CN108288037A - A kind of tire coding identifying system - Google Patents
A kind of tire coding identifying system Download PDFInfo
- Publication number
- CN108288037A CN108288037A CN201810051589.6A CN201810051589A CN108288037A CN 108288037 A CN108288037 A CN 108288037A CN 201810051589 A CN201810051589 A CN 201810051589A CN 108288037 A CN108288037 A CN 108288037A
- Authority
- CN
- China
- Prior art keywords
- layer
- model
- tire
- coding
- identifying system
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/94—Hardware or software architectures specially adapted for image or video understanding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/63—Scene text, e.g. street names
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/06—Recognition of objects for industrial automation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention belongs to intelligent imagings to identify field, and in particular to a kind of tire coding identifying system, includes mobile terminal platform and identification model based on arm processor, and the training data of the identification model uses Semi-intelligent Modular mask method.The present invention tire coding identifying system used the mobile terminal platform based on arm processor, greatly reduce equipment volume, the cost of equipment substantially reduces, and operation cost substantially reduces, and can hold use, meet it is various under the conditions of flexible configuration.The tire coding identifying system of the present invention of the present invention utilizes neural metwork training, using identification model, and Semi-intelligent Modular mask method is used in terms of the training data of model, improves the efficiency of data mark.
Description
Technical field
The invention belongs to intelligent imagings to identify field, and in particular to a kind of tire coding identifying system.
Background technology
For automaker, the database for preserving the date of manufacture of tire to set up an automobile is particularly important.
Tire production date, tiremaker information (such as D.O.T. codes) on the side wall of tire, by embossment or it is outstanding in the form of go out
Present black rubber surface；Tire manufacturer needs to record this kind of data in the manufacturing process of tire；Tyre rebuilding manufacturer, turns over
Newly, the use of tracking tire is also required to the high efficiency extraction to this type of information；Much it is related to the scene of a large amount of tire coding identifications, mesh
It is preceding all without miniaturization, low cost, high robust solution.Large manufacturer is generally known using set of Industrial Optics at present
Other equipment, mostly with high costs, running cost is higher.Due to that can not be miniaturized, hand-held, to some customization tires be difficult
It is dimensionally uniformly processed, it is still necessary to manual entry.
Invention content
For problems of the prior art, the object of the present invention is to provide a kind of intelligent, wheels based on mobile terminal
Tire coding identifying system, to solve the problems, such as to mention in above-mentioned background technology.
To achieve the goals above, the present invention adopts the following technical scheme that：
A kind of tire coding identifying system includes mobile terminal platform and identification model based on arm processor, the knowledge
The training data of other model uses Semi-intelligent Modular mask method.
It advanced optimizes, the mobile terminal platform is by 3840*3840 high-pixel cameras, arm processor, Soc mainboards, 4G
Ram, 16GRom and other hardware support kit equipment composition.
It advanced optimizes, the image that the camera takes is read with PNG format by Soc processing modules, by identification mould
Type is handled.
It advanced optimizes, the identification model includes one or more layers process flow.
It advanced optimizes, the identification model includes three layers of process flow：
First layer, copy original image use bilinear interpolation algorithm scaled down to artwork size 1/8, i.e. 480*
480, ROI region is detected with trained OpenCV HaarCascade models, and return to ROI region camber line substantially coordinate；
The arc area coordinate that first layer is handled is remapped to artwork, the area is extracted in artwork by the second layer
The region is sheared in domain, and the image normalization being cut out is handled, is broadcasted using matrix, is filled, carries out second layer ROI region inspection
It surveys；By this wheel detection, frame to the single coding letter of tire, the frame of number；
Third layer, the coding frame found by the second layer are cut out each coding to be normalized into the square for 60*80
Battle array, is identified, recognition result is output to database by CNN convolutional networks.
It advanced optimizes, the identification model has used three layers of processing model, and first floor model and the second layer model are
OpenCV graders, third layer model are deep learning model.
It advanced optimizes, three layers of processing model is required to training data to train.
It advanced optimizes, the detailed process of the training data training is：
The data of first floor model are to be taken the edge of x, the vertex of y-axis and radian region maximum according to entire tab area
Value, and it is scaled to 1/8, it is trained；
The data of second layer model, which need trained first layer model being applied in artwork, extracts characteristic coordinates, shearing
Go out arc coding region, the individual coding quadrilateral frame coordinate recursive transformation of each of mark to arc coding region picture
On, uniform sizes are normalized to again, and using bilinear interpolation algorithm scaled down to artwork size 1/4, this data is used in combination
Collection training second layer model, can capture and detect each alphanumeric frame；
The training of third layer model is the single character picture sheared based on each, most to tri- channel selectings of RGB
Excellent channel, usually channel B data integrity degree highest, are normalized to the matrix of 60*80, carry out CNN convolution pond layer with
Full articulamentum is classified by softmax, to judge all characters of A-Z/0-9.
