CN110490022A - A kind of bar code method and device in identification picture - Google Patents
A kind of bar code method and device in identification picture Download PDFInfo
- Publication number
- CN110490022A CN110490022A CN201910784481.2A CN201910784481A CN110490022A CN 110490022 A CN110490022 A CN 110490022A CN 201910784481 A CN201910784481 A CN 201910784481A CN 110490022 A CN110490022 A CN 110490022A
- Authority
- CN
- China
- Prior art keywords
- picture
- bar code
- convolutional neural
- neural networks
- frame information
- 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
- G06K7/1439—Methods for optical code recognition including a method step for retrieval of the optical code
- G06K7/1443—Methods for optical code recognition including a method step for retrieval of the optical code locating of the code in an image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
- G06K7/00—Methods or arrangements for sensing record carriers, e.g. for reading patterns
- G06K7/10—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation
- G06K7/14—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation using light without selection of wavelength, e.g. sensing reflected white light
- G06K7/1404—Methods for optical code recognition
- G06K7/146—Methods for optical code recognition the method including quality enhancement steps
- G06K7/1482—Methods for optical code recognition the method including quality enhancement steps using fuzzy logic or natural solvers, such as neural networks, genetic algorithms and simulated annealing
Abstract
The invention discloses the bar code method and device in a kind of identification picture, the embodiment of the present invention receives the first picture;During identifying bar code, the second picture of multiple different resolution sizes is generated based on first picture, bar code candidate frame is obtained in each second picture, and acquired multiple bar code candidate frames are merged, bar code candidate frame information is obtained;Based on bar code candidate frame information, determine that bar code selectes frame information, so that the bar code selected in the first picture that frame information acquires again based on the bar code is identified.In this way, make the first picture acquired again includes the characteristics of bar code point information of the application identification of available identified picture completely by the positioning to the bar code in bar code picture, the identification to bar code is realized, to successfully identify bar code.
Description
Technical field
Bar code method and device the present invention relates to field of computer technology, in particular in a kind of identification picture.
Background technique
It is a large amount of mutual with the rapid development and application of mobile Internet and bar codes technique, especially planar bar code technology
On-line off-line scene of networking uses various types of bar codes, especially using two dimensional code as information storage, transmitting and identification skill
Art.It is illustrated so that bar code is two dimensional code as an example below.The main application scenarios of planar bar code technology include such as business card,
The acquisition of the information such as figure, internet password or data, website jump, ad promotions, mobile phone electric business, it is anti-fake trace to the source, preferential promotion,
Member management, and/or terminal payment etc..Planar bar code technology starts from late 1980s, including acquisition, identification and generation skill
Art, before mobile Internet is greatly developed, acquisition is to be got by dedicated scanning device to scan, such as two dimensional code
Scanner is quickly scanned the two-dimensional code by short distance bidimensional image technology, and scanning speed is fast and makes subsequent identification accurate.With
The development of development of Mobile Internet technology, each application provider of terminal, especially electric business application and payment application both provide
The function that two dimensional code acquisition is carried out using the camera unit of terminal itself setting, substantially increases the utilization rate of two dimensional code, wealthy width
The usage scenario of two dimensional code, so that can such as occur with market, streets and lanes and vehicle for public transport all kinds of
Two dimensional code for acquire and subsequent identification, facilitate the product that market publicizes oneself, also facilitate user makes in mobile context
With each mobile application.
When identifying two dimensional code, after the camera unit acquisition of terminal has the picture of two dimensional code, directly to two in picture
Dimension code is identified, the meaning of two dimensional code expression is obtained.
But it is acquired using aforesaid way and identifies that two dimensional code has the disadvantage that the camera unit due to terminal itself setting
When acquiring two-dimension code image, it is understood that there may be the limitation of the distance objective condition such as farther out, obtained two-dimension code image is not
The identification to the two dimensional code in picture can be cannot achieve including available completely with identified two dimensional code characteristic point information by having.
Summary of the invention
In view of this, the embodiment of the present invention provides a kind of bar code method identified in picture, this method can successfully be known
Bar code in other picture.
The embodiment of the present invention also provides a kind of bar code device identified in picture, which can successfully identify in picture
Bar code.
