CN106846011A - Business license recognition methods and device - Google Patents
Business license recognition methods and device Download PDFInfo
- Publication number
- CN106846011A CN106846011A CN201611265344.0A CN201611265344A CN106846011A CN 106846011 A CN106846011 A CN 106846011A CN 201611265344 A CN201611265344 A CN 201611265344A CN 106846011 A CN106846011 A CN 106846011A
- Authority
- CN
- China
- Prior art keywords
- image
- text
- business license
- module
- line
- 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
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000001514 detection method Methods 0.000 claims description 50
- 238000001303 quality assessment method Methods 0.000 claims description 29
- 238000004458 analytical method Methods 0.000 claims description 25
- 230000009467 reduction Effects 0.000 claims description 19
- 230000009466 transformation Effects 0.000 claims description 15
- 238000005520 cutting process Methods 0.000 claims description 14
- 238000013441 quality evaluation Methods 0.000 claims description 14
- 230000007935 neutral effect Effects 0.000 claims description 10
- 238000012549 training Methods 0.000 claims description 7
- 238000003708 edge detection Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 239000000203 mixture Substances 0.000 claims description 5
- 230000006870 function Effects 0.000 description 11
- 210000002569 neuron Anatomy 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000015654 memory Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 230000000638 stimulation Effects 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 239000002131 composite material Substances 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000013016 learning Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000003607 modifier Substances 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/018—Certifying business or products
- G06Q30/0185—Product, service or business identity fraud
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5083—Techniques for rebalancing the load in a distributed system
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/94—Hardware or software architectures specially adapted for image or video understanding
- G06V10/95—Hardware or software architectures specially adapted for image or video understanding structured as a network, e.g. client-server architectures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/50—Indexing scheme relating to G06F9/50
- G06F2209/5013—Request control
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Development Economics (AREA)
- Biophysics (AREA)
- Finance (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Economics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Entrepreneurship & Innovation (AREA)
- Character Input (AREA)
Abstract
The present invention relates to a kind of business license recognition methods and device, the business license identification request that server receiving terminal is uploaded, business license identification request carries the original image obtained after being taken pictures to business license, use the mode of load balancing that business license identification request is assigned into server carries out Text region with to the original image for carrying, generate the recognition result of business license, recognition result is stored in server, and recognition result is back to terminal.Automatic identification to business license is realized by server, time and manpower is saved.
Description
Technical field
The present invention relates to technical field of character recognition, more particularly to a kind of business license recognition methods and device.
Background technology
Business license is that the department in charge of industry and commerce issues industrial and commercial enterprises, a certain production warp is engaged in the allowance of individual operator
The voucher of battalion's activity.Its item in registration includes：Title, address, director, amount of the fund, economic sector, business scope, operation
Mode, number of employees, operating period etc..Business license is generally used for the identification of enterprise identity, determine enterprise whether necessary being,
Whether the corresponding and for achieving and reviewing of bearer and the information described in business license proved.Handle open a bank account,
Telecommunication service, enterprise's bid, the tax registration, the business such as sign a contract, and can be commonly using to battalion in some commercial activitys
Industry license by the item in registration information on business license, it is necessary to carry out registration typing.Traditional registration input method is mostly logical
Cross and item in registration information important on business license is manually registered into typing, take a substantial amount of time and manpower.
The content of the invention
Based on this, it is necessary to for above-mentioned technical problem, there is provided the business license identification side of a kind of time-consuming and manpower
Method and device.
A kind of business license recognition methods, methods described includes：
The business license identification request that receiving terminal is uploaded, the business license identification request is carried takes pictures to business license
The original image for obtaining afterwards；
Use the mode of load balancing that business license identification request is assigned into server to carry described pair
The original image carries out Text region, generates the recognition result of business license；
The recognition result is stored in server, and the recognition result is back to terminal.
A kind of business license identifying device, described device includes：
Original image acquisition module, request, the business license identification are recognized for the business license that receiving terminal is uploaded
Request carries the original image obtained after being taken pictures to business license；
Original image identification module, for business license identification request to be assigned into clothes by the way of load balancing
Business device carries out Text region with the original image to described pair of carrying, generates the recognition result of business license；
Recognition result output module, for the recognition result to be stored in into server, and the recognition result is returned
To terminal.
Above-mentioned business license recognition methods and device, the business license identification request that server receiving terminal is uploaded, business
License identification request carries the original image obtained after being taken pictures to business license, is known business license by the way of load balancing
Do not invite to ask and be assigned to server to carry out the described pair of original image for carrying Text region, the identification of generation business license
As a result, recognition result is stored in server, and recognition result is back to terminal.Realized to business license by server
Automatic identification, save time and manpower.
Brief description of the drawings
Fig. 1 is the applied environment figure of business license recognition methods in one embodiment；
Fig. 2 is the cut-away view of server in one embodiment；
Fig. 3 is the flow chart of business license recognition methods in one embodiment；
Fig. 4 is the flow chart of original image recognition in one embodiment；
Fig. 5 is the flow chart of printed page analysis treatment in one embodiment；
Fig. 6 is background and the flow chart of frame removal in one embodiment；
Fig. 7 is the flow chart that one embodiment Chinese one's own profession image is obtained；
Fig. 8 is the structural representation of business license identifying device in one embodiment；
Fig. 9 is the structural representation of original pattern recognition device in one embodiment；
Figure 10 is the structural representation of printed page analysis processing module in one embodiment；
Figure 11 is the structural representation of background and frame removal module in one embodiment；
Figure 12 is the structural representation of one embodiment Chinese one's own profession image collection module.
Specific embodiment
To enable the above objects, features and advantages of the present invention more obvious understandable, below in conjunction with the accompanying drawings to the present invention
Specific embodiment be described in detail.Many details are elaborated in the following description in order to fully understand this
Invention.But the present invention can be implemented with being much different from other manner described here, those skilled in the art can be
Without prejudice to doing similar improvement in the case of intension of the present invention, therefore the present invention is not limited by following public specific implementation.
