CN103617422B - A social relation management method based on business card recognition - Google Patents
A social relation management method based on business card recognition Download PDFInfo
- Publication number
- CN103617422B CN103617422B CN201310521182.2A CN201310521182A CN103617422B CN 103617422 B CN103617422 B CN 103617422B CN 201310521182 A CN201310521182 A CN 201310521182A CN 103617422 B CN103617422 B CN 103617422B
- Authority
- CN
- China
- Prior art keywords
- contact person
- node
- character
- user
- address
- 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.)
- Active
Links
Landscapes
- Character Input (AREA)
Abstract
Provided is a social relation management method based on business card recognition. The social relation management method comprises four steps: a step 1 of entering business card information, performing image acquisition on the business card with a camera or a scanner; dividing character blocks according to the characteristics of the business card image and performing character recognition with an OCR engine; performing word-dividing processing according to a key field; classifying and entering extracted information; filling the extracted information in corresponding forms; and interacting with a user in order to perform artificial confirmation and adjustment on possibly existed incapably-recognized information, a step 2 of establishing a social relation network, a step 3 of achieving intelligent retrieval, and a step 4 completing mobile terminal synchronization.
Description
Technical field
The present invention relates to card information Rapid input computer and the management of social networks that formed by business card
Method.
Background technology
Economic fast development promotes human communication day by day frequent, and a slight business card carries actually
It is resource, is business opportunity, it might even be possible to say it is benefit.
Although mobile phone at present, PDA (Personal Digital Assistant), panel computer, notebook even platform
Formula PC all can install corresponding name card management system, but the management method that these systems are used could not be well
Solving one is the most at all, also sixty-four dollar question, i.e. social networks management is produced by substantial amounts of contact person
Raw complicated inter personal contact, usually allow people cannot find with the fastest speed want cooperation or ask for help suitable
Person.
The method for managing name cards of main flow all uses the mode of packet to carry out business card classifying at present.Such as, contact person is pressed
The key word such as " friend ", " client ", " colleague ", " leader " is grouped.This grouping management mode is confined to tradition
Custom, the most directly perceived, and a suitable contact approach cannot be provided based on inter personal contact, can only
Can not be for " thing " for " people ".In inter personal contact, being usually present a kind of situation, someone had one
The edge in face, is not familiar with, and its name or the information such as phone number or work unit simply know wherein clip information,
The most only know he (she) at city one hospital work, but other information are difficult to remember, when in the address list of " I "
When having thousands of contact persons, it is difficult to find this contact person, therefore, it is necessary to provide more intelligent method
It is managed.The friend of side of the edge usually occurs in certain feature occasion, or closeer with oneself relation with one
Close friend is relevant, or is introduced by this friend, therefore, can be searched for by following the clues by certain Intelligentize query,
Find this friend.Such as: user now wants to go to a hospital to see a doctor, it is desirable to relate to a hospital.But press
According to traditional mode, he can only first position a suitable packet, further each and every one checks either with or without friend doctor,
The location obscured cannot be realized.And the most common situation is: user has the business card of friend doctor, but with
It is not yet done, needs the recommendation of one or more friend.At this moment, traditional grouping management mode just cannot provide one
Bar suitably recommends approach, and needs the user effort plenty of time to go thinking, arranges.And artificial arrangement is often
Being incomplete, thinkable recommendation approach is also not necessarily most suitable.This is also a large number of users head the most
The problem of pain, is also current various business cards and the insurmountable problem of social networks management method.Additionally, it is right
Friend's circle in oneself also cannot be given with social networks management method by conventional business card with oneself intimate degree
Go out, therefore, it is difficult to carry out efficient doings.
End is got up, and emphasis of the present invention solves the innovative problems of three aspects: (1) based on computer vision technique with
And semanteme segmenting method solves the Rapid input of card information;(2) intelligent positioning of social networks and optimum
Recommendation approach;(3) intelligent synchronization between mobile terminal and system database.
Summary of the invention
Instant invention overcomes the problems referred to above that the business card of employing at present exists with social networks management method, it is achieved that
Card information automatic input based on computer vision and the intelligent positioning of social networks, save user artificial
Time considered, and be provided that a most suitable contact channel and recommendation approach, provide optimum, most reliable
Contact scheme.
The technical solution adopted for the present invention to solve the technical problems includes four steps:
Step 1. typing card information
Utilize photographic head or scanner to carry out business card image collection, mark off character block according to the feature of business card image
And utilize OCR engine to carry out character recognition, carry out word segmentation processing according to critical field, the information extracted is returned
Class typing, is inserted corresponding list, finally interacts with user, believes None-identified that may be present
Breath carries out manual confirmation and adjustment.Described flow process should comprise following operation:
1.1, gray processing.First the business card image to photographic head or scanner collection uses weighted mean method to carry out ash
Degreeization:
f(i,j)=0.30R(i,j)+0.59G(i,j)+0.11B(i,j)
Wherein (i, j) represent pixel coordinate, in coloured image three components of red, green, blue be respectively R (i, j), G (i, j),
(i, j), (i j) is then the gray value of this point to f to B.
1.2, edge extracting.In conjunction with Sobel operator and LOG (Laplacian of Gauss) operator, formed and revise
LOG algorithm, i.e. chooses following Sobel operator (SxDifference, S are divided in center for horizontal directionyFor vertical direction
Center divide difference, choose wherein higher value as gradient S)
As the precondition of rim detection, reduce unnecessary zero cross point.Recycling is if minor function is as filtering
Device, carries out LOG detection.
LOG operator template is as follows
LOG operator template therein is the digital form of LOG, is convolution, σ as interior collecting image
Mean square deviation for Gauss distribution.
1.3, binaryzation.Calculate threshold value, and gray-scale map is carried out binaryzation.If a height of h of gray-scale map, a width of w,
Threshold value is tried to achieve, i.e. average gray Threshold by following formula:
Binaryzation is carried out further according to threshold value Threshold:
1.4, tilt detection and rectification.Edge by Hough transformation (Hough Transform) detection business card frame
Line, it is thus achieved that name panel region also judges that the angle of inclination of business card is corrected.
Owing to the linear equation of y=kx+b form cannot represent that (c is constant, i.e. with x-axis for the straight line of x=c form
Parallel straight line, slope k → ∞).Therefore use parametric equation ρ=x*cos θ+y*sin θ, Qi Zhongtong here
Cross and choose p1(x1,y1), p2(x2,y2) two monitoring points, following formula can obtain tiltangleθ:
According to tiltangleθ, artwork and bianry image carried out simultaneously affine transformation (using homogeneous matrix to represent),
P'=P R, obtains correcting matrix by-θ substitution 2D spin matrix By point Obtain after rectification.Concrete operation is as follows:
x'=xcosθ+ysinθ
y'=-xsinθ+ycosθ
1.5, image segmentation.Feature according to business card image marks off character block.Comprise the following steps that
1.5.1, definition detection density Density, for current pixel up and down and on clinodiagonal totally 8 directions
Neighbor in the quantity of black picture element, computing formula is as follows:
Wherein i ∈ [x-1, x+1] ∩ N*,j∈[y-1,y+1]∩N*, (x y) is current sensing point coordinate, N*For
Positive integer collection.