The embodiments of the present invention are implemented as follows:
A kind of bar code method in identification picture, this method comprises:
Receive the first picture;
The second picture that multiple different resolution sizes are generated based on first picture, is obtained in each second picture
Acquired multiple bar code candidate frames are merged, obtain bar code candidate frame information by bar code candidate frame;
It based on bar code candidate frame information, determines that bar code selectes frame information, frame information weight is selected based on the bar code
The first picture of new acquisition, so that having in the first picture resurveyed being capable of identified bar code information.
The bar code candidate frame information that obtains is obtained using the convolutional neural networks of setting;
The determining bar code, which selectes frame information, to be determined using the convolutional neural networks of setting.
The convolutional neural networks are arranged in the application of identification picture, and the identification picture is applied in identification picture
Bar code when, call the convolutional neural networks to execute.
Before the method, further includes:
Identify application the first picture of identification of picture;
The application of identification picture determines whether that identification obtains bar code information from the first picture, if it is, terminating this
Method, if it is not, then the convolutional neural networks is called to execute.
The convolutional neural networks are three-layer coil product neural network;
The convolutional neural networks are that the first picture based on setting is trained.
The calling convolutional neural networks, which execute, includes:
The convolutional neural networks zoom in and out received first picture, scale according to the different resolution of setting
At the second picture of multiple different resolution sizes;
First layer convolutional neural networks in the convolutional neural networks obtain multiple second pictures and are handled, and generate
Multiple bar code candidate frames of second picture simultaneously merge, and multiple bar code candidate frame information after merging are exported to the convolution
Second layer convolutional neural networks in neural network;
Second layer convolutional neural networks in the convolutional neural networks are believed according to multiple bar code candidate frames after merging
Breath determines the accurate candidate frame information in bar code picture, exports to the third layer convolutional Neural in the convolutional neural networks
Network;
Third layer convolutional neural networks in the convolutional neural networks are from the accurate candidate frame information in bar code picture
In, determine that bar code selectes frame information.
First layer convolutional neural networks in the convolutional neural networks are handled are as follows: by multiple different resolution sizes
Picture generate have the first pixel value picture;
Second layer convolutional neural networks in the convolutional neural networks determine that the accurate candidate frame in bar code picture is believed
Breath further include:
The picture of multiple different resolution sizes is generated to the picture with the second pixel value, based on the second pixel value
Picture and merging after multiple bar code candidate frame information determine the accurate candidate frame information in bar code picture, described second
The resolution ratio of pixel value is greater than the first pixel value;
Third layer convolutional neural networks in the convolutional neural networks determine that bar code selectes frame information further include:
The picture of multiple different resolution sizes is generated to the picture with third pixel value, the figure based on third pixel value
Accurate candidate frame information in piece and bar code picture determines that bar code selectes frame information, the resolution ratio of the third pixel value
Greater than the second pixel value.
Bar code device in a kind of identification picture, comprising: picture acquiring unit and detection unit, wherein
Picture acquiring unit, for receiving the first picture;
Detection unit, for generating the second picture of multiple different resolution sizes based on first picture, each
Bar code candidate frame is obtained in second picture, acquired multiple bar code candidate frames are merged, and obtains bar code candidate frame letter
Breath;It based on bar code candidate frame information, determines that bar code selectes frame information, frame information is selected based on the bar code and is resurveyed
First picture, so that having in the first picture resurveyed being capable of identified bar code information.
A kind of bar code device in identification picture, comprising:
Memory;And it is coupled to the processor of the memory, the processor is configured to described based on being stored in
Instruction in memory executes the bar code method in identification picture as described in any one of the above embodiments.
A kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
The bar code method in described in any item identification pictures among the above is realized when execution.
As above as it can be seen that the embodiment of the present invention receives the first picture;During identifying bar code, it is based on first picture
The second picture for generating multiple different resolution sizes obtains bar code candidate frame in each second picture, will be acquired
Multiple bar code candidate frames merge, and obtain bar code candidate frame information;Based on bar code candidate frame information, determine that bar code is selected
Frame information, so that the bar code selected in the first picture that frame information acquires again based on the bar code is identified.In this way, logical
The positioning to the bar code in bar code picture is crossed, make the first picture acquired again includes available identified picture completely
Using the characteristics of bar code point information of identification, the identification to bar code is realized, to successfully identify bar code.