The beneficial effects of the present invention are：
(1) tire coding identifying system of the invention has used the mobile terminal platform based on arm processor, greatly reduces and sets
Standby volume, the cost of equipment substantially reduce, and operation cost substantially reduces, and can hold use, meet it is various under the conditions of it is flexible
Configuration.
(2) the tire coding identifying system of the present invention of the invention utilizes neural metwork training, using identification model, in mould
Semi-intelligent Modular mask method is used in terms of the training data of type, improves the efficiency of data mark.
Description of the drawings
Fig. 1 is the overall flow schematic diagram of the present invention；
Fig. 2 is the tire identification ROI schematic diagrames of the present invention.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
A kind of tire coding identifying system includes mobile terminal platform and identification model based on arm processor, the knowledge
The training data of other model uses Semi-intelligent Modular mask method.
The mobile terminal platform used in inventive embodiments, the configurations of similar smart mobile phone, by high-pixel camera
(3840*3840), arm processor Soc mainboards, 4G Ram, 16GRom and other hardware support kit equipment composition.Camera is subsidiary
LED flash is controlled by processor, in photographic subjects, can increase exposure automatically according to image quality；Cam lens make
With the camera lens compared with high parsing power, big pixel particle, single imaging photosensitive particle diameter are used using camera CMOS in embodiment
Be 3 microns or more, under low light environment, under dim environment can use faster aperture time, can fast imaging, stablize at
Picture, cooperated with LED flash lighting can realize the higher image quality of no smear, clarity.Gui interface can prompt to make when shooting
User is the outer profile frame of tire into preset picture region.
The image that camera takes is read with PNG format by Soc processing modules, is handled by identification model；Identification
Model includes one layer or more process flow；Three layers of process flow are used in embodiment：
First layer, copy original image use bilinear interpolation algorithm scaled down to artwork size 1/8, i.e. 480*
480, ROI region is detected with trained OpenCV HaarCascade models, and return to ROI region camber line substantially coordinate；
The arc area coordinate that first layer is handled is remapped to artwork, the area is extracted in artwork by the second layer
The region is sheared in domain, and the image normalization being cut out is handled, is broadcasted using matrix, is filled, carries out second layer ROI region inspection
It surveys；By this wheel detection, frame to the single coding letter of tire, the frame of number；
Third layer, the coding frame found by the second layer are cut out each coding to be normalized into the square for 60*80
Battle array；It is identified by CNN convolutional networks；Recognition result is output to database.
For each layer of processing model, training data is needed to train, present invention uses three layer models, including two layers
OpenCV sorter models and layer depth neural network model, that is, deep learning model；For the first two layers a large amount of negative samples of acquisition
Data, as negative training set；
The annotation process of data records dependent coordinate using xml document, each coding letter or number quadrangle
Frame, which individually outlines, to be come, and letter, the numerical value in it are marked to each quadrilateral frame；After the completion of about 3000 similar picture marks
The cleaning again of data is carried out, three layer model is generated and corresponds to required data；The data of first floor model be according to entire tab area,
The edge maximum value of x, the vertex of y-axis and radian region are taken, and is scaled to 1/8, is trained；For second layer mould
The data of type, which need trained first layer model being applied in artwork, extracts characteristic coordinates, is cut out arc coding region,
The individual coding quadrilateral frame coordinate recursive transformation of each of mark to arc coding administrative division map on piece, system is normalized to again
One size is used in combination this data set to train the second layer model using bilinear interpolation algorithm scaled down to artwork size 1/4,
It can capture and detect each alphanumeric frame；The training of third layer model is the single word sheared based on each
Symbol image is normalized to 60*80 to tri- channel selecting optimal channels of RGB, usually channel B data integrity degree highest
Matrix, carry out CNN convolution pond layer and full articulamentum by softmax classification, to judge all characters of A-Z/0-9.
As shown in Figure 1, after the completion of first run data mark, preliminary model is trained with the data of first batch, utilizes this
A rudimentary model, including the model of first layer to third layer generate tire coding, are shown in mark interface, mark personnel can basis
The result of model is finely adjusted, for example adjusts the coordinate of some ROI region frame, corrects the identification character etc. in regional frame.
In annotation process, prediction is generated in advance in new mark pictures in the trained identification model at all levels of use
Xml data, these data are shown in mark interface, and mark work is exactly the process of error correction, when the portion for finding model automatic marking
Divide to greatly reduce the complexity of operation, accelerates the speed of mark.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, although with reference to aforementioned reality
Applying example, invention is explained in detail, for those skilled in the art, still can be to aforementioned each implementation
Technical solution recorded in example is modified or equivalent replacement of some of the technical features, all essences in the present invention
With within principle, any modification, equivalent replacement, improvement and so on should all be included in the protection scope of the present invention god.