1.5.2, the image of name panel region is converted into density matrix, removes remaining noise.Operational approach is as follows:
Judging the density of each pixel one by one, as Density, < when 2, homography element is designated as 0, i.e. as noise
Process.As Density >=2 time, homography element is designated as 1, show this pixel be character block a part.
1.5.3, pass through conversion formula Binary single color figure after correcting turns
Turn to following two-dimensional array form:
1.5.4, according to density matrix, character block region, location (so far has two kinds of strategies to be respectively applied to quickly know
Other pattern and accurate recognition mode), then according to region, business card image is split.
Quickly recognition strategy: judge density matrix line by line, the ratio that interior " 1 " element of every a line accounts for unit number exceedes
Certain threshold value, then be considered as line of text, be then considered as blank less than this threshold value.
Precisely recognition strategy: judge density matrix line by line, connects to " detection line " by interior for row continuous print " 0 " element,
Determine whether text filed according to " detection line " depth of initiating terminal, difference in length with terminal position feature.I.e.
Remove the length detection line less than threshold value;Then end and the Article 2 detection line of Article 1 detection line are marked
Top;After all row carry out detection and labelling, labelling the region surrounded is the most text filed.
1.5.5, definition Ri∈{d(i,y)|y∈[0,h]∩N*,And if only ifTime record j value.One group of region [j of last gained1,j2], [j2,j3], [j3,j4]…
The most some character blocks, wherein h represents that line number, w represent columns, N*For positive integer collection, RiFor step 3 two
The element of the i-th row, Sum in dimension groupjFor RiBe expert at all elements sum.
1.6, character recognition.OCR (Optical Character Recognition) technology is utilized to extract each literary composition
The information of block.Here the OCR module pair of MODI (Microsoft Office Document Imaging) is used
The image being partitioned into is identified one by one, and each character block image is all become one group of word.I.e. to character block
[j1,j2], [j2,j3], [j3,j4] ... call OCR recognition engine one by one, obtain the character corresponding with each region
Collection C1,2,C2,3,C3,4…。
1.7, word segmentation processing.Word segmentation processing is carried out, to the information categorization typing extracted according to critical field.This needs
Keywords database to be set up and semantic base, such as: company, position, address, telephone number, email, road,
Number etc..Definition of keywords set W={ company, position, address, telephone number, email ... and definition of keywords
Semantic base, such as Address, Addr, address, contact address etc. are the near synonym of address, i.e. exist
It is semantically consistent, sets up corresponding mapping relations, form semantic base.
According to key word, based on contextual feature, extract and often organize the corresponding informance in word, insert respective table
Single, complete word segmentation processing.Specifically comprise the following steps that
1.7.1, find ": " separator, utilize separator to define key word and content.If the key defined
Word is not then transferred to the user decide whether that being regarded as key word is indexed in set in set of keywords.By this
Step realizes the learning functionality of participle strategy.
1.7.2, keyword Key=C is then matchedx,y∩ W, by character stringInsert entitled
In the form item of Key.WhenTime, by character set Cx,yInsert " unfiled " form item.
Strategy based on semantic base includes:
Location name strategy (Chinese): getting rid of other semantemes, number of characters, between 2~4, is followed by position, head
Rank, or with " name " class label;
Locating cellphone number strategy: with 11 pure digi-tal character strings of 13,15,18 beginnings, or with " hands
Machine " class label;
Location telephone number strategy: 7~8 pure digi-tal character strings, or start with area code, comprise bracket, loigature
The separators such as symbol, space, or with " phone " class label;
Location Business Name strategy: get rid of other semantemes, occurs that " company ", " group ", each institutional settings are crucial
Word, or with " company " class label;
Positioning address strategy: occur " province ", " city ", " county ", " township ", " town ", " road ", " district ", " building ",
The character string that " unit ", " room " mix with numeral, or with " address " class label;
Location postcode strategy: 6 pure digi-tal character strings, or with " postcode " class label;
Location mailbox strategy: numeral, letter, " ", ". " character occur, or with " mailbox " class label;
Location network address strategy: " http ", ". ", " www ", " com ", " cn ", " edu " character occur, or
With " network address " class label.
1.8, manual synchronizing.After participle completes, need to interact with user, None-identified that may be present is believed
Breath carries out manual confirmation and adjustment, i.e. " unfiled " form item is carried out manual sort, after categorizing process terminates,
To automatically learn to enter semantic base.
Step 2. sets up social networks network
Its step is as follows:
2.1, the relation of two contact persons is set up by moving contact person's node with mouse.The node that will be newly added
AjIt is attached to existing node AiOn (i.e. wanting the contact person of opening relationships), after determining operation both sides, utilize line
Form represents relation L set up between two contact personsi,j, it represents AiWith AjUnderstanding (two-way) each other.
The most optional) if newly-established contact person node AiIt is the first contact people of user O (i.e. " I "),
Then by new node AiIt is connected on user node O set up contact circuit OAi, its relation is L0,i。
The most optional) if newly-established contact person node AjIt is that user O is by friend AiKnow this connection
It is people, and in the case of the other side does not recognize user O, then by new node AjIt is connected to user friend node Ai
Upper foundation contacts circuit AiAj, its relation is Li,j。
The most optional) if being originally user O by friend AiThe contact person A learntjWith user friend Ai
It is the first contact people (unidirectional), and after further cooperation and exchange, user O and this contact person AjBecome
One contact people (two-way), then by this contact person node AjIt is connected on user node O, is formed and user's node
Directly internuncial pathway OAj, its relation is L0,j。
2.2, for each relation Li,jBy using weights Ki,jRepresent the power of relation.Li,jWeights Ki,jTake
Certainly in contact person AiWith contact person AjTime of getting to know T(unit: sky) and contact number of times Count.Specifically
Computational methods are as follows:
2.3, " cohesion " Intimacy of " I " and each contact person is drawn according to number of degrees D and weights K.Agreement is used
The number of degrees D of family node O0=0 passes through path OAiOn relation L0,iThe node A being connectediNumber of degrees Di=1+D0;
By path AiAjOn relation Li,jThe node A being connectedjNumber of degrees Dj=1+Di." I " and node Aj" parent
Density " computing formula is:
Intimacyj=Dj+Ki,j,Di∈N*,Ki,j∈[0,1)
The integer part of " cohesion " Intimacy represents the number of degrees in six degrees of separation, and fractional part will be with isocratic
Each relation Further Division under several is opened.Therefore, " cohesion " can preferably reflect and arrive centered by " I "
The close and distant relation of each node, wherein N*For positive integer collection.