Detailed description of the invention
Fig. 1 is the method flow diagram for the identification two dimensional code that the prior art provides;
Fig. 2 is the bar code method flow diagram in identification picture provided in an embodiment of the present invention;
Fig. 3 is the implementation method flow chart of convolutional neural networks provided in an embodiment of the present invention;
Fig. 4 is the first layer convolutional neural networks treatment process signal in convolutional neural networks provided in an embodiment of the present invention
Figure;
Fig. 5 is the second layer convolutional neural networks treatment process signal in convolutional neural networks provided in an embodiment of the present invention
Figure;
Fig. 6 is the third layer convolutional neural networks treatment process signal in convolutional neural networks provided in an embodiment of the present invention
Figure;
Fig. 7 is the bar code method specific example flow chart in identification picture provided in an embodiment of the present invention;
Fig. 8 is the bar code device structural schematic diagram in identification picture provided in an embodiment of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention more comprehensible, right hereinafter, referring to the drawings and the embodiments,
The present invention is further described.
May cause using background technique identification bar codes technique can not be to the bar code in bar code picture collected
Information accurately identifies.In the background technology, acquire and identify that planar bar code technology is based primarily upon the identification picture application of setting, such as
The application is two-dimensional scanning PostgreSQL database (such as ZXing) algorithm packet of open source, loads the ZXing algorithm packet at the terminal, should
After ZXing algorithm packet calls the camera unit acquisition two-dimension code image of terminal, then identify two dimensional code therein.
Be illustrated below by two dimensional code of bar code, certainly, bar code can also for one-dimension code, three-dimension code or its
The bar code of his type, does not limit here.
Specifically, the method flow diagram of the identification two dimensional code provided as shown in FIG. 1, FIG. 1 is the prior art, specific to walk
Suddenly are as follows:
The camera unit of step 101, initialization terminal itself setting;
In this step, camera unit, specifically camera are initialized and opened, the view-finder of camera is set, is prompted
The two dimensional code that user will acquire is placed into view-finder, to acquire two dimensional code;
Step 102, camera unit are to video flow processing;
In this step, the video flowing obtained to camera is handled, and extracts each frame picture in video flowing;
Step 103, camera unit focus detection, determine whether the picture in acquired video flowing focuses, if it is,
Execute step 104;If it is not, then the position of adjustment camera unit, carries out after focusing again, return step 103;
Step 104, camera unit carry out binary conversion treatment to the picture of focusing, form black and white picture;
The ZXing algorithm packet that step 105, terminal are arranged identifies the two dimensional code in the black and white picture;
In this step, ZXing algorithm packet can call recognizer, progressively scan the black and white picture, identify the artwork master
Three characteristic points of the two dimensional code in piece, i.e. the three of upper left, lower-left and upper right darkened features point;
Step 106, terminal ZXing algorithm packet determine whether three characteristic points in the black and white picture identify success, such as
Fruit is to then follow the steps 107;If it is not, then return step 105, waits next black and white picture to be processed;
The analytical algorithm of the ZXing algorithm packet calling setting of step 107, terminal, parsing obtain two in the black and white picture
Tie up code information;
Step 108, terminal ZXing algorithm packet judge whether successfully resolved, if so, thening follow the steps 109;If
No, then return step 105, wait next black and white picture to be processed;
Step 109 exports two-dimensional barcode information in the black and white picture.
Identification described in Fig. 1 obtains the method for two-dimensional barcode information there are many disadvantages, and one of disadvantage is exactly using eventually
What the function that the camera unit for holding itself to be arranged carries out two dimensional code acquisition generated.The equipment of previous acquisition two dimensional code, it is such as two-dimentional
Code scanner, is the content for needing close to two dimensional code faster more accurately to acquire out two-dimensional code scanning rifle two dimensional code.But
It is when being acquired using the camera unit of terminal itself setting to two dimensional code, is to need to utilize camera unit in most of scenes
The farther away two dimensional code of shooting distance, such as it is posted in the two dimensional code of shop metope, the two dimensional code being posted in compartment advertisement, display
Two dimensional code etc. on the large screen of distal end, these scenes because two dimensional code apart from camera lens farther out, if directly by the two of acquisition
It ties up code picture to identify using ZXing algorithm packet, then can not identify to obtain three characteristic points in the two-dimension code image, also just can not
Success parses and obtains two-dimensional barcode information.Although the camera unit of terminal itself setting is provided with focusing function, basis is needed
Instruction carries out, that is, to position two dimensional code in the position of two-dimension code image.And terminal when acquiring two-dimension code image not
The function is provided, this is problem to be solved of the embodiment of the present invention.
