sample set again, obtain two dimensional code candidate region extractor, next can collecting test image, pass through previously obtained two dimensional code candidate region extractor, to extract two-dimension code area that may be present in tested image in 2 D code, SHAPE DETECTION is carried out to the region again and realizes accurate positioning, finally by decoding algorithm, completes the extraction of two dimensional code clip information;The present invention is not only stable, but also high-efficient, while accuracy rate is high.
Description
Technical field
The invention belongs to mechanical vision inspection technology fields, and in particular to what a kind of deep learning was combined with SHAPE DETECTION
Two dimensional code localization method
Background technique
In industrial circle, industrial products and components are identified using two dimensional code, are realized to product and components
The functions such as tracking, assembly management, life cycle maintenance are generated, the professional standard of automatic industrial is had become.Wherein Data
Matrix (DM) two dimensional code is because its outstanding data compression capability and powerful error correcting capability are by industry and the blueness of logistic industry
It looks at.It is packed for that two dimensional code is different from food, drug and other consumer products, the application environment of industrial two dimensional code usually compares
Badly, the identification of two dimensional code usually along with noise, overexposure, abrasion, pollution the problems such as.Therefore fixed for the DM code of complex background
Position algorithm is of great significance and the urgent market demand.
But existing Datamatrix two dimensional code positioning system exists that time-consuming, can not identify under complex environment and accurately
The low disadvantage of rate.
Summary of the invention
1, goal of the invention.
The present invention provides two that a kind of high and stable deep learning of time-consuming short, locating accuracy is combined with SHAPE DETECTION
Tie up code localization method and system.
2, the technical solution adopted in the present invention.
The invention discloses the two dimensional code localization method that a kind of deep learning is combined with SHAPE DETECTION, specific steps are as follows:
Step 1 acquires image in 2 D code to be measured under complex light environment, the two dimension which is obtained by deep learning
Code candidate region extractor and the improvement Hough transformation two dimensional code potential figure location algorithm of gradient direction grouping obtain two dimension
Code candidate contours;
Step 2, the two dimensional code railway line that Gray Projection is carried out to it determine, obtain accurate two dimensional code localization region, into
And quickly carry out the acquisition of two dimensional code clip information;
Step 3 obtains accurate two-dimensional barcode information finally by RS decoding algorithm.
Further, in step 1, two dimensional code candidate region extractor obtains great amount of samples image, passes through Faster-Rcnn
Convolutional network may be detected in sample image containing the region of Datamatrix, be rectangle ROI by tested image segmentation,
Picture under varying environment containing two dimensional code is manually marked, Faster-Rcnn network instruction is carried out to the picture after mark
Practice, obtains the two dimensional code rectangle ROI extractor an of robust.
Further, in step 1, two dimensional code candidate contours are obtained, and are obtained test image, are mentioned by two dimensional code rectangle ROI
It takes device to carry out operation to image to be detected, obtains two dimensional code rectangle ROI region, then mean value down-sampling is carried out to ROI rectangle and is obtained
Image pyramid, the top convolution for carrying out 3*3 Puli and tieing up special operator of pyramid, obtains gradient direction and gradient modulus value, to institute
Obtained gradient direction carries out histogram calculation, choose the gradient direction value at each peak in histogram within a preset range pair
The pixel answered carries out hough transformation and seeks candidate straight line, screens to candidate straight line, chooses the four of all composition quadrangles
Item candidate's straight line finds out 4 intersection points as two dimensional code candidate contours.
Further, in step 2, accurate two dimensional code localization region, that is, accurate positioning point of two dimensional code four is obtained, to institute
There is each side of two dimensional code candidate contours quadrangle to carry out width vertical gray-level projection appropriate, calculates to obtain grey scale change
Waveform, such as find the Wave crest and wave trough spacing of the Gray Projection waveform of adjacent two edges meet a certain range and transition times unanimously i.e.
Assert that the railway line that the adjacent both sides are Datamatrix two dimensional code positions side, then detects the Gray Projection wave on two other side
Shape, flatness a certain range i.e. it is believed that current candidate profile there are two dimensional codes, it can be deduced that the intersection point of four straight lines is
Four anchor points of two dimensional code, the Gray Projection Wave crest and wave trough number of railway line are two-dimentional code bit number.