Step 3. realizes intelligent retrieval
The search key inputted by user carries out fuzzy matching with the personal information of typing, filters out possibility
Object listing, it specifically comprises the following steps that
3.1, the key word setting user's input is Char as character string WORD, element therein, business card data collection
DATA, ifAndThen return the contact person of its correspondence
A(Char∩DATA)。
3.2, user is from the list returned, and according to the actual requirements, selects objectives, navigates to a certain
It is people Ai。
3.3, according to the weights K in relation L, by shortest path first (using dijkstra's algorithm here),
Obtain the communication approach of " most effective " between " I " and target, i.e. by the most intimate (or minimum) contact
People relates to the contact person needed most.
Here contact person node A is regarded as the point set in dijkstra's algorithm;Relation L regards as the limit collection in algorithm;
The a length of corresponding K value on limit;User node O is as the source in algorithm.
Step 4. completes mobile terminal synchronization
Its step is as follows:
4.1, service is started.PC server end requires to use the design service of the correlation technique such as Socket and multithreading
Device program, it can be resident service program, i.e. starts along with os starting, it is also possible to start-up by hand,
To set up with database server after startup and be connected, the then request of AM automatic monitoring mobile terminal.
4.2, IP is automatically configured.When user clicks on the sync client program of mobile terminal, it will be from mobile whole
Configuration file in end reads server ip address, server end slogan PORT;If PC cannot be connected
Server end, it is by automatic for intelligence scanning server end program.
Disposable IP scanning configuration strategy: after terminal adds the PC server place network segment, terminal program obtains automatically
Take its IP address, according to trizonal data before algorithm for text string location acquisition IP address, utilize circulation journey
Sequence scans the IP address being made up of 0-255 with first three area data automatically (as possible PC server end
Program IP address), if the merit of can be connected to, then it is stored in configuration file, for automatic Connection Service next time
Device uses.
4.3, system is set.The system setting function technical characteristic of mobile terminal is: when terminal program scans automatically
When cannot connect PC server, can interactively revise IP address, port numbers PORT, with
Time can also arrange backup and import time be completely covered, difference cover two options.
4.4, backup mobile terminal.The backup functionality technical characteristic of mobile terminal is: user's point on mobile terminals
When hitting backup functionality button, the option in system is arranged is for being completely covered, and program of mobile terminal reads this terminal
Address list in i-th contact person (i=1,2 ..., n), then by i-th contact person by PC server end journey
Sequence is written in the data base of correspondence, if data base exists this contact person, the most more new data, if not existing,
Then insert.If system arrange in option when being difference, program of mobile terminal reads in the address list of this terminal
I-th contact person (i=1,2 ..., n), it is right then i-th contact person to be written to by PC server
In the data base answered, if data base exists this contact person, then ignoring, if not existing, then inserting.
4.5, mobile terminal is imported.The import feature technical characteristic of mobile terminal is: user's point on mobile terminals
When hitting import feature button, the option in system is arranged is for being completely covered, and PC server reads data
I-th contact person in the address list of storehouse (i=1,2 ..., n), then i-th contact person is write by program of mobile terminal
Entering in the address list of terminal, if address list exists this contact person, the most more new data, if not existing, then inserting
Enter.If when the option during system is arranged is difference, in PC server reading database address list
I contact person (i=1,2 ..., n), then i-th contact person is written to the logical of terminal by program of mobile terminal
In news record, if address list exists this contact person, then ignoring, if not existing, then inserting.
Accompanying drawing explanation
Fig. 1 is business card typing flow chart.
When Fig. 2 is Density value, eight adjacent pixel location schematic diagrams.
Fig. 3 is the citing of Density value.
Fig. 4 is the density matrix that Fig. 3 changes.
Fig. 5 is business card image Processing Algorithm schematic diagram
Fig. 6 is card information word segmentation processing algorithm schematic diagram
Fig. 7 is a kind of social network structure schematic diagram centered by " I ".
Fig. 8 is the result of retrieval concrete manifestation effect on social networks.
Detailed description of the invention
The each process embodiment of the technical solution adopted in the present invention is as follows:
Step 1. typing card information
Utilize photographic head or scanner to carry out business card image collection, mark off character block according to the feature of business card image
And utilize OCR engine to carry out character recognition, carry out word segmentation processing according to critical field, the information extracted is returned
Class typing, is inserted corresponding list, finally interacts with user, believes None-identified that may be present
Breath carries out manual confirmation and adjustment.
Whole process exports from image to result, must input through image, image pre-treatment, character features extraction,
Matching identification, after word correction that will admit one's mistake through manual synchronizing, result is stored in data base, with reference to Fig. 1.
Described flow process should comprise following operation:
1.1, gray processing.First the business card image to photographic head or scanner collection uses weighted mean method to carry out ash
Degreeization.Owing to human eye is the highest to green sensitivity, minimum to blue-sensitive, therefore, use following formula can obtain relatively
Reasonably gray level image.
f(i,j)=0.30R(i,j)+0.59G(i,j)+0.11B(i,j)
Wherein (i, j) represent pixel coordinate, in coloured image three components of red, green, blue be respectively R (i, j), G (i, j),
(i, j), (i j) is then the gray value of this point to f to B.
1.2, edge extracting.In conjunction with Sobel operator and LOG (Laplacian of Gauss) operator, formed and revise
LOG algorithm, i.e. chooses following Sobel operator (SxDifference, S are divided in center for horizontal directionyFor vertical direction
Center divide difference, choose wherein higher value as gradient S)
As the precondition of rim detection, reduce unnecessary zero cross point.Recycling is if minor function is as filtering
Device, carries out LOG detection.
LOG operator template is as follows
LOG operator template therein is the digital form of LOG, is convolution, σ as interior collecting image
Mean square deviation for Gauss distribution.
1.3, binaryzation.Calculate threshold value, and gray-scale map is carried out binaryzation.If a height of h of gray-scale map, a width of w,
Threshold value is tried to achieve, i.e. average gray Threshold by following formula:
Binaryzation is carried out further according to threshold value Threshold:
1.4, tilt detection and rectification.Edge by Hough transformation (Hough Transform) detection business card frame
Line, it is thus achieved that name panel region also judges that the angle of inclination of business card is corrected.