Therefore, in background technique, the ZXing algorithm packet of terminal do not provide the two dimensional code in two-dimension code image detection and
Positioning function causes two dimensional code in the farther away situation of camera unit being arranged apart from terminal, and the ZXing algorithm packet of terminal is from adopting
Three characteristic points that can't detect two dimensional code in the two-dimension code image of collection, also just can not correctly identify two-dimensional barcode information, and
The embodiment of the present invention is able to solve this problem.
In order to solve the problems in background technique, method used in the embodiment of the present invention are as follows: receive the first picture;It is identifying
During bar code, the second picture of multiple different resolution sizes is generated based on first picture, in each second picture
Acquired multiple bar code candidate frames are merged, obtain bar code candidate frame information by middle acquisition bar code candidate frame;Based on item
Shape code candidate frame information determines that bar code selectes frame information, so as to select that frame information acquires again based on the bar code
Bar code in one picture is identified.
In this way, making the first picture acquired again include completely can by the positioning to the bar code in bar code picture
The characteristics of bar code point information of the application identification of identified picture, realizes the identification to bar code, to successfully identify item
Shape code.
Fig. 2 is the bar code method flow diagram in identification picture provided in an embodiment of the present invention, the specific steps are that:
Step 201 receives the first picture;
Step 202, the second picture that multiple different resolution sizes are generated based on first picture, in each second figure
Bar code candidate frame is obtained in piece, and acquired multiple bar code candidate frames are merged, bar code candidate frame information is obtained;
Step 203 is based on bar code candidate frame information, determines that bar code selectes frame information;
Step 204 resurveys the first picture based on the selected frame information of the bar code, so that the first figure resurveyed
Having in piece being capable of identified bar code information.
In embodiments of the present invention, the bar code candidate frame information that obtains is obtained using the convolutional neural networks of setting
's;The determining bar code, which selectes frame information, to be determined using the convolutional neural networks of setting.
In embodiments of the present invention, convolutional neural networks are arranged in the application of identification picture, for example are arranged in ZXing
In algorithm packet, ZXing algorithm packet calls the convolutional neural networks to carry out the bar code in the first picture when identifying the first picture
Detection and positioning.
The embodiment of the present invention has been used for reference the convolutional neural networks based on deep learning and can be realized in picture detection and picture
Target positioning function, such as Face datection and positioning function.During the entire process of Face datection and positioning, Face datection is calculated
Method is used to monitor in a picture with the presence or absence of face, Face detection algorithm be for include in detecting picture face it
Afterwards, face is positioned, the position of face is specifically outlined with rectangle, and coordinate of the rectangle frame in picture is exported.This hair
Bright embodiment utilize this thinking, the bar code in the first picture is detected and is positioned, specifically first detect picture in whether
Comprising bar code, the position of rectangle frame bar code is then used again, and this is selected into frame information as location information and is exported, so that
It centainly include bar code information in subsequent the first picture based on the locating information acquisition, so that the application of identification picture can be known
Do not go out.
In embodiments of the present invention, convolutional neural networks are specifically constructed using multitask convolutional neural networks (MTCNN),
It is the barcode detection and location algorithm frame of a multitask, uses three-layer coil product neural network cascade, realize entire
The detection of bar code image and position fixing process.
It is described in detail so that bar code is two dimensional code as an example below.
Fig. 3 is the implementation method flow chart of convolutional neural networks provided in an embodiment of the present invention, is specifically included:
Step 301, convolutional neural networks zoom in and out the two-dimension code image of acquisition, contract according to the different resolution of setting
Put into the picture of multiple different resolution sizes;
First layer convolutional neural networks in step 302, convolutional neural networks obtain the figure of multiple different resolution sizes
Piece is simultaneously handled, and is generated multiple two dimensional code candidate frames in two-dimension code image and is merged, multiple two dimensional codes after merging are waited
Frame information is selected to export to the second layer convolutional neural networks in convolutional neural networks;
Second layer convolutional neural networks in step 303, convolutional neural networks are candidate according to multiple two dimensional codes after merging
Frame information determines the accurate candidate frame information in two-dimension code image, exports to the third layer convolutional Neural in convolutional neural networks
Network;
Third layer convolutional neural networks in step 304, convolutional neural networks are from the accurate candidate frame in two-dimension code image
In information, determines that two dimensional code selectes frame information, exported.