Further, in step 3, obtain accurate two-dimensional barcode information two dimensional code decoder module, by four anchor points with
And two-dimentional code bit number the average gray in each grid to be calculated, to being averaged in all grids to two dimensional code grid division point
Gray scale, which carries out OTSU binaryzation, can be obtained the binary code of Datamatrix two dimensional code, inquire Datamatrix agreement to two dimension
Chip extracts in code, obtains final result by error correction calculations.
The invention discloses the two dimensional code positioning systems that a kind of deep learning is combined with SHAPE DETECTION, including two dimensional code to wait
Favored area extractor obtains module, two dimensional code candidate contours obtain module, four accurate positioning points of two dimensional code obtain module, two dimension
Code decoder module;
Two dimensional code candidate region extractor obtains module, obtains great amount of samples image, passes through Faster-Rcnn convolutional network
It may be detected containing the region of Datamatrix in sample image, be rectangle ROI by tested image segmentation, to different rings
Picture under border containing two dimensional code is manually marked, and is carried out Faster-Rcnn network training to the picture after mark, is obtained one
The two dimensional code rectangle ROI extractor of a robust;
Two dimensional code candidate contours obtain module, test image are obtained, by two dimensional code rectangle ROI extractor to figure to be detected
As carrying out operation, two dimensional code rectangle ROI region is obtained, then mean value down-sampling is carried out to ROI rectangle and obtains image pyramid, golden word
The top convolution for carrying out 3*3 Puli and tieing up special operator of tower, obtains gradient direction and gradient modulus value, to obtained gradient direction into
Column hisgram calculates, and choosing the gradient direction value at each peak in histogram, corresponding pixel carries out within a preset range
Candidate straight line is sought in hough transformation, is screened to candidate straight line, is chosen the candidate straight line of all four for constituting quadrangle and is found out
4 intersection points are as two dimensional code candidate contours;
Four accurate positioning points of two dimensional code obtain module, carry out to each side of all two dimensional code candidate contours quadrangles
Width vertical gray-level projection appropriate calculates to obtain grey scale change waveform, such as finds the Gray Projection waveform of adjacent two edges
Wave crest and wave trough spacing meet a certain range and transition times unanimously assert that the adjacent both sides are Datamatrix two dimensional code
Railway line positions side, then detects the Gray Projection waveform on two other side, and flatness is in a certain range i.e. it is believed that current candidate
There are two dimensional codes for profile, it can be deduced that the intersection point of four straight lines is four anchor points of two dimensional code, the Gray Projection wave of railway line
Spike paddy number is two-dimentional code bit number;
Two dimensional code decoder module, to two dimensional code grid division point, is calculated by four anchor points and two-dimentional code bit number
Average gray in each grid, carrying out OTSU binaryzation to the average gray in all grids can be obtained Datamatrix bis-
The binary code of code is tieed up, inquiry Datamatrix agreement extracts chip in two dimensional code, obtains finally by error correction calculations
As a result.
Further, the two dimensional code candidate region extractor obtains module, shooting and network using industrial camera
Picture library obtains great amount of samples image.
Further, the two dimensional code candidate contours obtain module, and test image is obtained by industrial camera, choose
The gradient direction value at each peak pixel corresponding in positive and negative 22.5 degree carries out hough transformation and seeks candidate in histogram
Straight line.
3, technical effect caused by the present invention.
(1) this hair obtains a large amount of sample image, is trained by Faster-Rcnn convolutional network to sample image,
And then two dimensional code candidate region extractor is obtained, this method makes next two dimensional code candidate contours and precise positioning feature area
The operation in domain cost saved the time;
(2) precise positioning feature zone algorithm of the invention even more accurately calculates four anchor points of two dimensional code, makes
Two dimensional code chip recognizer can be carried out quickly and accurately;
(3) RS decoding algorithm of the invention can quickly and accurately obtain two dimensional code decoding result.Entire scheme is not only steady
It is fixed and high-efficient, while accuracy rate is high, up to 99.95%.
Detailed description of the invention
Fig. 1 is the visioning procedure for the two dimensional code localization method that a kind of deep learning of the invention is combined with SHAPE DETECTION
Figure.