Unless otherwise defined, all of technologies and scientific terms used here by the article with belong to technical field of the invention
The implication that technical staff is generally understood that is identical.The term for being used in the description of the invention herein is intended merely to description tool
The purpose of the embodiment of body, it is not intended that in the limitation present invention.Each technical characteristic of above example can carry out arbitrary group
Close, to make description succinct, combination not all possible to each technical characteristic in above-described embodiment is all described, however,
As long as the combination of these technical characteristics does not exist contradiction, the scope of this specification record is all considered to be.
Business license recognition methods provided in an embodiment of the present invention can be applied in environment as shown in Figure 1.With reference to Fig. 1 institutes
Show, terminal 102 is connected by network with server 104.The business license identification request that 104 receiving terminal of server 102 is uploaded,
Business license identification request carries the original image obtained after being taken pictures to business license, is held business by the way of load balancing
Be assigned to server 104 according to identification request carries out Text region with to the original image for carrying, and generates the identification knot of business license
Really, recognition result is stored in server 104, and recognition result is back to terminal 102.
In one embodiment, as shown in Fig. 2 additionally providing a kind of server, the server includes passing through system bus
The processor of connection, non-volatile memory medium, built-in storage, network interface, be stored with operation in non-volatile memory medium
System and a kind of business license identifying device, the business license identifying device are used to perform a kind of business license recognition methods.Should
Processor is used to improve calculating and control ability, supports the operation of whole server.Built-in storage is used to be non-volatile memories
The operation of the business license identifying device in medium provides environment, and computer-readable instruction can be stored in the built-in storage, should
Computer-readable instruction is when executed by, and may be such that a kind of business license recognition methods of the computing device.Network interface
For carrying out network service with terminal, data are received or send, the request of the business license identification that for example receiving terminal sends, with
And send business license identification etc. to terminal.
In one embodiment, as shown in Figure 3, there is provided a kind of business license recognition methods, Fig. 1 is applied in this way
In server as a example by illustrate, specifically include：
Step 310, the business license identification request that receiving terminal is uploaded, business license identification request is carried to business license
The original image obtained after taking pictures.
Business license identification function is made into micro services and is placed on high in the clouds, then open API (Application are provided
Programming Interface, application programming interface), be linked into the API in the application of oneself by user, it is possible to
Use business license identification function.Can be used under any operating system, not limited by operating system.
When user uploads business license by terminal to server recognizes request, because being taken in business license identification request
With business license original image, as business license original image is uploaded to server.User obtains business using terminal
License original image, for example, user can be entered by mobile phone, camera or other terminal-pair business licenses with camera function
Row is taken pictures, it is also possible to which by scanner etc., the terminal-pair business license with scan function is scanned.To take pictures again or sweep
Retouch the original image for obtaining to upload onto the server, upload mode supports local upload and URL (URL, Uniform
Resource Locator) upload, certainly in other embodiments, it is possibility to have other upload modes.
Step 320, uses the mode of load balancing that business license identification request is assigned into server with to the original for carrying
Beginning image carries out Text region, generates the recognition result of business license.
The business license that server is uploaded according to terminal recognizes request, and load balancing is carried out to the request, by the request point
The corresponding server of dispensing is processed.
Load balancing is made up of a server set multiple servers in a symmetrical manner, and every server all has
Status of equal value, individually can externally provide service and without the auxiliary of other servers.By load balancing technology, will be outer
The request that portion sends is evenly distributed on a certain server in symmetrical structure, and receives the server of request independently
Respond the request of client.Equally loaded can mean allocation client request to server array, take this to provide quick obtaining important
Data, solve the problems, such as a large amount of concurrent access services.
The server of request is received, the original image to being carried in the request carries out Text region, generate business license
Recognition result.
Step 330, is stored in server, and recognition result is back into terminal by recognition result.
On the one hand recognition result is stored in server, in case calling.On the other hand recognition result is returned into terminal, terminal
Recognition result can be directly read carries out registration typing.
In the present embodiment, business license identification function is made into micro services and is placed on high in the clouds, then open API, user are provided
The API is linked into the application of oneself, it is possible to will directly be used by the recognition result of identification output.It is truly realized logical
Cross automatic identification of the server to business license, time-consuming and manpower.
In one embodiment, as shown in figure 4, business license identification request is assigned into clothes by the way of load balancing
Business device carries out Text region with to the original image for carrying, and generates the recognition result of business license, specifically includes：
Step 322, the original image to receiving carries out printed page analysis treatment.
The server of request is received, the original image to being carried in request carries out printed page analysis treatment.Specifically, service
Device is ajusted to original image first, then to ajusting after image carry out quality evaluation.Typically can be to the color of image, clear
Clear degree, colour cast and brightness etc. carry out quality evaluation, generate quality assessment result, and incongruent project is processed, to cause
Image is qualified.If image some projects cannot pass through the later stage, and treatment is qualified to reach, then quality assessment result is fed back into end
End.So that Non-Compliance of the user in quality assessment result, targetedly shooting image, efficiently to obtain qualified
Image.Avoid and only feed back to user images and do not meet information, user does not know where unqualified image is but, it is impossible to corresponding
Adjust, the image or unqualified that may be shot next time, so as to waste substantial amounts of time and efforts.
Step 324, to being processed through printed page analysis after image carry out background and frame removal.
Server carries out background and frame removal to the image after ajusting.Specifically, the image after first to ajusting enters
Row noise reduction removes texture, and texture floating filter small in image is fallen.Rim detection is carried out to image again, is detected and clapped in image
Take the photograph the edge of business license.Straight-line detection is carried out according to the edge for detecting, the four edges circle straight line of business license is drawn.Most
Afterwards, four intersection points of boundary straight line are determined according to four edges circle straight line, the inframe constituted to four edges circle straight line and four intersection points
Image, carry out perspective transform, obtain removing the image of background and frame.