Step 301 in Fig. 3 is in order to which understanding to depth in three-dimensional is added in two-dimensional space, such as in picture
In object it is far when possible scale be using centimetre as unit, the object in picture just needs to be surveyed with millimeter when close
Only, but for convolutional neural networks, what ruler is each two-dimension code image that not can determine that be using to amount
Degree is to measure, so just to obtained two-dimension code image, i.e. the first picture is handled, so that it is big to obtain multiple different resolutions
Small picture.
In the step 302 in Fig. 3, specifically according to the picture for being scaled to multiple different resolution sizes, generation has
Whether the picture of first pixel value, then detecting in the picture of the pixel size with the first pixel value includes two dimensional code,
If generating two dimensional code candidate frame information comprising if.Herein, the result obtained is relatively rough, will include many two dimensional codes and waits
Frame is selected, using the merging algorithm of setting, multiple two dimensional code candidate frames of generation can be merged, convolutional neural networks are formed
In second layer convolutional neural networks input data, as shown in figure 4, use for reference recognition of face network model, Fig. 4 be the present invention
The first layer convolutional neural networks treatment process schematic diagram in convolutional neural networks that embodiment provides.
In the step 303 in Fig. 3, the second layer convolutional neural networks specifically in convolutional neural networks will be set
The picture of different resolution size generates the picture with the second pixel value respectively, then according to multiple two dimensional codes after merging
Candidate frame information determines the accurate candidate frame information in two-dimension code image, exports to the third layer convolution in convolutional neural networks
The resolution ratio of neural network, second pixel value is greater than the first pixel value.As shown in figure 5, using for reference the network mould of recognition of face
Type, Fig. 5 are the second layer convolutional neural networks treatment process schematic diagram in convolutional neural networks provided in an embodiment of the present invention.
In the step 304 in Fig. 3, the third layer convolutional neural networks specifically in convolutional neural networks will be set
The picture of different resolution size generates the picture with third pixel size respectively, then according in determining two-dimension code image
Accurate candidate frame information, determines the selected frame information of two dimensional code, exports.As shown in fig. 6, the network model of recognition of face is used for reference,
The resolution ratio of the third pixel value is greater than the second pixel value.Fig. 6 is in convolutional neural networks provided in an embodiment of the present invention
Third layer convolutional neural networks treatment process schematic diagram.
Herein, the first pixel value can be set to 12*12 pixel, and the second pixel value can be set to 24*24 pixel, the
Three pixel values can be set to 48*48 pixel, be illustrated by taking specific pixel value as an example below.
As can be seen that convolutional neural networks are constructed using MTCNN, three-layer coil product nerve is included in convolutional neural networks
Network needs to be trained the convolutional neural networks using two-dimension code image before the application convolutional neural networks.Training
The step of process is with the application convolutional neural networks are identical, three layers in result adjustment convolutional neural networks obtained according to training
Convolutional neural networks, until finally obtained training result is accurate.Convolutional neural networks have been trained when using two-dimension code image
Cheng Hou, then carry out the monitoring of two-dimension code image using the trained convolutional neural networks of institute and position two dimensional code.