Wherein module includes: that two dimensional code candidate region extractor obtains the acquisition of 1, two dimensional code candidate contours 2, two dimensional code four
It is accurately positioned point and obtains 3, two dimensional code decoding 4.
Specific embodiment
Embodiment 1
A kind of two dimensional code localization method that deep learning is combined with SHAPE DETECTION, refers to and acquires under complex light environment
Image in 2 D code to be measured, two dimensional code candidate region extractor and the gradient direction grouping which is obtained by deep learning change
Two dimensional code candidate contours are obtained into the potential figure location algorithm of Hough transformation two dimensional code, the two of Gray Projection then are carried out to it
It ties up code railway line to determine, obtains accurate two dimensional code localization region, and then quickly carry out the acquisition of two dimensional code clip information, finally
Accurate two-dimensional barcode information is obtained by RS decoding algorithm.
Further, two dimensional code candidate region extractor is by carrying out Faster- to the samples pictures after mark
Rcnn network training obtains.
Further, it is the improvement Hough transformation being grouped using gradient direction that the rapidly two dimensional code candidate contours, which obtain,
The potential figure location algorithm of two dimensional code is realized.
Further, it is the two dimensional code railway line decision algorithm by Gray Projection that the two dimensional code four, which are accurately positioned point,
It realizes.
Further, the two dimensional code decoding is to differentiate to realize with RS decoding algorithm by threshold value.
Technical solution of the present invention is described in further detail with reference to the accompanying drawing:
Referring to Fig. 1, the present invention includes four module, is the acquisition of two dimensional code candidate region extractor, two dimensional code candidate respectively
Profile obtains, four accurate positioning points of two dimensional code obtain, two dimensional code decoding.
Firstly, obtaining great amount of samples image using the shooting and Network Picture Database of industrial camera, then pass through Faster-
Rcnn convolutional network may be detected in sample image containing the region of Datamatrix, be rectangle by tested image segmentation
ROI manually marks the picture under varying environment containing two dimensional code, carries out Faster-Rcnn net to the picture after mark
Network training obtains the two dimensional code rectangle ROI extractor an of robust.
Later, test image is obtained by industrial camera, by two dimensional code rectangle ROI extractor to image to be detected into
Row operation obtains two dimensional code rectangle ROI region, then carries out mean value down-sampling to ROI rectangle and obtain image pyramid, and pyramid is most
High level carries out the convolution that 3*3 Puli ties up special operator, obtains gradient direction and gradient modulus value, carries out to obtained gradient direction straight
Side's figure calculates, and the gradient direction value for choosing each peak in histogram is carried out in positive and negative 22.5 degree interior corresponding pixels
Candidate straight line is sought in hough transformation, is screened to candidate straight line, is chosen the candidate straight line of all four for constituting quadrangle and is found out
4 intersection points are as two dimensional code candidate contours.
Later, width vertical gray-level projection appropriate is carried out to each side of all two dimensional code candidate contours quadrangles,
It calculates to obtain grey scale change waveform, such as finds that the Wave crest and wave trough spacing of the Gray Projection waveform of adjacent two edges meets certain model
It encloses and transition times unanimously assert that the railway line that the adjacent both sides are Datamatrix two dimensional code positions side, then detect other two
The Gray Projection waveform on side, flatness a certain range i.e. it is believed that current candidate profile there are two dimensional codes, it can be deduced that four
The intersection point of straight line is four anchor points of two dimensional code, and the Gray Projection Wave crest and wave trough number of railway line is two-dimentional code bit number.
Finally, to two dimensional code grid division point, being calculated in each grid by four anchor points and two-dimentional code bit number
Average gray, to the average gray in all grids carry out OTSU binaryzation can be obtained the two of Datamatrix two dimensional code into
Code processed, inquiry Datamatrix agreement extract chip in two dimensional code, obtain final result by error correction calculations.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (8)
1. the two dimensional code localization method that a kind of deep learning is combined with SHAPE DETECTION, it is characterised in that:
Step 1 acquires image in 2 D code to be measured under complex light environment, which is waited by the two dimensional code that deep learning obtains
Favored area extractor and the potential figure location algorithm of improvement Hough transformation two dimensional code of gradient direction grouping are waited to obtain two dimensional code
Select profile;
Step 2, the two dimensional code railway line that Gray Projection is carried out to it determine, obtain accurate two dimensional code localization region, and then fast
Speed carries out the acquisition of two dimensional code clip information;
Step 3 obtains accurate two-dimensional barcode information finally by RS decoding algorithm.