Step 326, to carrying out line of text detection and row cutting through the image after background and frame removal, obtains line of text figure
Picture.
Server removes color information to through the image after background and frame removal, first carrying out greyscale transformation.Again to image
Text edges detection is carried out, protrusion obtains text filed on business license.The edge detection graph text filed to each is entered again
Row is expanded by row, is specially expanded according to pixel column, obtains the Connection operator block connected with pixel column.Next, utilizing
Character features carry out connected domain detection to Connection operator block, reject non-legible Connection operator block, filter out text filed company
Logical domain.Finally, determine the boundary rectangle of text filed connected domain, according to boundary rectangle positional information, obtain a line a line
Line of text image.
Step 328, carries out Text region to line of text image, and generate by row according to the good neutral net of training in advance
Recognition result.
Using the good neutral net of precondition to carrying out Classification and Identification in the line of text image that obtains, it is possible to obtain text
The result of word identification.The result to Text region carries out preliminary Text extraction again, it is possible to recognizes business license and ties
Fruit returns to user by row.
Neutral net can be chosen for convolutional neural networks, and convolutional neural networks (CNN) are a kind of common deep learnings
Framework, is inspired by biological natural vision Cognition Mechanism, mainly by input layer, convolutional layer, and pond layer (also referred to as down-sampling layer),
Full articulamentum, output layer etc. are constituted, and wherein convolutional layer and pond layer can be multiple.Input layer is to treat training data, i.e., original
View data；Convolutional layer, with convolution kernel migration on image array, and correspondence position element multiplication, then multiplied result
It is added, last addition result forms new image array, and this layer has mainly carried out the extraction of feature to image, makes primary signal
Enhancing, reduces noise；Pond layer, maximum or weighted average in selection pond/sampling window, or average value etc. form new
Image array, fuzzy filter can be regarded as, play Further Feature Extraction, reduce parameter, reduce what is easily occurred in training
Over-fitting is acted on；Full articulamentum, an one-dimensional vector is pulled into by input matrix, then carries out dot product to this one-dimensional vector, mainly
The composite signal of some high values is selected for the re-sampling from convolution results, some is abandoned and is repeated low-quality composite signal,
To the feature learning of whole image, classification prediction.Output layer, for the output of result, obtains the classification results of image.
Neutral net can also be chosen for BP neural network, that is, choose based on BP (Back Propagation, reverse biography
Broadcast) the multilayer perceptron model of algorithm is trained to target.Multilayer perceptron model is by an input layer, one or more are hidden
Hide the neutral net composition of layer and an output layer.Each layer is interlinked by one or more neurons, neuron
Output can be just the input of another neuron.
Stimulation is passed to hidden layer by input layer, and hidden layer is by the intensity (weight) contacted between neuron and transmission rule
Then stimulation is passed to output layer by (activation primitive), and the stimulation that output layer arranges after hidden layer treatment produces final result.If having just
True result, then correct result and the result for producing are compared, error, then backstepping is obtained to the chain in neutral net
Connecing weight carries out feedback modifiers, so as to come complete study process.
For the activation primitive of neuron output in itself, Sigmoid functions are in general chosen.Sigmoid functions are
One function of common S types in biology, also referred to as S-shaped growth curve.Sigmoid functions are a good threshold value letters
Number, with continuous, smooth, strictly monotone characteristic.
In order to train grader, real character is trained during a part of business license is chosen in advance.For example, 10
Arabic numerals, 34 province information, type information, composition form information etc. are trained.
In the present embodiment, printed page analysis treatment is carried out to original image, quality appraisal report is generated, and feed back to use
Family.So that Non-Compliance of the user in quality assessment result, targetedly shooting image, efficiently to obtain qualified
Image.Avoid and only feed back to user images and do not meet information, user does not know where unqualified image is but, it is impossible to corresponding
Adjust, the image or unqualified that may be shot next time, so as to waste substantial amounts of time and efforts.
Text region is carried out to line of text image according to the good neutral net of training in advance, and by row generation recognition result.
Choose in a part of business license real character in advance to be trained neural network so that the identification of business license is accurate
Rate is substantially improved.And by row generation recognition result so that recognition result standardization, sharpening, readability enhancing.
In one embodiment, as shown in figure 5, the original image to receiving carries out printed page analysis treatment, specifically include：
Step 322a, obtains the anglec of rotation, according to rotation according to the words direction information on original image and graphical information
Angle is ajusted to original image.
Because the original image that user uploads not all is ajusted, therefore firstly the need of according to the word side on business license
Original image is ajusted to national emblem badge information, seal information etc. in information and figure, you can 90 degree, 180 degree are done to original image
Rotation ajust, or other are arbitrarily angled also according to rotation is needed.So that the text information on original image be all vertically to
On, facilitate follow-up Text region to work.
Step 322b, at least one in color, definition, colour cast and brightness to the image after ajusting carries out quality
Assessment, generates quality assessment result.
The original image quality that user uploads is uneven, it is therefore desirable to which the quality to the image after ajusting is commented
Estimate, generate quality assessment result.Quality assessment result is qualified, you can do the image procossing of next step, if unqualified, by quality
Assessment result feeds back to terminal.
Specifically, quality evaluation includes the detection to black and white, coloured image.Including being black to the original image that user uploads
Still cromogram is detected white figure, for example, the original image that some users upload is the copy of the business license, detects it is black
Bai Tu；If what is uploaded is the original business license image photograph for shooting, detect it is cromogram.Also include to black and white in cromogram
Partly, the detection of chrominance section.There is auxiliary to the determination of the detection to the external frame of line of text later of black and white, coloured image
Effect.Color table, hsv color space, the distribution of color of LAB color spaces of artwork master and the corresponding passages of RGB 3 of cromogram
It is different.According to above-mentioned distinctive information, the text message of the information such as badge, red chapter that image is enameled and black and white is identified.