Claims (8)
1. a kind of tire coding identifying system, which is characterized in that include mobile terminal platform and identification mould based on arm processor
The training data of type, the identification model uses Semi-intelligent Modular mask method.
2. a kind of tire coding identifying system according to claim 1, which is characterized in that the mobile terminal platform by
3840*3840 high-pixel cameras, arm processor, Soc mainboards, 4G Ram, 16GRom and other hardware support kit equipment groups
At.
3. a kind of tire coding identifying system according to claim 2, which is characterized in that the figure that the camera takes
As being read by Soc processing modules with PNG format, handled by identification model.
4. a kind of tire coding identifying system according to claim 3, which is characterized in that the identification model includes one layer
Or Multilevel method flow.
5. a kind of tire coding identifying system according to claim 3, which is characterized in that the identification model includes three layers
Process flow：
First layer, copy original image are used using bilinear interpolation algorithm scaled down to artwork size 1/8, i.e. 480*480
Trained OpenCV HaarCascade models detect ROI region, and return to ROI region camber line substantially coordinate；
The arc area coordinate that first layer is handled is remapped to artwork, the region is extracted in artwork, is cut by the second layer
The region is cut, the image normalization being cut out is handled, is broadcasted using matrix, is filled, carries out second layer ROI region detection；It is logical
Cross this wheel detection, frame to the single coding letter of tire, the frame of number；
Third layer, the coding frame found by the second layer each coding be cut out Lai, be normalized into the matrix for 60*80,
It is identified by CNN convolutional networks, recognition result is output to database.
6. a kind of tire coding identifying system according to claim 5, which is characterized in that the identification model has used three
Layer processing model, first floor model and the second layer model are OpenCV graders, and third layer model is deep learning model.
7. a kind of tire coding identifying system according to claim 6, which is characterized in that three layers of processing model is both needed to
Training data is wanted to train.
8. a kind of tire coding identifying system according to claim 7, which is characterized in that the tool of the training data training
Body process is：
The data of first floor model be the edge maximum value of x, the vertex of y-axis and radian region are taken according to entire tab area, and
It is scaled to 1/8, is trained；
The data of second layer model, which need trained first layer model being applied in artwork, extracts characteristic coordinates, is cut out arc
Shape coding region, the individual coding quadrilateral frame coordinate recursive transformation of each of mark to arc coding administrative division map on piece, then
It is secondary to normalize to uniform sizes, using bilinear interpolation algorithm scaled down to artwork size 1/4, this data set is used in combination to train
Second layer model can capture and detect each alphanumeric frame；
The training of third layer model is the single character picture sheared based on each, optimal to tri- channel selectings of RGB logical
Road, usually channel B data integrity degree highest are normalized to the matrix of 60*80, are carried out CNN convolution pond layer and are connected entirely
It connects layer by softmax to classify, to judge all characters of A-Z/0-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810051589.6A CN108288037B (en) | 2018-01-19 | 2018-01-19 | Tire code identification system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810051589.6A CN108288037B (en) | 2018-01-19 | 2018-01-19 | Tire code identification system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108288037A true CN108288037A (en) | 2018-07-17 |
| CN108288037B CN108288037B (en) | 2021-08-06 |
Family
ID=62835628
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810051589.6A Active CN108288037B (en) | 2018-01-19 | 2018-01-19 | Tire code identification system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108288037B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110705560A (en) * | 2019-10-14 | 2020-01-17 | 上海眼控科技股份有限公司 | Tire text acquisition method and device and tire specification detection method |
| CN111507325A (en) * | 2020-03-16 | 2020-08-07 | 重庆大学 | Industrial visual OCR recognition system and method based on deep learning |
| WO2020229132A1 (en) * | 2019-05-14 | 2020-11-19 | Wheelright Limited | Tyre sidewall imaging method |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN2507071Y (en) * | 2001-11-06 | 2002-08-21 | 华南理工大学 | Automatic tyre number identifier |
| CN1396093A (en) * | 2002-08-13 | 2003-02-12 | 华南理工大学 | System for recognizing and managing tyre sizes and its recognizing method |
| KR100511004B1 (en) * | 1999-06-01 | 2005-08-31 | 한국타이어 주식회사 | A apparatus for detecting the letter of tire |
| CN102446266A (en) * | 2010-09-30 | 2012-05-09 | 北京中远通科技有限公司 | Device, system and method for automatically identifying industrial number |
| CN102867185A (en) * | 2012-10-31 | 2013-01-09 | 江苏大学 | Method and system for identifying automobile tire number |