2. the two dimensional code localization method that a kind of deep learning according to claim 1 is combined with SHAPE DETECTION, feature
Be: in step 1, two dimensional code candidate region extractor obtains great amount of samples image, by Faster-Rcnn convolutional network to sample
It may be detected containing the region of Datamatrix in this image, be rectangle ROI by tested image segmentation, under varying environment
Picture containing two dimensional code is manually marked, and is carried out Faster-Rcnn network training to the picture after mark, is obtained a Shandong
The two dimensional code rectangle ROI extractor of stick.
3. the two dimensional code localization method that a kind of deep learning according to claim 2 is combined with SHAPE DETECTION, feature
Be: in step 1, two dimensional code candidate contours are obtained, and test image are obtained, by two dimensional code rectangle ROI extractor to be detected
Image carries out operation, obtains two dimensional code rectangle ROI region, then carry out mean value down-sampling to ROI rectangle and obtain image pyramid, gold
The top convolution for carrying out 3*3 Puli and tieing up special operator of word tower, obtains gradient direction and gradient modulus value, to obtained gradient direction
Histogram calculation is carried out, choosing the gradient direction value at each peak in histogram, corresponding pixel carries out within a preset range
Candidate straight line is sought in hough transformation, is screened to candidate straight line, is chosen the candidate straight line of all four for constituting quadrangle and is found out
4 intersection points are as two dimensional code candidate contours.
4. the two dimensional code localization method that a kind of deep learning according to claim 1 is combined with SHAPE DETECTION, feature
Be: in step 2, accurate two dimensional code localization region, that is, accurate positioning point of two dimensional code four is obtained, candidate to all two dimensional codes
Each side of profile quadrangle carries out width vertical gray-level projection appropriate, calculates to obtain grey scale change waveform, such as finds
The Wave crest and wave trough spacing of the Gray Projection waveform of adjacent two edges meets a certain range and transition times unanimously assert that this is adjacent
Both sides are that the railway line of Datamatrix two dimensional code positions side, then detects the Gray Projection waveform on two other side, and flatness exists
A certain range is believed that current candidate profile, and there are two dimensional codes, it can be deduced that the intersection point of four straight lines is two dimensional code four fixed
Site, the Gray Projection Wave crest and wave trough number of railway line are two-dimentional code bit number.
5. the two dimensional code localization method that a kind of deep learning according to claim 1 is combined with SHAPE DETECTION, feature
It is: in step 3, obtains accurate two-dimensional barcode information two dimensional code decoder module, passes through four anchor points and two-dimentional code bit number
To calculate the average gray in each grid to two dimensional code grid division point, OTSU is carried out to the average gray in all grids
The binary code of Datamatrix two dimensional code can be obtained in binaryzation, and inquiry Datamatrix agreement carries out chip in two dimensional code
It extracts, obtains final result by error correction calculations.