Quality evaluation also includes the detection to image definition.Average gradient is used in the present embodiment
(meangradient) method is detected to the definition of image.Because average gradient can sensitively reflect image to minor detail
The ability of contrast expression, so can be used to the definition of evaluation image.Average gradient is bigger, represents that image level is more, clearly
Degree is also higher.Definition image high, nearby gray scale has notable difference, i.e. rate of gray level for the border or hachure both sides of its image
Greatly.Its corresponding gradient formula is：
Wherein I represents image, and I (i, j) represents the i-th row in image, and the gray value of jth row pixel, W, H represents figure respectively
The width and height of picture.
Quality evaluation also includes the assessment to colour cast.Shoot image process in, occasionally there are certain color form and aspect,
Saturation degree has obvious difference with real image, that is, there is color offset phenomenon.Be converted to for RGB image first by the detection of colour cast
Lab space, wherein L represent brightness of image, and a represents image red green component, and b represents image yellow blue component.Generally there is colour cast
Image, average on a and b components can deviation from origin it is far, variance also can be less than normal；By calculating image on a and b components
Average and variance, so that it may assess image with the presence or absence of colour cast.Image for there is colour cast needs to carry out white balance to picture
Treatment.The correlation formula for assessing colour cast is as follows：
K=D/M
Wherein, MaRepresent average of the image on a components, MbRepresent average of the image on b components；DaRepresent image in a
Variance on component, DbRepresent variance of the image on b components；M represents colour cast average of the image in Lab space；D represents image
In the colour cast variance of Lab space；Colour cast judges that factor K value is bigger, and colour cast degree is bigger.
Quality evaluation also includes being estimated the brightness of image.Specifically, calculate average of the picture on gray-scale map and
Variance, when there is brightness and be abnormal, average can deviate average point, it can be assumed that be average point 128, and variance also can be less than normal.Pass through
Calculate the average and variance of gray-scale map, so that it may assess image with the presence or absence of overexposure or under-exposure.
After color, definition, colour cast and brightness to the image after ajusting etc. all carry out quality evaluation, generation quality is commented
Estimate result.Certainly in other embodiments, it is also possible to which other characteristics of the image after to ajusting carry out quality evaluation.Quality evaluation
Result specifically include definition it is whether qualified, with the presence or absence of colour cast, with the presence or absence of overexposure or under-exposure etc..If judging quality
Assessment result is unqualified, then image is adjusted to qualified according to quality assessment result, if still unqualified, needs to clap again
According to upload.
Step 322c, terminal is back to by quality assessment result.
Quality assessment result is back to terminal by server, if quality assessment result is unqualified, user need to be according to matter
The Non-Compliance in assessment result is measured, targetedly shooting image uploads onto the server again, efficiently to obtain qualified
Image.
In the present embodiment, original image is ajusted so that the text information on original image is all straight up
, facilitate follow-up Text region to work.Further, to the image after ajusting color, definition, colour cast and brightness etc.
Quality evaluation is carried out, quality assessment result is generated.If quality assessment result is judged for unqualified, according to quality assessment result pair
Image is adjusted to qualified, if still unqualified, needs upload of taking pictures again.User need to not be inconsistent in quality assessment result
Item is closed, targetedly shooting image uploads onto the server again, efficiently to obtain qualified image.
In one embodiment, as shown in fig. 6, the image after to being processed through printed page analysis carries out background and frame removal,
Specifically include：
Step 324a, to being processed through printed page analysis after image carry out noise reduction and remove texture.
Texture on business license is generally the texture of light color, therefore using meanshift algorithms to line small in image
Reason floating is filtered.Meanshift algorithms are also called mean shift algorithm.Image is carried out after noise reduction removes texture, then carried out follow-up
Treatment, it will improve the accuracy of image recognition.
Step 324b, to removing texture through noise reduction after image carry out rim detection, draw the edge of image.
Rim detection is carried out to image using canny Edge checks, the edge of captured business license in image is detected.
Canny Edge Detections are a kind of edge detection algorithms calculated based on image gradient.
Step 324c, straight-line detection is carried out to edge, draws the boundary straight line of image.
According to the edge of the business license for detecting, the possible boundary straight line of business license is gone out using hough straight-line detections,
Angle of inclination, relative distance and angle further according to straight line filter out most possible boundary straight line.But can't reject and side
The frame straight line of boundary's straight line parallel, is now accomplished by carrying out consistency detection to the edge of image flat with boundary straight line to reject
Capable frame straight line, so that it is determined that the four edges circle straight line of image.
Step 324d, the intersection point of boundary straight line is determined according to boundary straight line, to boundary straight line and the intersection point structure of boundary straight line
Into inframe the image that goes after texture of noise reduction, carry out perspective transform, obtain removing the image of background and frame.
Four intersection points of boundary straight line are determined according to four edges circle straight line, four edges circle straight line and four intersection points are constituted
The image of inframe, carries out perspective transform, obtains removing the image of background and frame.Perspective transform is that picture projection is new to one
View plane, also referred to as projection mapping.
In the present embodiment, image is carried out after noise reduction removes texture, texture small in image being floated and being filtered, then entered
Row subsequent treatment, it will improve the accuracy of image recognition.Can not be positive figure by shooting angle by perspective transform
As being adjusted to the image that shooting angle shoots for front.By above-mentioned pretreated image, picture quality is optimized, significantly
Improve the accuracy of subsequent treatment.
In one embodiment, as shown in fig. 7, carrying out line of text detection and row to the image after being removed through background and frame
Cutting, obtains line of text image, specifically includes：
Step 326a, the image to removal background and frame carries out greyscale transformation.
Server removes color information to through the image after background and frame removal, first carrying out greyscale transformation.
Step 326b, to carrying out text edges detection through the image after greyscale transformation, obtains text filed.
Text edges detection is carried out to image using canny Edge checks, protrusion obtains the edge of word on business license
Detection figure, it is as text filed.
Step 326c, is expanded text filed by row, is obtained with the Connection operator block of row connection.