| CN103559237A (en) * | 2013-10-25 | 2014-02-05 | 南京大学 | Semi-automatic image annotation sample generating method based on target tracking |
| CN103870803A (en) * | 2013-10-21 | 2014-06-18 | 北京邮电大学 | Vehicle license plate recognition method and system based on coarse positioning and fine positioning fusion |
| CN104298976A (en) * | 2014-10-16 | 2015-01-21 | 电子科技大学 | License plate detection method based on convolutional neural network |
| CN106446895A (en) * | 2016-10-28 | 2017-02-22 | 安徽四创电子股份有限公司 | License plate recognition method based on deep convolutional neural network |
-
2018
- 2018-01-19 CN CN201810051589.6A patent/CN108288037B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100511004B1 (en) * | 1999-06-01 | 2005-08-31 | 한국타이어 주식회사 | A apparatus for detecting the letter of tire |
| CN2507071Y (en) * | 2001-11-06 | 2002-08-21 | 华南理工大学 | Automatic tyre number identifier |
| CN1396093A (en) * | 2002-08-13 | 2003-02-12 | 华南理工大学 | System for recognizing and managing tyre sizes and its recognizing method |
| CN102446266A (en) * | 2010-09-30 | 2012-05-09 | 北京中远通科技有限公司 | Device, system and method for automatically identifying industrial number |
| CN102867185A (en) * | 2012-10-31 | 2013-01-09 | 江苏大学 | Method and system for identifying automobile tire number |
| CN103870803A (en) * | 2013-10-21 | 2014-06-18 | 北京邮电大学 | Vehicle license plate recognition method and system based on coarse positioning and fine positioning fusion |
| CN103559237A (en) * | 2013-10-25 | 2014-02-05 | 南京大学 | Semi-automatic image annotation sample generating method based on target tracking |
| CN104298976A (en) * | 2014-10-16 | 2015-01-21 | 电子科技大学 | License plate detection method based on convolutional neural network |
| CN106446895A (en) * | 2016-10-28 | 2017-02-22 | 安徽四创电子股份有限公司 | License plate recognition method based on deep convolutional neural network |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020229132A1 (en) * | 2019-05-14 | 2020-11-19 | Wheelright Limited | Tyre sidewall imaging method |
| US11669952B2 (en) | 2019-05-14 | 2023-06-06 | Wheelright Limited | Tyre sidewall imaging method |
| CN110705560A (en) * | 2019-10-14 | 2020-01-17 | 上海眼控科技股份有限公司 | Tire text acquisition method and device and tire specification detection method |
| CN111507325A (en) * | 2020-03-16 | 2020-08-07 | 重庆大学 | Industrial visual OCR recognition system and method based on deep learning |
| CN111507325B (en) * | 2020-03-16 | 2023-04-07 | 重庆大学 | Industrial visual OCR recognition system and method based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108288037B (en) | 2021-08-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109636787B (en) | High-precision real-time battery spot welding quality detection method based on deep learning | |
| CN112686812B (en) | Bank card inclination correction detection method and device, readable storage medium and terminal | |
| CN108875821A (en) | The training method and device of disaggregated model, mobile terminal, readable storage medium storing program for executing | |
| KR101182173B1 (en) | Method and system for recognizing vehicle plate | |
| CN108288037A (en) | A kind of tire coding identifying system | |
| WO2005022465A2 (en) | Detecting and correcting redeye in an image | |
| CN112069902A (en) | Transformer substation cabinet pressing plate identification method and system | |
| CN113128507B (en) | License plate recognition method and device, electronic equipment and storage medium | |
| CN106295645B (en) | A kind of license plate character recognition method and device | |
| CN113159014B (en) | Objective question reading method, device, equipment and storage medium based on handwritten question number | |
| CN106355180B (en) | A kind of license plate locating method combined based on color with edge feature | |
| CN109190742B (en) | Decoding method of coding feature points based on gray feature | |
| CN111723805B (en) | Method and related device for identifying foreground region of signal lamp | |
| CN115170550A (en) | Deep learning-based battery defect detection method and system | |
| CN111160328A (en) | Automatic traffic marking extraction method based on semantic segmentation technology | |
| CN112528917A (en) | Zebra crossing region identification method and device, electronic equipment and storage medium | |
| CN112149535A (en) | Lane line detection method and device combining SegNet and U-Net | |
| CN111008979A (en) | Robust night image semantic segmentation method | |
| CN110807743B (en) | Image defogging method based on convolutional neural network | |
| CN112750113A (en) | Glass bottle defect detection method and device based on deep learning and linear detection | |
| CN112651913A (en) | Invoice seal desalination method, system and device and computer storage medium | |
| CN117541820A (en) | Method for identifying silica gel color change of transformer substation main transformer equipment based on semantic segmentation | |
| CN114882486B (en) | License plate detection and recognition method based on deep learning under night vision scene | |
| CN115690578A (en) | Image fusion method and target identification method and device | |
| CN112950599B (en) | Large intestine cavity area and intestine content labeling method based on deep learning |
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 |