6. the two dimensional code positioning system that a kind of deep learning is combined with SHAPE DETECTION, it is characterised in that: including two dimensional code candidate
Extracted region device obtains module, two dimensional code candidate contours obtain module, four accurate positioning points of two dimensional code obtain module, two dimensional code
Decoder module;
Two dimensional code candidate region extractor obtains module, great amount of samples image is obtained, by Faster-Rcnn convolutional network to sample
It may be detected containing the region of Datamatrix in this image, be rectangle ROI by tested image segmentation, under varying environment
Picture containing two dimensional code is manually marked, and is carried out Faster-Rcnn network training to the picture after mark, is obtained a Shandong
The two dimensional code rectangle ROI extractor of stick;
Two dimensional code candidate contours obtain module, obtain test image, by two dimensional code rectangle ROI extractor to image to be detected into
Row operation obtains two dimensional code rectangle ROI region, then carries out mean value down-sampling to ROI rectangle and obtain image pyramid, and pyramid is most
High level carries out the convolution that 3*3 Puli ties up special operator, obtains gradient direction and gradient modulus value, carries out to obtained gradient direction straight
Side's figure calculates, and choosing the gradient direction value at each peak in histogram, corresponding pixel carries out hough within a preset range
Candidate straight line is sought in transformation, is screened to candidate straight line, is chosen the candidate straight line of all four for constituting quadrangle and is found out 4 friendships
Point is used as two dimensional code candidate contours;
Four accurate positioning points of two dimensional code obtain module, carry out width to each side of all two dimensional code candidate contours quadrangles
Vertical gray-level projection appropriate calculates the wave that the Gray Projection waveform of adjacent two edges is such as found to obtain grey scale change waveform
Spike paddy spacing meets a certain range and transition times unanimously assert that the adjacent both sides are the railway of Datamatrix two dimensional code
Line positions side, then detects the Gray Projection waveform on two other side, and flatness is in a certain range i.e. it is believed that current candidate profile
There are two dimensional codes, it can be deduced that the intersection point of four straight lines is four anchor points of two dimensional code, the Gray Projection wave crest wave of railway line
Paddy number is two-dimentional code bit number;
Two dimensional code decoder module, to two dimensional code grid division point, is calculated each by four anchor points and two-dimentional code bit number
Average gray in grid, carrying out OTSU binaryzation to the average gray in all grids can be obtained Datamatrix two dimensional code
Binary code, inquiry Datamatrix agreement chip in two dimensional code is extracted, obtain final result by error correction calculations.
7. the two dimensional code positioning system that deep learning according to claim 6 is combined with SHAPE DETECTION, it is characterised in that:
The two dimensional code candidate region extractor obtains module, obtains great amount of samples using the shooting and Network Picture Database of industrial camera
Image.
8. the two dimensional code positioning system that deep learning according to claim 6 is combined with SHAPE DETECTION, it is characterised in that:
The two dimensional code candidate contours obtain module, and test image is obtained by industrial camera, choose each peak in histogram
Gradient direction value pixels corresponding in positive and negative 22.5 degree carry out hough transformation and seek candidate straight line.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811539959.7A CN109740395A (en) | 2018-12-17 | 2018-12-17 | The two dimensional code localization method and system that deep learning is combined with SHAPE DETECTION |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811539959.7A CN109740395A (en) | 2018-12-17 | 2018-12-17 | The two dimensional code localization method and system that deep learning is combined with SHAPE DETECTION |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109740395A true CN109740395A (en) | 2019-05-10 |
Family
ID=66360336
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811539959.7A Pending CN109740395A (en) | 2018-12-17 | 2018-12-17 | The two dimensional code localization method and system that deep learning is combined with SHAPE DETECTION |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109740395A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107220577A (en) * | 2017-05-12 | 2017-09-29 | 广州智慧城市发展研究院 | A kind of Quick Response Code localization method and system based on machine learning |
CN110472623A (en) * | 2019-06-29 | 2019-11-19 | 华为技术有限公司 | Image detecting method, equipment and system |
CN111027343A (en) * | 2019-12-12 | 2020-04-17 | 中科微至智能制造科技江苏有限公司 | Bar code area positioning method and device |
CN111046692A (en) * | 2019-11-26 | 2020-04-21 | 北京奇艺世纪科技有限公司 | Method and device for determining two-dimensional code area |