Again to each it is text filed carry out by row expand, specially expanded according to pixel column, obtain with pixel column connect
The Connection operator block for connecing.
I (i, (j-k))=255 (k=0 ..., n)
The minimum value of width (width) of the wherein n less than image I and height (height), uses through test of many times herein
N=5, the occurrence of n is not limited.I-th row of I (i, (j-k))==255 expression image I, the gray value of jth-k row pixels
During equal to 255, i.e. I (i, (j-k))==255 it is injunctive be it is true, then it is 0 to n to be counted as the span of 1, k；Thresh is
The threshold value of setting, when the numerical value on the left side in formula (9) is more than thresh threshold values, I (i, j-k) will be entered as in formula (10)
255, carried out line slip expansion.
Step 326d, connected domain detection is carried out according to character features to Connection operator block, filters out text filed connection
Domain.
Connected domain detection is carried out to Connection operator block using character features, non-legible Connection operator block is rejected, filtered out
Text filed connected domain.Character features specifically include the ratio of width to height of line of text, height etc..
Step 326e, determines the boundary rectangle of text filed connected domain, and every trade cutting is entered to text according to boundary rectangle,
Obtain line of text image.
Determine the boundary rectangle of text filed connected domain, every trade cutting is entered to text according to boundary rectangle, obtain a line
The line of text image of a line.Can be identified by row during follow-up identification, such that it is able to the semanteme of associated context.
In the present embodiment, greyscale transformation is first carried out, color information is removed,
In one embodiment, there is provided a kind of business license recognition methods, the method is being applied to ring as shown in Figure 1
It is illustrated in border.
Business license identification function is made into micro services and is placed on high in the clouds, then open API is provided, user accesses the API
To in the application of oneself, it is possible to use business license identification function.For example, the API can be accessed the business of oneself for bank
System, such that it is able to directly invoke business license recognition result, seamless linking.Can be used under any operating system, do not received
The limitation of operating system.
When banking person is shot by the camera of computer to business license, business license original graph is obtained
Picture, business license original image is uploaded by being transmitted to server on local.Server receives the original image, and load
Weighing apparatus, distributes to the original image corresponding server and is processed.
Because the original image that user uploads not all is ajusted, therefore firstly the need of according to the word side on business license
Original image is ajusted to national emblem badge information, seal information etc. in information and figure, you can 90 degree, 180 degree are done to original image
Rotation ajust, or other are arbitrarily angled also according to rotation is needed.So that the text information on original image be all vertically to
On, facilitate follow-up Text region to work.Then to ajusting after the color of image, definition, colour cast and brightness etc. carry out
Quality evaluation, generates quality assessment result.Quality assessment result shows qualified, proceeds subsequent treatment, if unqualified,
Incongruent project is processed, to cause image qualified.Image for there is colour cast needs white by carrying out to picture
Balance Treatment is with so that image is qualified.If image some projects cannot pass through the later stage, and treatment is qualified to reach, then comment quality
Estimate result and feed back to the computer for uploading image.So that Non-Compliance of the banking person in quality assessment result, is directed to
Property shooting image, and upload until qualified again.
Further, server carries out noise reduction and removes texture to qualified image using meanshift algorithms, will be small in image
Texture floating filter fall.Recycle canny Edge checks carries out rim detection to image, detects captured business in image
The edge of license.Hough straight-line detections are utilized according to the edge for detecting, the possible boundary straight line of business license, then root is drawn
Most possible boundary straight line is filtered out according to the angle of inclination of straight line, relative distance and angle.Now, however it remains straight with border
The parallel unnecessary frame straight line of line, the edge to image carries out consistency detection to reject the frame parallel with boundary straight line
Straight line, so that it is determined that the four edges circle straight line of image.Four intersection points of boundary straight line are determined according to four edges circle straight line, to four
Boundary straight line and four images of the inframe of intersection point composition, carry out perspective transform, obtain removing the image of background and frame.
Further, the image after server through background and frame to removing, first carries out greyscale transformation, removal color letter
Breath.Recycling canny Edge checks carries out text edges detection to image, and protrusion obtains the rim detection of word on business license
Figure, it is as text filed.Again to each it is text filed carry out by row expand, specially expanded according to pixel column, obtain with
The Connection operator block of pixel column connection.Connected domain inspection is carried out to Connection operator block using features such as the ratio of width to height of line of text, height
Survey, reject non-legible Connection operator block, such as the mark for being added with pen etc., filter out text filed connected domain.Really
The boundary rectangle of fixed text filed connected domain, according to boundary rectangle to it is text filed enter every trade cutting, obtain a line a line
Line of text image.
Further, real character, such as 10 Arabic numerals, 34 provinces in a part of business license are chosen in advance
Part information, type information, composition form information etc. are trained to BP neural network.Then using the BP nerves that precondition is good
Network generates recognition result to carrying out Classification and Identification in line of text image by row.Banking system can directly invoke battalion
Industry license recognition result.
In one embodiment, as shown in figure 8, additionally providing a kind of business license identifying device, the device includes：It is original
Image collection module 810, original image identification module 820, recognition result output module 830.
Original image acquisition module 810, request is recognized for the business license that receiving terminal is uploaded, and business license identification please
The original image for asking carrying to be obtained after being taken pictures to business license；
Original image identification module 820, for business license identification request to be assigned into clothes by the way of load balancing
Business device is with to carrying out Text region, the recognition result of generation business license to the original image for carrying；
Recognition result output module 830, for recognition result to be stored in into server, and is back to end by recognition result
End.
In one embodiment, as shown in figure 9, original image identification module 820 includes：Printed page analysis processing module 822,
Background and frame removal module 824, line of text image collection module 826, line of text picture recognition module 828.
Printed page analysis processing module 822, for carrying out printed page analysis treatment to the original image for receiving；
Background and frame removal module 824, background and frame removal are carried out for the image after to being processed through printed page analysis；
Line of text image collection module 826, for the image after to being removed through background and frame carry out line of text detection and
Row cutting, obtains line of text image；
Line of text picture recognition module 828, for entering style of writing to line of text image according to the good neutral net of training in advance
Word is recognized, and generates recognition result by row.