Function involved in above-mentioned convolutional neural networks and identification two dimensional code can be integrated in setting by the embodiment of the present invention
Identification picture application in, such as in Zxing algorithm packet, so that operation Zxing algorithm packet, so that it may realize the whole of two dimensional code
A identification process, specific as described in Figure 7, Fig. 7 is the method flow diagram of identification two dimensional code provided in an embodiment of the present invention, specific
Step are as follows:
The camera unit of the Zxing algorithm packet that step 701, operation are arranged, the setting of Zxing algorithm packet triggering terminal is opened,
Intercept each frame two-dimension code image in video flowing;
The identification two dimensional code function being arranged in step 702, Zxing algorithm packet identifies the two-dimensional barcode information in picture, if
It identifies two-dimensional barcode information, then return to two-dimensional barcode information and closes camera unit, it, should if not detecting two-dimensional barcode information
Two-dimension code image is sent to the convolutional neural networks in Zxing algorithm packet;
Convolutional neural networks in step 703, Zxing algorithm packet zoom in and out two-dimension code image, are scaled to multiple points
Resolution picture of different sizes;
The first layer convolutional neural networks in convolutional neural networks in step 704, Zxing algorithm packet are more to what is received
Picture is handled, and the picture of 12*12 pixel is obtained, and the picture based on 12*12 pixel generates in two-dimension code image
Multiple two dimensional code candidate frames are simultaneously merged according to the merging algorithm of setting, and multiple two dimensional code candidate frame information after merging are defeated
Out to the second layer convolutional neural networks in convolutional neural networks;
The second layer convolutional neural networks in convolutional neural networks in step 705, Zxing algorithm packet are to multiple resolution ratio
Picture of different sizes is handled, and the picture of 24*24 pixel is obtained, according to multiple two dimensional code candidate frame information after merging
And the picture of 24*24 pixel is obtained, it determines the accurate candidate frame information in two-dimension code image, exports in convolutional neural networks
Third layer convolutional neural networks;
The third layer convolutional neural networks in convolutional neural networks in step 706, Zxing algorithm packet are to multiple resolution ratio
Picture of different sizes is handled, and the picture of 48*48 pixel is obtained, according to the accurate candidate frame information in two-dimension code image
And the picture of 48*48 pixel is obtained, determine that two dimensional code selectes frame information, Zxing algorithm packet is given in output, if not obtaining two
It ties up code and selectes frame information, be then returned directly to step 701;
Step 707, Zxing algorithm packet are based on two dimensional code and select frame information, call the camera unit of terminal setting, are become
Coke operation, amplification two-dimensional code picture region image return again to step 701 execution.
Fig. 8 is the bar code device schematic diagram in identification picture provided in an embodiment of the present invention, comprising: picture acquiring unit
And detection unit, wherein
Picture acquiring unit, for receiving the first picture;
Detection unit, for generating the second picture of multiple different resolution sizes based on first picture, each
Bar code candidate frame is obtained in second picture, acquired multiple bar code candidate frames are merged, and obtains bar code candidate frame letter
Breath;It based on bar code candidate frame information, determines that bar code selectes frame information, frame information is selected based on the bar code and is resurveyed
First picture, so that having in the first picture resurveyed being capable of identified bar code information.
The embodiment of the present invention also provides a kind of bar code device identified in picture, comprising:
Memory;And it is coupled to the processor of the memory, the processor is configured to described based on being stored in
Instruction in memory executes the bar code method in above-mentioned identification picture.
The embodiment of the present invention also provides a kind of computer readable storage medium, is stored thereon with computer program, the program
The bar code method in above-mentioned identification picture is realized when being executed by processor.
The embodiment of the present invention can be seen that convolutional neural networks and terminal of the embodiment of the present invention based on setting acquired certainly
Bar code picture training obtains convolutional neural networks, and convolutional neural networks are arranged in the application of identification picture.Work as use
During the application of identification picture identifies bar code, the barcode detection in bar code picture is helped through and has determined
Bit function so that identification picture application can according to convolutional neural networks export as a result, to terminal setting camera unit into
Row adjustment, to get bar code information, improves bar code recognition accuracy rate.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Within mind and principle, any modification, equivalent substitution, improvement and etc. done be should be included within the scope of the present invention.
Claims (10)
1. a kind of bar code method in identification picture, which is characterized in that this method comprises:
Receive the first picture;
The second picture that multiple different resolution sizes are generated based on first picture, obtains bar shaped in each second picture
Code candidate frame, acquired multiple bar code candidate frames are merged, bar code candidate frame information is obtained;
It based on bar code candidate frame information, determines that bar code selectes frame information, frame information is selected based on the bar code and is adopted again
Collect the first picture, so that having in the first picture resurveyed being capable of identified bar code information.
2. the method as described in claim 1, which is characterized in that the bar code candidate frame information that obtains is the volume using setting
Product neural network obtains;
The determining bar code, which selectes frame information, to be determined using the convolutional neural networks of setting.