Owing to the linear equation of y=kx+b form cannot represent that (c is constant, i.e. with x-axis for the straight line of x=c form
Parallel straight line, slope k → ∞).Therefore use parametric equation ρ=x*cos θ+y*sin θ here, wherein pass through
Choose p1(x1,y1), p2(x2,y2) two monitoring points, following formula can obtain tiltangleθ:
According to tiltangleθ, artwork and bianry image carried out simultaneously affine transformation (using homogeneous matrix to represent),
P'=P R, obtains correcting matrix by-θ substitution 2D spin matrix By point Obtain after rectification.Concrete operation is as follows:
x'=xcosθ+ysinθ
y'=-xsinθ+ycosθ
1.5, image segmentation.Feature according to business card image marks off character block.Comprise the following steps that
1.5.1, definition detection density Density, for current pixel up and down and on clinodiagonal totally 8 directions
Neighbor in the quantity of black picture element, with reference to Fig. 2, computing formula is as follows:
Wherein i ∈ [x-1, x+1] ∩ N*,j∈[x-1,x+1]∩N*, (x y) is current sensing point coordinate, N*For
Positive integer collection.
1.5.2, the image of name panel region is converted into density matrix, removes remaining noise.Operational approach is as follows:
Judging the density of each pixel one by one, as Density, < when 2, homography element is designated as 0, i.e. as noise
Process.As Density >=2 time, homography element is designated as 1, show this pixel be character block a part.As
In Fig. 3, pixel Density=0 of upper right mark;Pixel Density=2 of upper left mark, it is converted into close
Degree matrix is shown in Fig. 4.
1.5.3, pass through conversion formula Binary single color figure after correcting turns
Turn to following two-dimensional array form:
1.5.4, according to density matrix, character block region, location (this time has two kinds of strategies to differentiate to be applied to quickly know
Other pattern and accurate recognition mode), then according to region, business card image is split.
Quickly recognition strategy: judge density matrix line by line, the ratio that interior " 1 " element of every a line accounts for unit number exceedes
Certain threshold value, then be considered as line of text, be then considered as blank less than this threshold value.
Precisely recognition strategy: judge density matrix line by line, connects to " detection line " by interior for row continuous print " 0 " element,
Determine whether text filed according to " detection line " depth of initiating terminal, difference in length with terminal position feature.I.e.
Remove the length detection line less than threshold value;Then end and the Article 2 detection line of Article 1 detection line are marked
Top;After all row carry out detection and labelling, labelling the region surrounded is the most text filed.
1.5.5, definition Ri∈{d(i,y)|y∈[0,h]∩N*,And if only ifTime record j value.One group of region [j of last gained1,j2], [j2,j3], [j3,j4]…
The most some character blocks, wherein h represents that line number, w represent columns, N*For positive integer collection, RiFor step 3 two
The element of the i-th row, Sum in dimension groupjFor RiBe expert at all elements sum.
Owing to Chinese business card mostly is horizontal typesetting, it is in vertical distribution between character block, so place divides only with level
Cut character block.If if the business card running into complicated space of a whole page pattern needs to carry out vertical segmentation, can by same method again
Segment vertical character block.
1.6, character recognition.OCR (Optical Character Recognition) technology is utilized to extract each literary composition
The information of block.Here the OCR module pair of MODI (Microsoft Office Document Imaging) is used
The image being partitioned into is identified one by one, and each character block image is all become one group of word.
I.e. to character block [j1,j2], [j2,j3], [j3,j4] ... call OCR recognition engine one by one, obtain with each
Character set C that region is corresponding1,2,C2,3,C3,4…
Business card image processes and so far terminates, and Fig. 5 has sketched this processing procedure and algorithm used.
1.7, word segmentation processing.Word segmentation processing is carried out, to the information categorization typing extracted according to critical field.This needs
Keywords database to be set up and semantic base, such as: company, position, address, telephone number, email, road,
Number etc..Definition of keywords set W={ company, position, address, telephone number, email ... and definition of keywords
Semantic base, such as Address, Addr, address, contact address etc. are the near synonym of address, i.e. exist
It is semantically consistent, sets up corresponding mapping relations, form the semantic base with self-learning function.
Then according to key word, based on contextual feature, extract and often organize the corresponding informance in word, insert phase
Answer list, complete word segmentation processing.Specifically comprise the following steps that
1.7.1, find ": " separator, utilize separator to define key word and content.If the key defined
Word is not then transferred to the user decide whether that being regarded as key word is indexed in set in set of keywords.By this
Step realizes the self-learning function of participle strategy.
1.7.2, keyword Key=C is then matchedx,y∩ W, by character stringInsert entitled
In the form item of Key.WhenTime, by character set Cx,yInsert " unfiled " form item.
Strategy based on semantic base includes:
Location name strategy (Chinese): getting rid of other semantemes, number of characters, between 2~4, is followed by position, head
Rank, or with " name " class label;
Locating cellphone number strategy: with 11 pure digi-tal character strings of 13,15,18 beginnings, or with " hands
Machine " class label;
Location telephone number strategy: 7~8 pure digi-tal character strings, or start with area code, comprise bracket, loigature
The separators such as symbol, space, or with " phone " class label;
Location Business Name strategy: get rid of other semantemes, occurs that " company ", " group ", each institutional settings are crucial
Word, or with " company " class label;
Positioning address strategy: occur " province ", " city ", " county ", " township ", " town ", " road ", " district ", " building ",
The character string that " unit ", " room " mix with numeral, or with " address " class label;
Location postcode strategy: 6 pure digi-tal character strings, or with " postcode " class label;
Location mailbox strategy: numeral, letter, " ", ". " character occur, or with " mailbox " class label;
Location network address strategy: " http ", ". ", " www ", " com ", " cn ", " edu " character occur, or
With " network address " class label;
1.8 manual synchronizing.After participle completes, need to interact with user, to None-identified information that may be present
Carry out manual confirmation and adjustment, i.e. " unfiled " form item is carried out manual sort, after categorizing process terminates, will
Automatically study enters semantic base.
Card information word segmentation processing so far terminates, and Fig. 6 has sketched its processing procedure and logic.
Step 2. sets up social networks network
According to the six-point implicit scheme in sociology, set up a kind of social networks centered by " I ".
Being embodied in based on " I " is starting point O, and the form radiated to surrounding in arachnoid is shown in schematic diagram 7.
Using the familiarity between contact person and " I " as foundation, with the contact person Ai that " I " be directly related
Directly it is attached at " I " around by line, becomes the ground floor contact person the closest with " I ".With
This analogizes, by the contact person of each layer to the direction OAi divergence contrary with " I ".