In one embodiment, as shown in Figure 10, printed page analysis processing module 822 includes：Module 822a, quality is ajusted to comment
Estimate module 822b, feedback module 822c.
Module 822a is ajusted, for obtaining the anglec of rotation according to the words direction information on original image and graphical information,
Original image is ajusted according to the anglec of rotation；
Quality assessment modules 822b, in the color to the image after ajusting, definition, colour cast and brightness at least
One kind carries out quality evaluation, generates quality assessment result；
Feedback module 822c, for quality assessment result to be back into terminal.
In one embodiment, as shown in figure 11, background and frame removal module 824 include：Noise reduction goes texture module
824a, Image Edge-Detection module 824b, straight-line detection module 824c, perspective transform module 824d.
Noise reduction removes texture module 824a, and carrying out noise reduction for the image after to being processed through printed page analysis removes texture；
Image Edge-Detection module 824b, rim detection is carried out for the image after to removing texture through noise reduction, draws image
Edge；
Straight-line detection module 824c, for carrying out straight-line detection to edge, draws the boundary straight line of image；
Perspective transform module 824d, the intersection point for determining boundary straight line according to boundary straight line, to boundary straight line and border
The image that the noise reduction of the inframe that the intersection point of straight line is constituted is gone after texture, carries out perspective transform, obtains removing the figure of background and frame
Picture.
In one embodiment, as shown in figure 12, line of text image collection module 826 includes：Greyscale transformation module 826a,
Text edges detection module 826b, Connection operator block acquisition module 826c, connected domain detection module 826d, row cutting module
826e。
Greyscale transformation module 826a, greyscale transformation is carried out for the image to removal background and frame；
Text edges detection module 826b, for carrying out text edges detection through the image after greyscale transformation, obtaining text
One's respective area；
Connection operator block acquisition module 826c, for being expanded text filed by row, obtains with the Connection operator of row connection
Block；
Connected domain detection module 826d, for carrying out connected domain detection to Connection operator block according to character features, filters out
Text filed connected domain；
Row cutting module 826e, the boundary rectangle for determining text filed connected domain, according to boundary rectangle to text
Enter every trade cutting, obtain line of text image.
Embodiment described above only expresses several embodiments of the invention, and its description is more specific and detailed, but simultaneously
Can not therefore be construed as limiting the scope of the patent.It should be pointed out that coming for one of ordinary skill in the art
Say, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to protection of the invention
Scope.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (10)
1. a kind of business license recognition methods, methods described includes：
The business license that receiving terminal is uploaded recognizes request, and the business license identification request is carried after being taken pictures to business license and obtained
The original image for arriving；
Use the mode of load balancing that business license identification request is assigned into server with to described in described pair of carrying
Original image carries out Text region, generates the recognition result of business license；
The recognition result is stored in server, and the recognition result is back to terminal.
2. method according to claim 1, it is characterised in that it is described by the way of load balancing by the business license
Identification request is assigned to server and carries out Text region with the original image to described pair of carrying, generates the knowledge of business license
Other result, including：
The original image to receiving carries out printed page analysis treatment；
Image after to being processed through printed page analysis carries out background and frame removal；
Image after to being removed through background and frame carries out line of text detection and row cutting, obtains line of text image；
Text region is carried out to the line of text image according to the good neutral net of training in advance, and by row generation recognition result.
3. method according to claim 2, it is characterised in that the described pair of original image for receiving carries out the space of a whole page point
Analysis is processed, including：
The anglec of rotation is obtained according to the words direction information on the original image and graphical information, according to the anglec of rotation pair
The original image is ajusted；
At least one in color, definition, colour cast and brightness to the image after ajusting carries out quality evaluation, generates quality
Assessment result；
The quality assessment result is back to terminal.
4. method according to claim 2, it is characterised in that described pair processed through printed page analysis after image carry out background
And frame removal, including：
Image after to being processed through printed page analysis carries out noise reduction and removes texture；
Image after to removing texture through noise reduction carries out rim detection, draws the edge of described image；
Straight-line detection is carried out to the edge, the boundary straight line of described image is drawn；
The intersection point of the boundary straight line is determined according to the boundary straight line, to the boundary straight line and the intersection point of the boundary straight line
The image that the noise reduction of the inframe of composition is gone after texture, carries out perspective transform, obtains removing the image of background and frame.
5. method according to claim 2, it is characterised in that described pair removed through background and frame after image enter style of writing
One's own profession is detected and row cuts, and obtains line of text image, including：
Image to removal background and frame carries out greyscale transformation；
To carrying out text edges detection through the image after greyscale transformation, obtain text filed；
Will be described text filed by row expansion, obtain with the Connection operator block of row connection；
Connected domain detection is carried out to the Connection operator block according to character features, text filed connected domain is filtered out；
Determine the boundary rectangle of the text filed connected domain, every trade cutting is entered to text according to the boundary rectangle, obtain
Line of text image.
6. a kind of business license identifying device, it is characterised in that described device includes：
Original image acquisition module, request, the business license identification request are recognized for the business license that receiving terminal is uploaded
The original image that carrying is obtained after being taken pictures to business license；
Original image identification module, for business license identification request to be assigned into server by the way of load balancing
Text region is carried out with the original image to described pair of carrying, the recognition result of business license is generated；
Recognition result output module, for the recognition result to be stored in into server, and is back to end by the recognition result
End.
7. device according to claim 6, it is characterised in that the original image identification module includes：
Printed page analysis processing module, for carrying out printed page analysis treatment to the original image for receiving；
Background and frame removal module, background and frame removal are carried out for the image after to being processed through printed page analysis；
Line of text image collection module, line of text detection and row cutting are carried out for the image after to being removed through background and frame,
Obtain line of text image；
Line of text picture recognition module, for carrying out word knowledge to the line of text image according to the good neutral net of training in advance
Not, and by row generation recognition result.