3. method according to claim 2, which is characterized in that the application of identification picture is arranged in the convolutional neural networks
In, applying when identifying the bar code in picture for the identification picture calls the convolutional neural networks to execute.
4. method as claimed in claim 3, which is characterized in that before the method, further includes:
Identify application the first picture of identification of picture;
The application of identification picture determines whether that identification obtains bar code information from the first picture, if it is, terminate this method,
If it is not, then the convolutional neural networks is called to execute.
5. method according to claim 2, which is characterized in that the convolutional neural networks are three-layer coil product neural network;
The convolutional neural networks are that the first picture based on setting is trained.
6. detection method as claimed in claim 5, which is characterized in that the calling convolutional neural networks, which execute, includes:
The convolutional neural networks zoom in and out received first picture, are scaled to according to the different resolution of setting more
Open the second picture of different resolution size;
First layer convolutional neural networks in the convolutional neural networks obtain multiple second pictures and are handled, and generate second
Multiple bar code candidate frames of picture simultaneously merge, and multiple bar code candidate frame information after merging are exported to the convolutional Neural
Second layer convolutional neural networks in network;
Second layer convolutional neural networks in the convolutional neural networks are according to multiple bar code candidate frame information after merging, really
Determine the accurate candidate frame information in bar code picture, exports to the third layer convolutional neural networks in the convolutional neural networks;
Third layer convolutional neural networks in the convolutional neural networks are from the accurate candidate frame information in bar code picture, really
Determine bar code and selectes frame information.
7. detection method as claimed in claim 6, which is characterized in that the first layer convolutional Neural in the convolutional neural networks
Network is handled are as follows: the picture of multiple different resolution sizes is generated to the picture with the first pixel value;
Second layer convolutional neural networks in the convolutional neural networks determine the accurate candidate frame information in bar code picture also
Include:
The picture of multiple different resolution sizes is generated to the picture with the second pixel value, based on the figure with the second pixel value
Multiple bar code candidate frame information after piece and merging determine the accurate candidate frame information in bar code picture, second pixel
The resolution ratio of value is greater than the first pixel value;
Third layer convolutional neural networks in the convolutional neural networks determine that bar code selectes frame information further include:
The pictures of multiple different resolution sizes is generated to the picture with third pixel value, picture based on third pixel value and
Accurate candidate frame information in bar code picture determines that bar code selectes frame information, and the resolution ratio of the third pixel value is greater than
Second pixel value.
8. the bar code device in a kind of identification picture characterized by comprising picture acquiring unit and detection unit, wherein
Picture acquiring unit, for receiving the first picture;
Detection unit, for generating the second picture of multiple different resolution sizes based on first picture, each second
Bar code candidate frame is obtained in picture, and acquired multiple bar code candidate frames are merged, bar code candidate frame information is obtained;Base
It in bar code candidate frame information, determines that bar code selectes frame information, frame information is selected based on the bar code and resurveys first
Picture, so that having in the first picture resurveyed being capable of identified bar code information.
9. the bar code device in a kind of identification picture characterized by comprising
Memory;And it is coupled to the processor of the memory, the processor is configured to based on the storage is stored in
Instruction in device is executed such as the bar code method in identification picture of any of claims 1-7.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
The bar code method in identification picture of any of claims 1-7 is realized when execution.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910784481.2A CN110490022A (en) | 2019-08-23 | 2019-08-23 | A kind of bar code method and device in identification picture |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910784481.2A CN110490022A (en) | 2019-08-23 | 2019-08-23 | A kind of bar code method and device in identification picture |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110490022A true CN110490022A (en) | 2019-11-22 |
Family
ID=68553299