In addition, the power of relation also can be determined by the number of degrees D of six topology degree.1 degree of relation i.e. " I " with
Relation between the friend that I is recognized, the friend that the friend of 2 degree of relations i.e. " I " and " I " is recognized it
Between relation, by that analogy, it is to be appreciated that: 1 degree of relation is certainly strong than 2 degree of relations.
Here, the concept introducing " cohesion " Intimacy expresses the familiarity of " I " and each contact person.It takes
The power of relation is certainly illustrated in weights K() and the number of degrees D of six topology degree.Here, using contact person as
Node, and provide visual interactive tool to be grouped contact person with opening relationships network.Set up this network
Step as follows:
2.1, the relation of two contact persons is set up by mobile contact person's node.The node A that will be newly addedjAdditional
To existing node AiOn (i.e. wanting the contact person of opening relationships), after determining operation both sides, utilize the form table of line
Show relation L set up between two contact personsi,j.Herein presented all relations such as Li,j, show AiWith AjThat
This recognizes mutually (two-way).
The most optional) if newly-established contact person node AiIt is the first contact people of user O (i.e. " I "),
Then by new node AiIt is connected on user node O set up contact circuit OAi, its relation is L0,i。
The most optional) if newly-established contact person node AjIt is that user O is by friend AiKnow this connection
It is people, and in the case of the other side does not recognize user O, then by new node AjIt is connected to user friend node Ai
Upper foundation contacts circuit AiAj, its relation is Li,j。
The most optional) if being originally user O by friend AiThe contact person A learntjWith user friend Ai
It is the first contact people (unidirectional), and after further cooperation and exchange, user O and this contact person AjBecome
One contact people (two-way), then by this contact person node AjIt is connected on user node O, is formed and user's node
Directly internuncial pathway OAj, its relation is L0,j。
2.2, for each relation L by using weights K to represent the power of relation.Li,jWeights Ki,jTake
Certainly in contact person AiWith contact person AjTime of getting to know T(unit: sky) and contact number of times Count.Specifically
Computational methods are as follows:
2.3, " cohesion " Intimacy of " I " and each contact person is drawn according to number of degrees D and weights K.Agreement is used
The number of degrees D of family node O0=0 passes through path OAiOn relation L0,iThe node A being connectediThe number of degrees
Di=1+D0;By path AiAjOn relation Li,jThe node A being connectedjNumber of degrees Dj=1+Di." I "
With node Aj" cohesion " computing formula be:
Intimacyj=Dj+Ki,j,Di∈N*,Ki,j∈[0,1)
The integer part of " cohesion " Intimacy represents the number of degrees in six degrees of separation, and fractional part is by the equal number of degrees
Under each relation Further Division open, wherein N*For positive integer collection.Therefore, " cohesion " can preferably react
Go out the close and distant relation to each node centered by " I ".
Step 3. realizes intelligent retrieval
By the social networks centered by " I ", intelligent retrieval most preferably contacts approach with location.
The result of retrieval by the performance of visual social network diagram is: by the contact person's node on path and line
It is highlighted, sees schematic diagram 8.When user selects highlighted node contact person, contact person's essential information is at node
Side manifests.Thus, the route that user can pass through to be highlighted be checked that user can check and can pass through which position
Friend can relate to required contact person.Described retrieval flow should realize:
● machine screens: fuzzy matching retrieval information, filters out possible object listing.
● it is positioned manually: from object listing, manuallys locate target.
● result is fed back: obtain the communication approach of " most effective " between " I " and target, i.e. by the most intimate
(or minimum) contact person relate to the contact person that needs most.
When being embodied as, the search key inputted by user carries out fuzzy matching with the personal information of typing,
Filtering out possible object listing, step is as follows:
3.1, the key word setting user's input is Char as character string WORD, element therein, business card data collection
DATA, ifMeetThen return the contact person of its correspondence
A(Char∩DATA)。
3.2, user is from the list returned, and according to the actual requirements, selects objectives, navigates to a certain
It is people Ai。
3.3, according to the weights K in relation L, by shortest path first (using dijkstra's algorithm here),
Obtain the communication approach of " most effective " between " I " and target, i.e. by the most intimate (or minimum) contact
People relates to the contact person needed most.
Contact person node A is regarded as the point set in dijkstra's algorithm;Relation L regards as the limit collection in algorithm;Limit
A length of corresponding K value;User node O is as the source in algorithm.
Step 4. completes mobile terminal synchronization
This step is by the card information realized in data base and mobile-terminal platform such as the mobile phones such as Android, iOS
Intelligent synchronization between address list.When being embodied as, following steps should be completed:
4.1, service is started.PC server end requires to use the design service of the correlation technique such as Socket and multithreading
Device program, it can be resident service program, i.e. starts along with os starting, it is also possible to start-up by hand,
To set up with database server after startup and be connected, the then request of AM automatic monitoring mobile terminal.
4.2, IP is automatically configured.When user clicks on the sync client program of mobile terminal, it will be from mobile whole
Configuration file in end reads server ip address, server end slogan PORT;If PC cannot be connected
Server end, it is by automatic scanning server end program.
Disposable IP scanning configuration strategy: after terminal adds the PC server place network segment, terminal program obtains automatically
Take its IP address, according to trizonal data before algorithm for text string location acquisition IP address, utilize circulation journey
Sequence scans the IP address being made up of 0-255 with first three area data automatically (as possible PC server end
Program IP address), if the merit of can be connected to, then it is stored in configuration file, for automatic Connection Service next time
Device uses.
4.3, system is set.The system setting function technical characteristic of mobile terminal is: when terminal program scans automatically
When cannot connect PC server, can interactively revise IP address, port numbers PORT, with
Time can also arrange backup and import time be completely covered, two options of difference.
4.4, backup mobile terminal.The backup functionality technical characteristic of mobile terminal is: user's point on mobile terminals
When hitting backup functionality button, the option in system is arranged is for being completely covered, and program of mobile terminal reads this terminal
Address list in i-th contact person (i=1,2 ..., n), then by i-th contact person by PC server end journey
Sequence is written in the data base of correspondence, if data base exists this contact person, the most more new data, if not existing,
Then insert.If when the option during system is arranged is difference, program of mobile terminal reads the address list of this terminal
Middle i-th contact person (i=1,2 ..., n), then i-th contact person is written to by PC server
In corresponding data base, if data base exists this contact person, then ignoring, if not existing, then inserting.
4.5, mobile terminal is imported.The import feature technical characteristic of mobile terminal is: user's point on mobile terminals
When hitting import feature button, the option in system is arranged is for being completely covered, and PC server reads data
I-th contact person in the address list of storehouse (i=1,2 ..., n), then i-th contact person is write by program of mobile terminal
Entering in the address list of terminal, if address list exists this contact person, the most more new data, if not existing, then inserting
Enter.If when the option during system is arranged is difference, in PC server reading database address list
I contact person (i=1,2 ..., n), then i-th contact person is written to the logical of terminal by program of mobile terminal
In news record, if address list exists this contact person, then ignoring, if not existing, then inserting.