8. device according to claim 7, it is characterised in that the printed page analysis processing module includes：
Module is ajusted, for obtaining the anglec of rotation according to the words direction information on the original image and graphical information, according to
The anglec of rotation is ajusted to the original image；
Quality assessment modules, are carried out at least one in the color to the image after ajusting, definition, colour cast and brightness
Quality evaluation, generates quality assessment result；
Feedback module, for the quality assessment result to be back into terminal.
9. device according to claim 7, it is characterised in that the background and frame removal module include：
Noise reduction goes texture module, and carrying out noise reduction for the image after to being processed through printed page analysis removes texture；
Image Edge-Detection module, rim detection is carried out for the image after to removing texture through noise reduction, draws the side of described image
Edge；
Straight-line detection module, for carrying out straight-line detection to the edge, draws the boundary straight line of described image；
Perspective transform module, the intersection point for determining the boundary straight line according to the boundary straight line, to the boundary straight line and
The image that the noise reduction of the inframe that the intersection point of the boundary straight line is constituted is gone after texture, carries out perspective transform, obtains the removal back of the body
The image of scape and frame.
10. device according to claim 7, it is characterised in that the line of text image collection module includes：
Greyscale transformation module, greyscale transformation is carried out for the image to removal background and frame；
Text edges detection module, for carrying out text edges detection through the image after greyscale transformation, obtaining text filed；
Connection operator block acquisition module, for will be described text filed by row expansion, obtains with the Connection operator block of row connection；
Connected domain detection module, for carrying out connected domain detection to the Connection operator block according to character features, filters out text
The connected domain in region；
Row cutting module, the boundary rectangle for determining the text filed connected domain, according to the boundary rectangle to text
Enter every trade cutting, obtain line of text image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611265344.0A CN106846011A (en) | 2016-12-30 | 2016-12-30 | Business license recognition methods and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611265344.0A CN106846011A (en) | 2016-12-30 | 2016-12-30 | Business license recognition methods and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106846011A true CN106846011A (en) | 2017-06-13 |
Family
ID=59117876
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611265344.0A Pending CN106846011A (en) | 2016-12-30 | 2016-12-30 | Business license recognition methods and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106846011A (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107145879A (en) * | 2017-06-23 | 2017-09-08 | 依通（北京）科技有限公司 | A kind of floristics automatic identifying method and system |
| CN108256530A (en) * | 2017-12-29 | 2018-07-06 | 北京城市网邻信息技术有限公司 | Image-recognizing method, device and equipment |
| CN108764226A (en) * | 2018-04-13 | 2018-11-06 | 顺丰科技有限公司 | Image text recognition methods, device, equipment and its storage medium |
| CN108921158A (en) * | 2018-06-14 | 2018-11-30 | 众安信息技术服务有限公司 | Method for correcting image, device and computer readable storage medium |
| CN109034050A (en) * | 2018-07-23 | 2018-12-18 | 顺丰科技有限公司 | ID Card Image text recognition method and device based on deep learning |
| CN109670423A (en) * | 2018-12-05 | 2019-04-23 | 依通（北京）科技有限公司 | A kind of image identification system based on deep learning, method and medium |
| CN110070045A (en) * | 2019-04-23 | 2019-07-30 | 杭州智趣智能信息技术有限公司 | A kind of text recognition method of business license, system and associated component |
| CN110135431A (en) * | 2019-05-16 | 2019-08-16 | 深圳市信联征信有限公司 | The automatic identifying method and system of business license |
| CN110473015A (en) * | 2019-08-09 | 2019-11-19 | 南京智骋致想电子科技有限公司 | A kind of smart ads system and advertisement placement method |
| CN112396050A (en) * | 2020-12-02 | 2021-02-23 | 上海优扬新媒信息技术有限公司 | Image processing method, device and storage medium |
| CN112766275A (en) * | 2021-04-08 | 2021-05-07 | 金蝶软件（中国）有限公司 | Seal character recognition method and device, computer equipment and storage medium |
| CN115376142A (en) * | 2022-07-20 | 2022-11-22 | 北大荒信息有限公司 | Image-based business license information extraction method, computer equipment and readable storage medium |
| CN117935296A (en) * | 2024-02-06 | 2024-04-26 | 广东度才子集团有限公司 | Employment quality report generation system |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101510258A (en) * | 2009-01-16 | 2009-08-19 | 北京中星微电子有限公司 | Certificate verification method, system and certificate verification terminal |
| CN103208004A (en) * | 2013-03-15 | 2013-07-17 | 北京英迈杰科技有限公司 | Automatic recognition and extraction method and device for bill information area |
| CN103678305A (en) * | 2012-08-31 | 2014-03-26 | 北京网秦天下科技有限公司 | Method and system for displaying inquiring information based on image recognition |
| CN103810485A (en) * | 2014-01-22 | 2014-05-21 | 深圳市东信时代信息技术有限公司 | Recognition device, character recognition system and method |
| CN104079587A (en) * | 2014-07-21 | 2014-10-01 | 深圳天祥质量技术服务有限公司 | Certificate identification device and certificate check system |
| US8879783B1 (en) * | 2013-06-28 | 2014-11-04 | Google Inc. | Comparing extracted card data with user data |
| CN104408450A (en) * | 2014-11-21 | 2015-03-11 | 深圳天源迪科信息技术股份有限公司 | Identity card identifying method, device and system |
| CN105528601A (en) * | 2016-02-25 | 2016-04-27 | 华中科技大学 | Identity card image acquisition and recognition system as well as acquisition and recognition method based on contact type sensor |
-
2016