CN111950308A (en) * | 2020-07-15 | 2020-11-17 | 江苏理工学院 | Two-dimensional code positioning method for AGV |
CN112329514A (en) * | 2020-09-07 | 2021-02-05 | 江苏感创电子科技股份有限公司 | Book checking method and system based on fast R-CNN algorithm |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101398895A (en) * | 2007-09-26 | 2009-04-01 | 杨高波 | Image preprocess method based on data matrix two-dimension bar code identification |
US20120138695A1 (en) * | 2008-03-27 | 2012-06-07 | Denso Wave Incorporated | Two-dimensional code having rectangular region provided with specific patterns for specify cell positions and distinction from background |
CN108416412A (en) * | 2018-01-23 | 2018-08-17 | 浙江瀚镪自动化设备股份有限公司 | A kind of logistics compound key recognition methods based on multitask deep learning |
-
2018
- 2018-12-17 CN CN201811539959.7A patent/CN109740395A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101398895A (en) * | 2007-09-26 | 2009-04-01 | 杨高波 | Image preprocess method based on data matrix two-dimension bar code identification |
US20120138695A1 (en) * | 2008-03-27 | 2012-06-07 | Denso Wave Incorporated | Two-dimensional code having rectangular region provided with specific patterns for specify cell positions and distinction from background |
CN108416412A (en) * | 2018-01-23 | 2018-08-17 | 浙江瀚镪自动化设备股份有限公司 | A kind of logistics compound key recognition methods based on multitask deep learning |
Non-Patent Citations (3)
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》 * |
郭浩铭: "复杂背景下的DataMatrix二维码识别算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
黄翀等: "局部特征自适应的DM二维码结构提取方法", 《计算机科学》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107220577A (en) * | 2017-05-12 | 2017-09-29 | 广州智慧城市发展研究院 | A kind of Quick Response Code localization method and system based on machine learning |
CN110472623A (en) * | 2019-06-29 | 2019-11-19 | 华为技术有限公司 | Image detecting method, equipment and system |
CN110472623B (en) * | 2019-06-29 | 2022-08-09 | 华为技术有限公司 | Image detection method, device and system |
CN111046692A (en) * | 2019-11-26 | 2020-04-21 | 北京奇艺世纪科技有限公司 | Method and device for determining two-dimensional code area |
CN111046692B (en) * | 2019-11-26 | 2023-04-14 | 北京奇艺世纪科技有限公司 | Method and device for determining two-dimensional code area |
CN111027343A (en) * | 2019-12-12 | 2020-04-17 | 中科微至智能制造科技江苏有限公司 | Bar code area positioning method and device |
CN111950308A (en) * | 2020-07-15 | 2020-11-17 | 江苏理工学院 | Two-dimensional code positioning method for AGV |
CN112329514A (en) * | 2020-09-07 | 2021-02-05 | 江苏感创电子科技股份有限公司 | Book checking method and system based on fast R-CNN algorithm |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109740395A (en) | The two dimensional code localization method and system that deep learning is combined with SHAPE DETECTION | |
CN107301368B (en) | Recognition method of DataMatrix two-dimensional code | |
CN102176228B (en) | Machine vision method for identifying dial plate information of multi-pointer instrument | |
CN110415257B (en) | Gas-liquid two-phase flow overlapped bubble image segmentation method | |
CN105894069B (en) | A kind of generation method and recognition methods of the CRC two dimensional code of vision guided navigation | |
CN109376740A (en) | A kind of water gauge reading detection method based on video | |
CN109190742B (en) | Decoding method of coding feature points based on gray feature | |
CN103632137B (en) | A kind of human eye iris segmentation method | |
CN104794421A (en) | QR (quick response) code positioning and recognizing methods | |
EP3330887A1 (en) | Chinese-sensitive code feature pattern detection method and system | |
CN104268499B (en) | Method for detecting barcode tags of commodities with barcodes | |
CN106803087A (en) | A kind of car number automatic identification method and system | |
CN113780200A (en) | Computer vision-based pavement multi-disease area detection and positioning method | |
CN108961262A (en) | A kind of Bar code positioning method under complex scene | |
Zhang et al. | Detection and identification method of medical label barcode based on deep learning | |
CN110648323A (en) | Defect detection classification system and method thereof | |
CN110634131B (en) | Crack image identification and modeling method | |
CN113506276B (en) | Marker and method for measuring structural displacement | |
CN111724354A (en) | Image processing-based method for measuring spike length and small spike number of multiple wheat | |
CN103699876A (en) | Method and device for identifying vehicle number based on linear array CCD (Charge Coupled Device) images | |
CN113894799A (en) | Robot and marker identification method and device for assisting environment positioning | |
CN115661586B (en) | Model training and people flow statistical method, device and equipment | |
CN115169375B (en) | AR and gun ball linkage-based high-level material visualization method | |
Sirmacek et al. | Road detection from remotely sensed images using color features | |
Cui et al. | Real-time missing seedling counting in paddy fields based on lightweight network and tracking-by-detection algorithm |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190510 |
|
RJ01 | Rejection of invention patent application after publication |