Claims (1)
1. a social networks management method based on business card recognition, including following four step:
Step 1, typing card information, photographic head or scanner is utilized to carry out business card image collection, feature according to business card image marks off character block and utilizes OCR engine to carry out character recognition, word segmentation processing is carried out according to critical field, to the information categorization typing extracted, inserted corresponding list, finally interact with user, None-identified information that may be present is carried out manual confirmation and adjustment;
The described feature according to business card image divide character block and utilize OCR engine carry out character recognition specifically comprise the following steps that
1.1, gray processing: the first business card image to photographic head or scanner collection uses weighted mean method to carry out gray processing, owing to human eye is the highest to green sensitivity, minimum to blue-sensitive, therefore f (i is used, j)=0.30R (i, j)+0.59G (i, j) (i, j) can obtain more rational gray level image to+0.11B, wherein (i, j) represent pixel coordinate, in coloured image three components of red, green, blue be respectively R (i, j), G (i, j), B (i, j), (i j) is then the gray value of this point to f;
1.2, edge extracting: combine Sobel operator and LOG (Laplacian of Gauss) operator, is formed and revises LOG algorithm, i.e. chooses following Sobel operator: S=max{ | Sx|,|Sy|, as the precondition of rim detection, reduce unnecessary zero cross point, wherein SxDifference, S are divided in center for Sobel operator horizontal directionyDifference is divided at center for Sobel operator vertical direction, choose wherein higher value as gradient S;Re-using template isLOG operator detect;
1.3, binaryzation: calculate threshold value and gray-scale map is carried out binaryzation, if a height of h of gray-scale map, a width of w, tries to achieve threshold value, i.e. average grayBinaryzation is carried out further according to threshold value Threshold:
1.4, tilt detection and rectification: by the edge line of Hough transformation (Hough Transform) detection business card frame, obtain name panel region and judge that the angle of inclination of business card is corrected, particularly as follows: edge line is expressed as parametric equation ρ=x*cos θ+y*sin θ, then inclination angleWherein p1(x1,y1), p2(x2,y2) represent two monitoring points chosen on edge line, according to tiltangleθ, artwork and bianry image are carried out affine transformation, i.e. P'=P R simultaneously, obtain correcting matrix by-θ substitution 2D spin matrixBy pointObtaining after rectification, concrete operation isCoordinate points during wherein P represents artwork or bianry image (x, homogeneous coordinates y) represent, P' represent Slant Rectify after image in the homogeneous coordinates of coordinate points (x', y') represent;
1.5, image segmentation: mark off character block according to the feature of business card image, comprise the following steps that
1.5.1, defining detection density Density, for the quantity of black picture element in current pixel neighbor up and down and on clinodiagonal totally 8 directions, computing formula is:Wherein i ∈ [x-1, x+1] ∩ N*,j∈[x-1,x+1]∩N*, (x y) is current sensing point coordinate, N*For positive integer collection;
1.5.2, the image of name panel region is converted into density matrix, remove remaining noise, operational approach is as follows: judge the density of each pixel one by one, when Density is<when 2, homography element is designated as 0, i.e. as noise processed, as Density>=2 time, homography element is designated as 1, shows that this pixel is a character block part;
1.5.3, pass through conversion formulaBinary single color figure after correcting is converted into following two-dimensional array form:
1.5.4, according to density matrix, quick recognition strategy or character block region, accurate recognition strategy location can be respectively adopted, then according to region, business card image is split, wherein quick recognition strategy: judge density matrix line by line, the ratio that interior " 1 " element of every a line accounts for unit number exceedes certain threshold value, then it is considered as line of text, is then considered as blank less than this threshold value;Precisely recognition strategy: judge density matrix line by line, interior for row continuous print " 0 " element is connected to " detection line ", determine whether text filed according to " detection line " depth of initiating terminal, difference in length with terminal position feature, i.e. remove the length detection line less than threshold value;Then the end of Article 1 detection line and the top of Article 2 detection line are marked;After all row carry out detection and labelling, labelling the region surrounded is the most text filed;
1.5.5, definition Ri∈{d(i,y)|y∈[0,h]∩N*,And if only ifTime record j value, one group of region [j of last gained1,j2], [j2,j3], [j3,j4] ... the most some character blocks, wherein h represents that line number, w represent columns, N*For positive integer collection, RiFor the element of the i-th row, Sum in the two-dimensional array of step 3jFor RiBe expert at all elements sum;
1.6, character recognition: utilize OCR (Optical Character Recognition) technology to extract the information of each character block, the image being partitioned into is identified by OCR module one by one that use MODI (Microsoft Office Document Imaging) here, each character block image is all become one group of word, i.e. to character block [j1,j2], [j2,j3], [j3,j4] ... call OCR recognition engine one by one, obtain character set C corresponding with each region1,2,C2,3,C3,4…;
Described carry out specifically comprising the following steps that of word segmentation processing according to critical field
1.7, word segmentation processing: carry out word segmentation processing according to critical field, to the information categorization typing extracted, particularly as follows: initially set up keywords database W and semantic base, wherein key word refers to that those have the vocabulary determining describing significance for unit, position, address, contact method, and semantic base is the mapping library set up between the near synonym for the different expression of key word;According to key word, based on contextual feature, extract and often organize the corresponding informance in word, insert corresponding list, complete word segmentation processing, specifically comprise the following steps that
1.7.1, ": " separator is found, separator is utilized to define key word and content, if the key word defined is not in set of keywords, transfer to the user decide whether that being regarded as key word is indexed in set, realizes the learning functionality of participle strategy by this step;
1.7.2, keyword Key=C is then matchedx,y∩ W, by character stringInsert in the form item of entitled Key, whenTime, by character set Cx,yInsert " unfiled " form item;
Strategy based on semantic base includes:
Location Chinese Name strategy: use this rule to judge in the case of character string does not meets the semanteme of cell-phone number, telephone number, address, postcode, mailbox, network address: number of characters is between 2~4 and is followed by the character string of position, title, or the character string with " name " class label;
Locating cellphone number strategy: with 11 pure digi-tal character strings of 13,15,18 beginnings, or the character string with " mobile phone " class label;