- 2016-12-30 CN CN201611265344.0A patent/CN106846011A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101510258A (en) * | 2009-01-16 | 2009-08-19 | 北京中星微电子有限公司 | Certificate verification method, system and certificate verification terminal |
| CN103678305A (en) * | 2012-08-31 | 2014-03-26 | 北京网秦天下科技有限公司 | Method and system for displaying inquiring information based on image recognition |
| CN103208004A (en) * | 2013-03-15 | 2013-07-17 | 北京英迈杰科技有限公司 | Automatic recognition and extraction method and device for bill information area |
| US8879783B1 (en) * | 2013-06-28 | 2014-11-04 | Google Inc. | Comparing extracted card data with user data |
| CN103810485A (en) * | 2014-01-22 | 2014-05-21 | 深圳市东信时代信息技术有限公司 | Recognition device, character recognition system and method |
| CN104079587A (en) * | 2014-07-21 | 2014-10-01 | 深圳天祥质量技术服务有限公司 | Certificate identification device and certificate check system |
| CN104408450A (en) * | 2014-11-21 | 2015-03-11 | 深圳天源迪科信息技术股份有限公司 | Identity card identifying method, device and system |
| CN105528601A (en) * | 2016-02-25 | 2016-04-27 | 华中科技大学 | Identity card image acquisition and recognition system as well as acquisition and recognition method based on contact type sensor |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107145879A (en) * | 2017-06-23 | 2017-09-08 | 依通（北京）科技有限公司 | A kind of floristics automatic identifying method and system |
| CN108256530B (en) * | 2017-12-29 | 2021-12-07 | 北京城市网邻信息技术有限公司 | Image recognition method, device and equipment |
| CN108256530A (en) * | 2017-12-29 | 2018-07-06 | 北京城市网邻信息技术有限公司 | Image-recognizing method, device and equipment |
| CN108764226A (en) * | 2018-04-13 | 2018-11-06 | 顺丰科技有限公司 | Image text recognition methods, device, equipment and its storage medium |
| CN108921158A (en) * | 2018-06-14 | 2018-11-30 | 众安信息技术服务有限公司 | Method for correcting image, device and computer readable storage medium |
| CN109034050A (en) * | 2018-07-23 | 2018-12-18 | 顺丰科技有限公司 | ID Card Image text recognition method and device based on deep learning |
| CN109670423A (en) * | 2018-12-05 | 2019-04-23 | 依通（北京）科技有限公司 | A kind of image identification system based on deep learning, method and medium |
| CN110070045A (en) * | 2019-04-23 | 2019-07-30 | 杭州智趣智能信息技术有限公司 | A kind of text recognition method of business license, system and associated component |
| CN110135431A (en) * | 2019-05-16 | 2019-08-16 | 深圳市信联征信有限公司 | The automatic identifying method and system of business license |
| CN110473015A (en) * | 2019-08-09 | 2019-11-19 | 南京智骋致想电子科技有限公司 | A kind of smart ads system and advertisement placement method |
| CN112396050A (en) * | 2020-12-02 | 2021-02-23 | 上海优扬新媒信息技术有限公司 | Image processing method, device and storage medium |
| CN112396050B (en) * | 2020-12-02 | 2023-09-15 | 度小满科技(北京)有限公司 | Image processing method, device and storage medium |
| CN112766275A (en) * | 2021-04-08 | 2021-05-07 | 金蝶软件（中国）有限公司 | Seal character recognition method and device, computer equipment and storage medium |
| CN112766275B (en) * | 2021-04-08 | 2021-09-10 | 金蝶软件（中国）有限公司 | Seal character recognition method and device, computer equipment and storage medium |
| CN115376142A (en) * | 2022-07-20 | 2022-11-22 | 北大荒信息有限公司 | Image-based business license information extraction method, computer equipment and readable storage medium |
| CN115376142B (en) * | 2022-07-20 | 2023-09-01 | 北大荒信息有限公司 | Image-based business license information extraction method, computer equipment and readable storage medium |
| CN117935296A (en) * | 2024-02-06 | 2024-04-26 | 广东度才子集团有限公司 | Employment quality report generation system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106846011A (en) | Business license recognition methods and device | |
| US20210224577A1 (en) | Classification with segmentation neural network for image-based content capture | |
| CN104463209B (en) | Method for recognizing digital code on PCB based on BP neural network | |
| CN109902715B (en) | Infrared dim target detection method based on context aggregation network | |
| CN109871845B (en) | Certificate image extraction method and terminal equipment | |
| CN106096610B (en) | A kind of file and picture binary coding method based on support vector machines | |
| CN106650721A (en) | Industrial character identification method based on convolution neural network | |
| CN110119687A (en) | Detection method based on the road surface slight crack defect that image procossing and convolutional neural networks combine | |
| CN111460947B (en) | BP neural network-based method and system for identifying metal minerals under microscope | |
| CN112651333A (en) | Silence living body detection method and device, terminal equipment and storage medium | |
| CN101650780A (en) | Identification method of container number | |
| US20230169554A1 (en) | System and method for automated electronic catalogue management and electronic image quality assessment | |
| CN111209858B (en) | Real-time license plate detection method based on deep convolutional neural network | |
| CN110135446B (en) | Text detection method and computer storage medium | |
| CN114444565B (en) | Image tampering detection method, terminal equipment and storage medium | |
| Jhang et al. | CNN training for face photo based gender and age group prediction with camera | |
| Cheddad | On box-cox transformation for image normality and pattern classification | |
| CN110569716A (en) | A kind of shelf image remake detection method | |
| Lecca et al. | GREAT: a gradient-based color-sampling scheme for Retinex | |
| Hussain et al. | Few-shot based learning recaptured image detection with multi-scale feature fusion and attention | |
| Yao et al. | End-to-end adaptive object detection with learnable Retinex for low-light city environment | |
| Joshi et al. | Source printer identification from document images acquired using smartphone | |
| CN112837329B (en) | A method and system for image binarization of Tibetan ancient book documents | |
| CN117636045B (en) | Wood defect detection system based on image processing | |
| CN112749731A (en) | Bill quantity identification method and system based on deep neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170613 |