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910784481.2A Pending CN110490022A (en) | 2019-08-23 | 2019-08-23 | A kind of bar code method and device in identification picture |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110490022A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111222355A (en) * | 2019-12-30 | 2020-06-02 | 新大陆数字技术股份有限公司 | Method and system for positioning bar code on PCB |
CN111709262A (en) * | 2020-05-19 | 2020-09-25 | 李峰 | Three-dimensional code information intelligent identification method and device |
CN115329795A (en) * | 2022-10-17 | 2022-11-11 | 北京搜狐新动力信息技术有限公司 | Method and device for identifying two-dimensional code |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150262031A1 (en) * | 2012-12-06 | 2015-09-17 | Tencent Technology (Shenzhen) Company Limited | Method And Apparatus For Identifying Picture |
CN108121931A (en) * | 2017-12-18 | 2018-06-05 | 广州市动景计算机科技有限公司 | two-dimensional code data processing method, device and mobile terminal |
CN109543486A (en) * | 2018-10-29 | 2019-03-29 | 华南理工大学 | Bar code localization method neural network based and system |
CN109858305A (en) * | 2019-01-17 | 2019-06-07 | 柳州康云互联科技有限公司 | A kind of two dimensional code positioning identification system and method based on deep learning |
-
2019
- 2019-08-23 CN CN201910784481.2A patent/CN110490022A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150262031A1 (en) * | 2012-12-06 | 2015-09-17 | Tencent Technology (Shenzhen) Company Limited | Method And Apparatus For Identifying Picture |
CN108121931A (en) * | 2017-12-18 | 2018-06-05 | 广州市动景计算机科技有限公司 | two-dimensional code data processing method, device and mobile terminal |
CN109543486A (en) * | 2018-10-29 | 2019-03-29 | 华南理工大学 | Bar code localization method neural network based and system |
CN109858305A (en) * | 2019-01-17 | 2019-06-07 | 柳州康云互联科技有限公司 | A kind of two dimensional code positioning identification system and method based on deep learning |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111222355A (en) * | 2019-12-30 | 2020-06-02 | 新大陆数字技术股份有限公司 | Method and system for positioning bar code on PCB |
CN111709262A (en) * | 2020-05-19 | 2020-09-25 | 李峰 | Three-dimensional code information intelligent identification method and device |
CN111709262B (en) * | 2020-05-19 | 2024-01-09 | 海南天鉴防伪科技有限公司 | Intelligent three-dimensional code information identification method and device |
CN115329795A (en) * | 2022-10-17 | 2022-11-11 | 北京搜狐新动力信息技术有限公司 | Method and device for identifying two-dimensional code |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111652223B (en) | Certificate identification method and device | |
US11393256B2 (en) | Method and device for liveness detection, and storage medium | |
CN110232369B (en) | Face recognition method and electronic equipment | |
JP5318122B2 (en) | Method and apparatus for reading information contained in bar code | |
CN108197618B (en) | Method and device for generating human face detection model | |
CN105956059A (en) | Emotion recognition-based information recommendation method and apparatus | |
CN110490022A (en) | A kind of bar code method and device in identification picture | |
CN110516517B (en) | Target identification method, device and equipment based on multi-frame image | |
CN110298281B (en) | Video structuring method and device, electronic equipment and storage medium | |
CN108509994B (en) | Method and device for clustering character images | |
CN104751093A (en) | Method and device for acquiring image identification code displayed by host equipment | |
CN109871826A (en) | Information displaying method, device, computer readable storage medium and computer equipment | |
CN114049512A (en) | Model distillation method, target detection method and device and electronic equipment | |
CN111738199B (en) | Image information verification method, device, computing device and medium | |
CN116311214B (en) | License plate recognition method and device | |
CN112183148A (en) | Batch bar code positioning method and identification system | |
WO2023142396A1 (en) | Test method and apparatus for relocation module, device, system, medium, computer program, and computer program product | |
CN115860026A (en) | Bar code detection method and device, bar code detection equipment and readable storage medium | |
WO2017219562A1 (en) | Method and apparatus for generating two-dimensional code | |
CN111091089B (en) | Face image processing method and device, electronic equipment and storage medium | |
CN109934045B (en) | Pedestrian detection method and device | |
CN111753766A (en) | Image processing method, device, equipment and medium | |
CN111898536A (en) | Certificate identification method and device | |
CN109583471A (en) | Method and device for smart phone identification | |
CN111862106A (en) | Image processing method based on light field semantics, computer device and storage medium |
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 | ||
CB02 | Change of applicant information |
Address after: Room 221, 2 / F, block C, 18 Kechuang 11th Street, Daxing District, Beijing, 100176 Applicant after: Jingdong Technology Holding Co.,Ltd. Address before: Room 221, 2 / F, block C, 18 Kechuang 11th Street, Beijing Economic and Technological Development Zone, 100176 Applicant before: JINGDONG DIGITAL TECHNOLOGY HOLDINGS Co.,Ltd. |