Location telephone number strategy: 7~8 pure digi-tal character strings, or start and include bracket, hyphen or the character string of space-separated symbol, or the character string with " phone " class label with area code;
Location Business Name strategy: use this rule to judge in the case of character string does not meets the semanteme of cell-phone number, telephone number, address, postcode, mailbox, network address: " company ", " group ", the character string of each institutional settings key word, or the character string with " company " class label occur;
Positioning address strategy: the character string that " province ", " city ", " county ", " township ", " town ", " road ", " district ", " building ", " unit ", " room " mix with numeral, or the character string with " address " class label occur;
Location postcode strategy: 6 pure digi-tal character strings, or the character string with " postcode " class label;
Location mailbox strategy: numeral, letter, " ", ". " character, or the character string with " mailbox " class label occur;
Location network address strategy: " http ", ". ", " www ", " com ", " cn ", " edu " character, or the character string with " network address " class label occur;
Step 2, set up social networks network;
The described step setting up social networks network is as follows:
2.1, the relation of two contact persons is set up by moving contact person's node with mouse: the node A that will be newly addedjIt is attached to existing node AiOn, utilize the form of line to represent relation L set up between two contact persons after determining operation both sidesi,j, it represents AiWith AjRecognize each other, wherein AjAnd AiRepresent two contact persons, specifically process respectively by following several situations:
If the most newly-established contact person node AiIt is the first contact people of user O, then by new node AiIt is connected on user node O set up contact circuit OAi, its relation is L0,i;
If the most newly-established contact person node AjIt is that user O is by friend AiKnow this contact person, and in the case of the other side does not recognize user O, then by new node AjIt is connected to user friend node AiUpper foundation contacts circuit AiAj, its relation is Li,j;
If being 2.1.3 originally user O by friend AiThe contact person A learntjWith user friend AiIt is the first contact people, and after further cooperation and exchange, user O and this contact person AjBecome the first contact people, then by this contact person node AjIt is connected on user node O, is formed and user's node direct internuncial pathway OAj, its relation is L0,j;
2.2, for each relation Li,jBy using weights Ki,jRepresent the power of relation, Li,jWeights Ki,jDepend on contact person AiWith contact person AjTime of getting to know T (unit: sky) and contact number of times Count, circular is:
2.3, " cohesion " Intimacy of " I " and each contact person, the number of degrees D of agreement user node O is drawn according to number of degrees D and weights K0=0 passes through path OAiOn relation L0,iThe node A being connectediNumber of degrees Di=1+D0;By path AiAjOn relation Li,jThe node A being connectedjNumber of degrees Dj=1+Di, " I " and node Aj" cohesion " computing formula be: Intimacyj=Dj+Ki,j,Di∈N*,Ki,j∈ [0,1), the integer part of " cohesion " Intimacy represents the number of degrees in six degrees of separation, each relation Further Division under the equal number of degrees is opened by fractional part, therefore, " cohesion " can preferably reflect the close and distant relation centered by " I " to each node, wherein N*For positive integer collection;
Step 3, realize intelligent retrieval;
Described realize intelligent retrieval, it is characterised in that: the search key inputted by user carries out fuzzy matching with the personal information of typing, filters out possible object listing intelligent retrieval, and it specifically comprises the following steps that
3.1, the key word setting user's input is Char as character string WORD, element therein, business card data collection DATA, ifAndThen return the contact person A of its correspondence(Char ∩ DATA);
3.2, user is from the list returned, and according to the actual requirements, selects objectives, navigates to a certain contact person Ai;
3.3, according to the weights K in relation L, pass through dijkstra's algorithm, obtain the communication approach of " most effective " between " I " and target, i.e. related to the contact person needed most by the most intimate contact person, here contact person node A is regarded as the point set in dijkstra's algorithm;Relation L regards as the limit collection in algorithm;The a length of corresponding K value on limit;User node O is as the source in algorithm;
Step 4, complete mobile terminal synchronization;
The described step completing mobile terminal synchronization is as follows:
4.1, service is started: PC server end requires to use Socket and multithreading design server program, it can be resident service program, i.e. start along with os starting, can also start-up by hand, to set up with database server after startup and be connected, the then request of AM automatic monitoring mobile terminal;
4.2, automatically configuring IP: when user clicks on the sync client program of mobile terminal, it will read server ip address, server end slogan PORT in the configuration file from mobile terminal;If PC server end cannot be connected, it is by automatic scanning server end program, scanning configuration strategy is: after terminal adds the PC server place network segment, terminal program obtains its IP address automatically, according to trizonal data before algorithm for text string location acquisition IP address, cyclic program is utilized automatically to scan the IP address being made up of 0-255 and first three area data as possible PC server IP address, if the merit of can be connected to, then it is stored in configuration file, for automatic Connection Service device next time;
4.3, system is arranged: the system setting function technical characteristic of mobile terminal is: when terminal program automatically scans and cannot connect upper PC server, can interactively revise IP address, port numbers PORT, being completely covered when backup can also be set simultaneously and import, two options of difference;
4.4, backup mobile terminal: the backup functionality technical characteristic of mobile terminal is: when user clicks on backup functionality button on mobile terminals, option in system is arranged is for being completely covered, and program of mobile terminal reads i-th contact person in the address list of this terminal, i=1,2, ..., n, then i-th contact person is written in the data base of correspondence by PC server, if data base exists this contact person, the most more new data, if not existing, then inserts;If when the option during system is arranged is difference, program of mobile terminal reads i-th contact person in the address list of this terminal, i=1,2 ..., n, then i-th contact person is written in the data base of correspondence by PC server, if data base exists this contact person, then ignores, if not existing, then insert;
4.5, mobile terminal is imported: the import feature technical characteristic of mobile terminal is: when user clicks on import feature button on mobile terminals, option in system is arranged for being completely covered, i-th contact person in PC server reading database address list, i=1,2, ..., n, then i-th contact person is written in the address list of terminal by program of mobile terminal, if address list exists this contact person, the most more new data, if not existing, then inserts;If when the option during system is arranged is difference, i-th contact person in PC server reading database address list, i=1,2 ..., n, then i-th contact person is written in the address list of terminal by program of mobile terminal, if address list exists this contact person, then ignores, if not existing, then insert.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310521182.2A CN103617422B (en) | 2013-10-29 | 2013-10-29 | A social relation management method based on business card recognition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310521182.2A CN103617422B (en) | 2013-10-29 | 2013-10-29 | A social relation management method based on business card recognition |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103617422A CN103617422A (en) | 2014-03-05 |
CN103617422B true CN103617422B (en) | 2017-01-11 |
Family
ID=50168125
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310521182.2A Active CN103617422B (en) | 2013-10-29 | 2013-10-29 | A social relation management method based on business card recognition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103617422B (en) |
Families Citing this family (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104020847A (en) * | 2014-05-15 | 2014-09-03 | 上海合合信息科技发展有限公司 | Character input method and device |
CN105631486A (en) * | 2014-10-27 | 2016-06-01 | 深圳Tcl数字技术有限公司 | Method and device for recognizing images and characters |
CN104504069A (en) * | 2014-12-22 | 2015-04-08 | 北京奇虎科技有限公司 | Building method and device for file index |
CN105490811A (en) * | 2015-11-25 | 2016-04-13 | 小米科技有限责任公司 | Method and device for obtaining signature information |
CN105718972B (en) * | 2016-01-15 | 2019-01-29 | 卢怀农 | A kind of information intelligent acquisition method |
CN105631752A (en) * | 2016-03-22 | 2016-06-01 | 南京信息工程大学 | Social network sampling generation algorithm for improving Dijkstra weight |
CN107239453B (en) * | 2016-03-28 | 2020-10-02 | 平安科技(深圳)有限公司 | Information writing method and device |
CN105791503A (en) * | 2016-04-28 | 2016-07-20 | 乐视控股(北京)有限公司 | Method of storing business card information in address list and apparatus thereof |
CN107402927A (en) * | 2016-05-19 | 2017-11-28 | 上海斯睿德信息技术有限公司 | A kind of enterprise's incidence relation topology method for building up and querying method based on graph model |
CN106326854B (en) * | 2016-08-19 | 2019-09-06 | 掌阅科技股份有限公司 | A kind of format document paragraph recognition methods |
CN106230968A (en) * | 2016-08-22 | 2016-12-14 | 杨玲 | Cross-platform file based on Apache Server transmission APP software system |
CN108009546B (en) * | 2016-10-28 | 2021-08-10 | 北京京东振世信息技术有限公司 | Information identification method and device |
CN106709488A (en) * | 2016-12-20 | 2017-05-24 | 深圳市深信服电子科技有限公司 | Business card identification method and device |
CN107679240B (en) * | 2017-10-27 | 2020-11-10 | 中国计量大学 | Virtual identity mining method |
CN108446697B (en) * | 2018-03-06 | 2019-11-12 | 平安科技(深圳)有限公司 | Image processing method, electronic device and storage medium |
CN108509783B (en) * | 2018-04-09 | 2020-06-30 | 托普朗宁(北京)教育科技有限公司 | Intelligent terminal application program management system |
CN109358067A (en) * | 2018-11-30 | 2019-02-19 | 桂林电子科技大学 | Motor ring varistor defect detecting system based on computer vision and method |
CN111147361B (en) * | 2019-12-30 | 2022-06-07 | 论客科技(广州)有限公司 | Method, device and storage medium for adding mailbox account |
CN111414536A (en) * | 2020-03-17 | 2020-07-14 | 支付宝(杭州)信息技术有限公司 | Service processing method and device |
CN112199545B (en) * | 2020-11-23 | 2021-09-07 | 湖南蚁坊软件股份有限公司 | Keyword display method and device based on picture character positioning and storage medium |
CN112733834B (en) * | 2021-03-30 | 2021-07-06 | 恒银金融科技股份有限公司 | Character area positioning method based on non-matrix window mode |
CN113569738A (en) * | 2021-07-28 | 2021-10-29 | 深圳思谋信息科技有限公司 | Tabulation information extraction processing method and device, computer equipment and storage medium |
CN115170040B (en) * | 2022-09-08 | 2022-12-20 | 南方电网数字电网研究院有限公司 | Method and system for dynamically updating asset directory |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101882264A (en) * | 2009-11-27 | 2010-11-10 | 镇立新 | Method for adding contacts |
CN101916271A (en) * | 2010-08-10 | 2010-12-15 | 上海合合信息科技发展有限公司 | Method and system for automatically acquiring contact head image |
CN102622592A (en) * | 2012-01-19 | 2012-08-01 | 北京友录在线科技发展有限公司 | Name card recognition method based on cloud technology |
-
2013
- 2013-10-29 CN CN201310521182.2A patent/CN103617422B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101882264A (en) * | 2009-11-27 | 2010-11-10 | 镇立新 | Method for adding contacts |
CN101916271A (en) * | 2010-08-10 | 2010-12-15 | 上海合合信息科技发展有限公司 | Method and system for automatically acquiring contact head image |
CN102622592A (en) * | 2012-01-19 | 2012-08-01 | 北京友录在线科技发展有限公司 | Name card recognition method based on cloud technology |
Non-Patent Citations (2)
Title |
---|
中文商务名片识别系统的实现;张纯等;《中文信息学报》;20000420;第14卷(第02期);第21-25页 * |
到底谁更实用 经纬名片通VS名片全能王;牛华网;《http://www.newhua.com/2013/0105/191386.shtml》;20130105;第11、12、14页 * |
Also Published As
Publication number | Publication date |
---|---|
CN103617422A (en) | 2014-03-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103617422B (en) | A social relation management method based on business card recognition | |
CN109993160B (en) | Image correction and text and position identification method and system | |
RU2668717C1 (en) | Generation of marking of document images for training sample | |
US9171013B2 (en) | System and method for providing objectified image renderings using recognition information from images | |
US8649572B2 (en) | System and method for enabling the use of captured images through recognition | |
US7809192B2 (en) | System and method for recognizing objects from images and identifying relevancy amongst images and information | |
US7809722B2 (en) | System and method for enabling search and retrieval from image files based on recognized information | |
CN109657665A (en) | A kind of invoice batch automatic recognition system based on deep learning | |
US8787673B2 (en) | System and method of determining building numbers | |
CN108875744B (en) | Multi-direction text line detection method based on rectangular frame coordinate transformation | |
CN104820817A (en) | Four-dimensional code, image identification system and method as well as retrieval system and method based on four-dimensional code | |
CN103577818A (en) | Method and device for recognizing image characters | |
CN115994230A (en) | Intelligent archive construction method integrating artificial intelligence and knowledge graph technology | |
CN110580302B (en) | Sketch image retrieval method based on semi-heterogeneous joint embedded network | |
CN102779140A (en) | Keyword acquiring method and device | |
WO2006122164A2 (en) | System and method for enabling the use of captured images through recognition | |
CN104142955A (en) | Method and terminal for recommending learning courses | |
CN109948598B (en) | Document layout intelligent analysis method and device | |
CN113901053A (en) | Teaching material index management system based on big data | |
WO2016069005A1 (en) | Text line detection | |
Choi et al. | Face annotation for personal photos using collaborative face recognition in online social networks | |
CN111582264B (en) | Method, device and system for accurate frame questions, electronic equipment and storage medium | |
CN110647646B (en) | Picture file batch reading method and device | |
Khlif | Multi-lingual scene text detection based on convolutional neural networks | |
CN117671619A (en) | Traffic scene image matching method and system based on deep learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |