CN108038496A - Love and marriage object matching data processing method, device, computer equipment and storage medium based on big data and deep learning - Google Patents
Love and marriage object matching data processing method, device, computer equipment and storage medium based on big data and deep learning Download PDFInfo
- Publication number
- CN108038496A CN108038496A CN201711260402.5A CN201711260402A CN108038496A CN 108038496 A CN108038496 A CN 108038496A CN 201711260402 A CN201711260402 A CN 201711260402A CN 108038496 A CN108038496 A CN 108038496A
- Authority
- CN
- China
- Prior art keywords
- matching
- duration
- deep learning
- married
- preset
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Evolutionary Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of love and marriage object matching data processing method based on big data and deep learning, including:Obtain the first object image data and the second object image data to be matched;By the trained deep learning neutral net of the first object image data and the input of the second object image data, the parameter of deep learning neutral net is to be used as input using the corresponding matching image big data of married matching object, the corresponding duration that is effectively matched of married matching object is trained to obtain with the corresponding label that the preset matching duration of deep learning neural network training model relatively obtains as anticipated output, input deep learning neural network training model;Obtain the corresponding matching result of trained deep learning neutral net output;The matching status of first object and the second object in the range of preset matching duration is determined according to matching result, it is also proposed that a kind of love and marriage object matching data processing equipment, computer equipment and readable storage medium storing program for executing, there is provided objectively match foundation.
Description
Technical field
The present invention relates to field of computer technology, more particularly to a kind of love and marriage pair based on big data and deep learning
As matched data processing method, device, computer equipment and storage medium.
Background technology
Big data, which refers to, needs new tupe to have stronger decision edge, see clearly discovery power and process optimization ability
Magnanimity, high growth rate and diversified information assets.Finding in daily life, man and wife often has man and wife's phase, because
The appearance of people be with gene-correlation, and each side such as gene and personality of people is related.
Love and marriage object is recommended in existing agency of matchmakers or website to user at random or according to the experience of matchmaker to user, does not have
Objective basis, it is impossible to predict that object to be recommended matches matching status of the object in future time scope with request.
The content of the invention
Based on this, it is necessary to for above-mentioned technical problem, there is provided a kind of love and marriage object matching data processing method, device,
Computer equipment and readable storage medium storing program for executing, based on big data and deep learning neural network object to be matched and default
Correlation with duration, there is provided objectively match foundation, can predict that object to be recommended matches object in future time model with request
The matching status enclosed.
A kind of love and marriage object matching data processing method, the described method includes:
Obtain the first object image data and the second object image data to be matched;
By the trained deep learning nerve net of first object image data and the input of the second object image data
Network, the parameter of the deep learning neutral net is using the corresponding matching image big data of married matching object as input, is incited somebody to action
The corresponding preset matching time length ratio for being effectively matched duration and deep learning neural network training model of the married matching object
The corresponding label relatively obtained trains to obtain as anticipated output, input deep learning neural network training model;
The corresponding matching result of deep learning neutral net trained output described in obtaining;
Of first object and the second object in the range of the preset matching duration is determined according to the matching result
With state.
In one of the embodiments, it is described to obtain the first object image data to be matched and the second object images number
According to the step of before, further include:
Training image big data is obtained, the training image big data includes the corresponding matching image of married matching object
Big data, from it is described matching image big data in extract it is corresponding it is each it is married matching object it is corresponding matching image pair;
Obtain that married matching object is corresponding to be effectively matched duration, obtain the pre- of deep learning neural network training model
If match duration;
Corresponding matching is schemed according to the corresponding relation for being effectively matched duration and preset matching duration of married matching object
As to being included into corresponding goal set, different goal sets corresponds to different labels;
Using each matching image to the input as the deep learning neutral net, to the deep learning nerve net
Network carries out unsupervised training;
The object matching image pair label corresponding with goal set obtained successively in each goal set forms matching pass
System, successively will be provided with the input as the deep learning neutral net object matching image in each goal set
Anticipated output of the label of matching relationship as the deep learning neutral net, carries out the deep learning neutral net
Training;
Obtain the trained deep learning neutral net.
In one of the embodiments, it is described to obtain married matching object corresponding the step of being effectively matched duration and include:
Obtain the matching initial time and current matching state of married matching object;
If current matching state is non-matching, obtains matching and terminate the time, the time is terminated with having matched according to matching
Time beginning obtains being effectively matched duration;
If current matching state is matching, married matching object both sides' state is obtained, if married matching object is double
Square state is normal, then obtains current time, obtains being effectively matched duration scope with matching initial time according to current time;
If at least one party's abnormal state in married matching object both sides, when will be effectively matched duration and be determined as default
It is long.
It is in one of the embodiments, described that according to married matching, object is corresponding is effectively matched duration and preset matching
Corresponding matching image is corresponded to different labels by the relation of duration to being included into corresponding goal set, different goal sets
The step of include:
If being effectively matched duration is greater than or equal to preset matching duration, by married corresponding of the object of matching
With image the first label is corresponded to being included into first set, the first set;
If being effectively matched duration is less than preset matching duration, by the corresponding matching image of the married matching object
To being included into second set, the second set corresponds to the second label;
If the relation for being effectively matched duration and preset matching duration does not determine, the married matching object is corresponded to
Matching image to being classified as invalid data.
In one of the embodiments, it is described to determine the first object with the second object described according to the matching result
The step of matching status in the range of preset matching duration, includes:
If the matching result is the first label, it is determined that first object and the second object are at described default
With being successful match state in the range of duration;
If the matching result is the second label, it is determined that first object and the second object are at described default
With being it fails to match state in the range of duration.
In one of the embodiments, it is described to obtain the first object image data to be matched and the second object images number
According to the step of include:
The matching request that request matching object user terminal is sent is received, the matching request includes request matching object
View data, matches object image data as first object image data using the request, obtains object diagram to be recommended
As data are as second object image data;
It is described to determine the first object and the second object in the range of the preset matching duration according to the matching result
Matching status the step of after, further include:
If request matching object is with matching status of the object to be recommended in the range of the preset matching duration
With success status, then the corresponding user information of object to be recommended is sent to the request matching object user terminal.
A kind of love and marriage object matching data processing equipment, described device include:
Acquisition module, for obtaining the first object image data and the second object image data to be matched;
Matching result acquisition module, for first object image data and the second object image data to be inputted
Trained deep learning neutral net, the parameter of the deep learning neutral net is with the corresponding matching of married matching object
Image big data instructs the corresponding duration that is effectively matched of the married matching object with deep learning neutral net as input
The corresponding label that the preset matching duration of white silk model relatively obtains is as anticipated output, input deep learning neutral net instruction
Practice model training to obtain, the corresponding matching result of deep learning neutral net trained output described in acquisition;
Matching status determining module, for determining the first object with the second object described pre- according to the matching result
If match the matching status in the range of duration.
In one of the embodiments, the acquisition module is additionally operable to receive what request matching object user terminal was sent
Matching request, the matching request include request matching object image data, using it is described request match object image data as
First object image data, obtains object image data to be recommended as second object image data;
Described device further includes:
Sending module, if for asking matching object with object to be recommended in the range of the preset matching duration
Matching status be successful match state, then send the corresponding use of object to be recommended to the request matching object user terminal
Family information.
A kind of computer equipment, including memory and processor, the memory storage have computer program, the meter
When calculation machine program is performed by the processor so that the processor performs the step of any of the above-described embodiment the method.
A kind of computer-readable recording medium, is stored with computer program, and the computer program is executed by processor
When so that the processor performs the step of any of the above-described embodiment the method.
Above-mentioned love and marriage object matching data processing method, device, computer equipment and readable storage medium storing program for executing, pass through acquisition
First object image data and the second object image data to be matched, by the first object image data and the second object images
The trained deep learning neutral net of data input, the parameter of deep learning neutral net are corresponded to married matching object
Matching image big data as input, by married matching, object is corresponding is effectively matched duration and deep learning neutral net
The corresponding label that the preset matching duration of training pattern relatively obtains inputs deep learning neutral net as anticipated output
Training pattern trains to obtain, and obtains the corresponding matching result of trained deep learning neutral net output, is tied according to matching
Fruit determines the matching status of the first object and the second object in the range of the preset matching duration, based on big data and depth
Degree learning neural network establishes the correlation of object to be matched and preset matching duration, there is provided objectively matches foundation, can predict
Object to be recommended matches matching status of the object in future time scope with request.
Brief description of the drawings
Fig. 1 is the applied environment figure of love and marriage object matching data processing method in one embodiment;
Fig. 2 is the cut-away view of one embodiment Computer equipment;
Fig. 3 is the flow chart of love and marriage object matching data processing method in one embodiment;
Fig. 4 is the flow chart of initial configuration in one embodiment;
Fig. 5 is the flow chart that one embodiment is trained deep learning neutral net;
Fig. 6 is the structure diagram of love and marriage object matching data processing equipment in one embodiment;
Fig. 7 is the structure diagram of love and marriage object matching data processing equipment in another embodiment;
Fig. 8 is the structure diagram of love and marriage object matching data processing equipment in further embodiment;
Fig. 9 is the structure diagram of love and marriage object matching data processing equipment in another embodiment.
Embodiment
As shown in Figure 1, being the applied environment figure in one embodiment, which includes terminal 1001, server
1002, it can wherein be communicated between terminal 1001, server 1002 by network.Server 1002 can be multiple things
The server cluster that server is formed is managed, the basic cloud computing such as Cloud Server, cloud database, cloud storage and CDN can be to provide
The Cloud Server of service.Terminal 1001 can be smart mobile phone, tablet computer, laptop, desktop computer, intelligent watch
Deng, but be not limited thereto.Terminal 1001 can send matching request to server 1002, and matching request carries request matching pair
As view data, will request matching object image data as the first object image data, server 1002 will successively will be each
Candidate's object image data to be recommended calculates each candidate object to be recommended and request successively as the second object image data
Matching status of the object in the range of preset matching duration is matched, and according to matching status, by candidate's object information to be recommended
Recommend the corresponding terminal of request matching object.Server 1002 can also store request matching object image data, work as presence
During the candidate of renewal object to be recommended, the candidate's object to be recommended for calculating renewal matches object in preset matching duration with request
In the range of matching status, actively to the corresponding terminal push of request matching object in the range of preset matching duration
With successful candidate object information to be recommended.The first object image data and the second object can also be obtained in terminal 1001
View data, inputs in terminal 1001 in the range of trained deep learning neutral net obtains preset matching duration
Matching status.
As shown in Fig. 2, the internal structure schematic diagram for one embodiment Computer equipment.The computer equipment includes
Processor, non-volatile memory medium, built-in storage and the network interface connected by system bus.Wherein, which sets
Standby non-volatile memory medium can storage program area and computer program, which is performed, and may be such that place
Manage device and perform a kind of love and marriage object matching data processing method.The processor of the computer equipment, which is used to provide, to be calculated and controls
Ability, supports the operation of whole computer equipment.Computer program can be stored in the built-in storage, the computer program quilt
When processor performs, it may be such that processor performs a kind of love and marriage object matching data processing method.The network of computer equipment connects
Mouth is used to carry out network service.It will be understood by those skilled in the art that structure shown in Figure 2, it is only and application scheme
The block diagram of relevant part-structure, does not form the restriction for the computer equipment being applied thereon to application scheme, tool
The computer equipment of body can include, than more or fewer components shown in figure, either combining some components or with not
Same component arrangement.
As shown in figure 3, in one embodiment it is proposed that a kind of love and marriage object matching data processing method, this method should
For the server or terminal in above application environment, following steps are specifically included:
Step S210, obtains the first object image data and the second object image data to be matched.
Wherein, the first object image data, the second object image data are to characterize the first object and the second object spy respectively
The data of sign, can be the picture of shooting or carry out the image feature data after image characteristics extraction to picture, its
In the first object and the second object be respectively to ask matching object and object to be recommended.First object image data, the second object
The first object images number that view data can be carried in the matching request that terminal is sent or obtained from server
According to, the second object image data.In one embodiment, the second object image data is the registered user from server storage
The view data of corresponding image information extraction, with registered user's information updating.In one embodiment, the first object diagram
As data and the second object image data are to be obtained respectively from the first object images and second pair as extracting human face region in image
's.
Step S220, by the trained deep learning god of the first object image data and the input of the second object image data
Through network, the parameter of deep learning neutral net is using the corresponding matching image big data of married matching object as input, is incited somebody to action
The corresponding duration that is effectively matched of married matching object is compared with the preset matching duration of deep learning neural network training model
The corresponding label arrived trains to obtain as anticipated output, input deep learning neural network training model.
Wherein, deep learning neutral net is to simulate the neutral net that human brain carries out analytic learning, imitates the mechanism of human brain
Carry out interpretation of images data, the first step of deep learning is substantially a network parameter initialization procedure, is different from traditional neural
Network initial value random initializtion, deep learning neutral net are obtained by the structure of unsupervised trained input data, thus
This initial value is closer to global optimum, so as to obtain more preferable effect.In one embodiment, by convolutional neural networks
As this deep learning neutral net, convolutional neural networks (Constitutional Neural Networks, CNN) be
The specially designed a kind of deep learning side for image classification and identification to grow up on the basis of multilayer neural network
Method.
The wherein parameter of trained deep learning neutral net, is big with the corresponding matching image of married matching object
Data are effectively matched the pre- of duration and deep learning neural network training model as input, by married matching object is corresponding
If the corresponding label that matching duration relatively obtains is as anticipated output, the neural network training model training of input deep learning
Obtain.Trained deep learning neutral net includes input layer, multiple hidden layers and output layer, the wherein number of plies, every layer of default section
Points, the weights of each network connection are obtained by training.
Wherein married matching object refers to marriage registration object in statutory force, obtains sufficiently large married of data volume
The corresponding matching image big data of matching object is trained as the training data of deep learning neutral net, obtains depth
The parameter of learning neural network.It is effectively matched duration and refers to matching duration in statutory force, can be double according to married matching object
Matching status in the life state and statutory force of side is determined.Default of deep learning neural network training model
It can customize as needed with duration, be such as defined as all one's life or 30 years.It is effectively matched duration and deep learning nerve net
The preset matching duration of network training pattern relatively obtains different comparative results, and different comparative results corresponds to different labels,
The corresponding matched label of input data is as anticipated output when obtaining training, so as to be trained.Wherein, label is used to retouch
State the matching degree in the range of preset matching duration, matching degree can be matching probability, matching whether successfully etc..
In one embodiment, matching the corresponding matching image big data of object includes the people of married matching object both sides
Face picture, can also include both sides region, marriage registration region, age information formed various dimensions input training data into
Row training.The corresponding different trained deep learning neutral net of different marriage registration areas generation can be directed to, is improved not
With the fitness and accuracy of area object matching.
Since the marriage registration object in statutory force represents successful match, according to the image information of successful match object,
Form objective scientific basis big data and carry out depth network analysis, improve the objective basis of matching object.
Step S230, obtains the corresponding matching result of trained deep learning neutral net output.
Specifically, the matching result form of deep learning neutral net output is unlimited, can be matching probability, can also
It is the matching label of mutual exclusion, such as represents the first object and the second object successful match in the range of preset matching duration with 1,
Representing the first object and the second object with 0, it fails to match in the range of preset matching duration.Trained deep learning nerve
Network can export corresponding objective matching result according to the analysis of image data of the object to be matched of input.
Step S240, the scope of the first object and the second object in the preset matching duration is determined according to matching result
Interior matching status.
Specifically, if matching result is matching probability, the first object and can be determined according to the size of matching probability
Matching probability of two objects in the range of preset matching duration.If matching result is the matching label of mutual exclusion, direct root
The matching status of corresponding first object and the second object in the range of preset matching duration is worth to according to matching label,
Such as successful match or it fails to match.Can be by the range of the first object and multiple objects progress preset matching durations to be matched
The calculating of matching status, obtains the object of maximum matching probability as object matching object.
In the present embodiment, by obtaining the first object image data and the second object image data to be matched, by first
The deep learning neutral net that object image data and the input of the second object image data have been trained, deep learning neutral net
Parameter be using the corresponding matching image big data of married matching object as input, married matching object is corresponding effective
Matching duration is with the corresponding label that the preset matching duration of deep learning neural network training model relatively obtains as expection
Output, input deep learning neural network training model train to obtain, and obtain trained deep learning neutral net output
Corresponding matching result, determines the first object and the second object in the range of the preset matching duration according to matching result
Matching status, the correlation based on big data and deep learning neural network object to be matched and preset matching duration,
Objectively matching foundation is provided, can predict that object to be recommended matches matching status of the object in future time scope with request.
In one embodiment, as shown in figure 4, before step S210, further include:
Step S310, obtains training image big data, and training image big data includes the corresponding matching of married matching object
Image big data, from matching image big data in extract it is corresponding it is each it is married matching object it is corresponding matching image pair.
Wherein, training image big data refers to the mass data for training deep learning neutral net, wherein married
Include multiple and different regional corresponding view data of married matching object both sides with the corresponding matching image big data of object,
Such as the picture of married object both sides.It is understood that the training image that corresponding area is respectively adopted in different regions can be directed to
Big data trains to obtain deep learning neutral net corresponding with area, because different areas is due to region, population, custom
There is also difference on object matching degree, form different deep learning neutral nets for different regions has the difference of culture
Beneficial to the matching objectivity and accuracy rate for improving locality.
In addition, matching image to referring to the married image for matching object both sides, can be facial image, whole body images etc., root
According to the image range for needing self-defined needs.Face can be therefrom extracted by face recognition algorithms and obtains each married matching pair
As corresponding matching image pair.Such as married matching object includes A, B, C, D, then matches image to being respectively { A1,A2, { B1,
B2, { C1,C2, { D1,D2}。
In one embodiment, it is marriage certificate data to match image big data, according to sex character and people from marriage certificate
Face feature carries out image recognition, extracts husband's photo and wife's photo forms the corresponding matching image of each married matching object
It is right;The predeterminated position of husband's photo and wife's photo in marriage certificate can also be utilized, according to position range to husband's photo
Directly extracted with wife's photo, because marriage certificate is all standard, fixed position scope of husband's photo in marriage certificate
It is interior, wife's photo also in the range of another fixed position of marriage certificate, can rapid extraction obtain matching image pair, improve data
The efficiency of preparation.
Step S320, the married matching object of acquisition is corresponding to be effectively matched duration, obtains deep learning neural metwork training
The preset matching duration of model.
Specifically, the corresponding duration that is effectively matched of married matching object can be from the legal day in legal matching status file
Phase determines that wherein statutory date includes matching from date, matching date of expiry etc..Go out for both sides in married matching object
Existing abnormal conditions, it can determine to be effectively matched duration exist in such as married matching object dead, missing according to abnormal conditions
Situation, it can determine to be effectively matched duration according to the time that abnormal conditions occur.Deep learning neural network training model it is pre-
If matching duration is determined that in training deep learning neural network training model, can train to obtain corresponding different default
Match the deep learning neural network training model of duration.The deep learning god of selection target preset matching duration as needed
Through network training model.
Step S330, will be right according to the corresponding relation for being effectively matched duration and preset matching duration of married matching object
The matching image answered corresponds to different labels to being included into corresponding goal set, different goal sets.
Specifically, the default function calculated relationship for including 2 variables can be obtained, duration and default will be effectively matched
Substitute into, be calculated as a result, according to result by corresponding matching image to returning respectively as the corresponding value of 2 variables with duration
Enter corresponding goal set, wherein different goal sets corresponds to different labels.In one embodiment, multiple areas are preset
Between section, different segments corresponds to different goal sets, there is corresponding label, determine to be effectively matched duration and preset matching
Segment where the difference of duration, so that by corresponding matching image to the corresponding goal set of segment where being included into.
Step S340, using each matching image to the input as the deep learning neutral net, to deep learning god
Unsupervised training is carried out through network.
Specifically, by each matching image pair, such as (A1,A2), (B1,B2), (C1,C2), (D1,D2) depth is inputted respectively
Neutral net is practised, carries out unsupervised training.Using unsupervised training from bottom to top, monolayer neuronal member is successively built, every layer is adopted
Tuning is carried out with wake-sleep algorithms.One layer is only adjusted every time, is successively adjusted, this process can be regarded as one
The process of feature learning, is to distinguish larger part with traditional neural network.Wherein wake-sleep algorithms are divided into
Wake stages and sleep stages, wherein wake stages are cognitive processes, by the input feature vector (Input) of lower floor and upward
Cognition (Encoder) weight produces each layer of abstract representation (Code), then is produced by current generation (Decoder) weight
A raw reconstruction information (Reconstruction), calculates input feature vector and reconstruction information residual error, declines modifying layer using gradient
Between downlink generation (Decoder) weight, that is, " if reality with I imagine it is different, change my generation weight
So that the thing that I imagines becomes as reality ".The sleep stages are generating process, by Upper Concept (Code) and downwards
Generation (Decoder) weight, generate the state of lower floor, recycle cognition (Encoder) weight to produce an abstract scene.
Using initial upper layer concept and the residual error of newly-built abstract scene, decline the upward cognition (Encoder) of modification interlayer using gradient
Weight.Namely " if the scene in dream is not the corresponding concepts in my brain, the cognition weight for changing me causes this scene
It is exactly this concept in my view ".
Step S350, obtains the object matching image pair label corresponding with goal set in each goal set successively
Matching relationship is formed, successively using the object matching image in each goal set to the input as deep learning neutral net,
Will be provided with anticipated output of the label of matching relationship as deep learning neutral net, to the deep learning neutral net into
Row Training.
Specifically, the label for possessing matching relationship is object matching image in each goal set to corresponding label,
Represent matching status of the corresponding object of two photos of matching image pair in the range of preset matching duration.It is such as each
Goal set is expressed as Pi, wherein 1≤i≤N, N represent the total quantity of goal set.The corresponding label point of each goal set
Wei not Mi, obtain PiObject matching image in set is to the input as deep learning neutral net, by corresponding MiAs
The anticipated output of deep learning neutral net, carries out Training.
Top-down supervised training is carried out, this step is to obtain each layer parameter on the basis of in first step study,
The coding layer most pushed up adds a grader, such as the special recurrence of Rogers, SVM etc., is then instructed by the supervision of tape label data
Practice, go to finely tune whole network parameter using gradient descent method.The first step of deep learning is substantially that a network parameter is initial
Change process, is different from traditional neural network initial value random initializtion, and deep learning neutral net is inputted by unsupervised training
What the structure of data obtained, thus this initial value is closer to global optimum, so as to obtain more preferable effect.
Step S360, the deep learning neutral net trained.
Specifically, after unsupervised training above and Training, the deep learning nerve net trained
Network.
In one embodiment, as shown in figure 5, obtaining married matching in step S320, object is corresponding when being effectively matched
Length includes:
Step S321, obtains the matching initial time and current matching state of married matching object.Judge current matching shape
Whether state matches, and enters step S322 if non-matching, otherwise enters step S323.
Specifically, the effective time matching initial time is the matching determined by legal matching files, current matching
State be the married matching object both sides determined according to legal matching files in the corresponding matching status of current time, including
With state and non-matched state.As obtained matching initial time including wedding date on marriage certificate, if in current time still
Marriage state, then current matching state is matching, if being in divorce state in current time, current matching state is non-
Match somebody with somebody.
Step S322, obtains matching and terminates the time, is effectively matched according to the matching termination time with matching initial time
Duration.
Specifically, it is the effective time that the matching determined by legal matching files terminates, such as current that matching, which terminates the time,
It is non-matching with state, then the acquisition divorce date obtains matching and terminates the time from divorce certificate.Will divorce date and wedding date
Subtract each other, obtain being effectively matched duration.
Step S323, obtains married matching object both sides' state, judges whether married matching object both sides state is normal,
If both sides' state is normal, current time is obtained, obtains being effectively matched duration model with matching initial time according to current time
Enclose, otherwise, enter step S324.
Specifically, married matching object both sides state is just referring to situation of the married matching object both sides there is no the death of one's spouse,
And matching status is still within, the current time date is subtracted matching initial time obtains current matching duration, when being effectively matched
Long scope refers to that being effectively matched duration is more than current matching duration.Such as the date 19970920 of marriage, do not divorce, wedding, both sides
Normally, current date 2017-9-20, was effectively matched duration more than 20 years.
Step S324, if at least one party's abnormal state in married matching object both sides, will be effectively matched duration and determine
For preset duration.
Specifically, if it is married matching object both sides in there is a situation where be bereft of one's spouse, will be effectively matched duration be determined as it is pre-
If duration, wherein preset duration can be self-defined as needed, is such as defined as all one's life, or be determined according to the time of the death of one's spouse.
In the present embodiment, corresponding be effectively matched is determined by current matching state and married matching object both sides' state
Duration, can flexibly and accurately determine to be effectively matched duration.
In one embodiment, as shown in fig. 6, step S330 includes:
Step S331, if being effectively matched duration is greater than or equal to preset matching duration, married matching object is corresponded to
Matching image to being included into first set, the first set corresponds to the first label.
Specifically, the matching image in identity set is to identical matching characteristic, the wherein matching in first set
Image is to be effectively matched duration to be greater than or equal to preset matching duration to corresponding feature, illustrates each in first set
Matching status is belonged in preset matching duration to corresponding matching object with image, so that the first tag representation matches object
Belong to matching status in preset matching duration.The occurrence of first label can be self-defined as needed, is such as defined as 1.
Step S332, if being effectively matched duration is less than preset matching duration, by the corresponding matching of married matching object
Image corresponds to the second label to being included into second set, second set.
Specifically, the matching image wherein in second set is to be effectively matched duration to be less than default to corresponding feature
With duration, illustrate that each matching image in second set does not belong to corresponding matching object in preset matching duration
Matching status, so that the second tag representation matching object belongs to non-matched state in preset matching duration.The tool of second label
Body value can be self-defined as needed, is such as defined as 0.
Step S333, if the relation for being effectively matched duration and preset matching duration does not determine, by married matching object
Corresponding matching image is to being classified as invalid data.
Specifically, for being effectively matched the undetermined image that matches of duration and the relation of preset matching duration to not being included into
Any one set, becomes invalid data, so as to further increase the accuracy of training data, ensures trained depth
The accuracy that neutral net is predicted.
In one embodiment, step S240 includes:If matching result is the first label, it is determined that the first object with
Second object is successful match state in the range of preset matching duration, if it is the second label to state matching result, it is determined that
First object and the second object are it fails to match state in the range of preset matching duration.
Specifically, whether the matching result and default label exported according to trained deep learning neutral net be identical,
To judge the matching status of the first object and the second object in the range of preset matching duration, matching result can be immediately arrived at,
It is simple and convenient.The form of first label and the second label can be self-defined as needed, is such as represented with flag bit, and such as 1 represents first
Label, 0 represents the second label.
In one embodiment, step S210 includes:The matching request that request matching object user terminal is sent is received,
Matching request includes request matching object image data, will request matching object image data as the first object image data,
Object image data to be recommended is obtained as the second object image data.
Specifically, request matching object refers to need to carry out matched object, and request matching object image data is request
The characteristic for matching the picture of object or being extracted according to picture, object to be recommended refer to that matching object with request is matched
Candidate target.Object image data to be recommended can be directly carried in matching request, can also be obtained from server to be recommended
Object image data.Object image data to be recommended is that picture or root that object carries out matched candidate target are matched with request
The characteristic extracted according to picture.Object image data to be recommended can dynamically be updated according to the renewal of object to be recommended.
In one embodiment, the picture information for registered user being obtained from love and marriage matching website forms object diagram to be recommended
As data, object to be recommended can be screened according to the basic document of request matching object.
After step S240, further include:If request matches the scope of object and object to be recommended in preset matching duration
Interior matching status is successful match state, then sends the corresponding user of object to be recommended to request matching object user terminal
Information.
If specifically, request matching object and matching status of the object to be recommended in the range of preset matching duration
For the matching relationship stabilization of successful match state, then explanation request matching object and object to be recommended, object to be recommended is corresponded to
User information pushing to ask matching object user's terminal.The only matching relationship of request matching object and object to be recommended
Stablize, can just be pushed, improve the validity and objectivity of request matching object acquisition object to be recommended, improve request
Match the Interest Measure of object, improve matching efficiency, avoid a large amount of invalid objects to be recommended push user is formed it is dry
Disturb.
In a specific embodiment, the detailed process of love and marriage object matching data processing method is as follows:
1st, image recognition is carried out according to sex character and face characteristic from marriage certificate, extracts the bridegroom's or husband's side per a pair of of man and wife
Photo and wife's side photo, and the bridegroom's or husband's side photo and wife's side photo of every a pair of of man and wife are pre-processed according to preset data form,
Matching image is added to set, if in the case of allowing same-sex marriage, man and wife's photo can be at the same time man, can also be at the same time
For female.Matching image is combined into collection in { male 1 photo, 1 photo of female;Male 2 photos, 2 photo of female;Male 3 photos, 3 photo of female;Man 4 is shone
Piece, 4 photo of female;Male 5 photos, 5 photo of female;Male 6 photos, 6 photo of female;Male 7 photos, 7 photo of female }.
2nd, the marital status of each man and wife can be obtained from the marriage registration data in marriage big data, obtained and got married
Date, acquisition whether divorced or be bereft of one's spouse or it is double die, if divorced, obtain the date of divorce, will divorce the date subtract knot
The wedding date obtains being effectively matched duration.If do not divorced so far, acquisition whether be bereft of one's spouse or it is double die or wedding, if be bereft of one's spouse
Or it is double die, then will be effectively matched duration be arranged to all one's life, if wedded, get Date, grow up when obtaining being effectively matched
The divorce date is subtracted in current date.
For example, male 1 female 1:Male 1 photo, 1 photo of female are obtained by marriage certificate;On the date 19780920 of marriage, divorce,
Divorce the date 19980920, be effectively matched duration 20 years.
Male 2 female 2:Male 2 photos, 2 photo of female are obtained by marriage certificate;On the date 19580920 of marriage, divorce, divorced
On the date 19680920, be effectively matched duration 10 years.
Male 3 female 3:Male 3 photos, 3 photo of female are obtained by marriage certificate;On the date 19880920 of marriage, do not divorce, lost
Even, current date 2017-9-20, is effectively matched duration all one's life.
Male 4 female 4:Male 4 photos, 4 photo of female are obtained by marriage certificate;On the date 19650920 of marriage, do not divorce, it is double
Die, current date 2017-9-20, be effectively matched duration all one's life.
Male 5 female 5:Male 5 photos, 5 photo of female are obtained by marriage certificate;On the date 19570920 of marriage, do not divorce, wedding,
Current date 2017-9-20, was effectively matched duration more than 60 years.
Male 6 female 6:Male 6 photos, 6 photo of female are obtained by marriage certificate;On the date 19970920 of marriage, do not divorce, wedding,
Current date 2017-9-20, was effectively matched duration more than 20 years.
Male 7 female 7:Male 6 photos, 6 photo of female are obtained by marriage certificate;On the date 20070920 of marriage, do not divorce, wedding,
Current date 2017-9-20, was effectively matched duration more than 10 years.
3rd, the preset matching duration of deep learning neural network training model is obtained, for example, a length of 16 years during preset matching
Or a length of all one's life during preset matching.
4th, obtained from marriage big data and be effectively matched photograph of the duration more than or equal to each man and wife of preset matching duration
Piece, adds the first set of man and wife's photo pair.Obtained from marriage big data and be effectively matched duration less than preset matching duration
The photo of each man and wife, adds the second set of man and wife's photo pair.
For example, when preset matching at a length of 16 years,
Male 1 female 1 is effectively matched duration 20 years, and more than preset matching duration 16 years, male 1 photo, 1 photo of female added man and wife
The first set of photo pair;
Because male 2 female 2 are effectively matched duration 10 years, less than preset matching duration 16 years, male 2 photos, 2 photo of female added
The second set of man and wife's photo pair;
Because male 3 female 3 are effectively matched duration all one's life, more than preset matching duration 16 years, male 3 photos, 3 photo of female added
The first set of man and wife's photo pair;
Because male 4 female 4 are effectively matched duration all one's life, more than preset matching duration 16 years, male 4 photos, 4 photo of female added
The first set of man and wife's photo pair;
Because male 5 female 5 were effectively matched duration more than 60 years, more than preset matching duration 16 years, male 5 photos, 5 photo of female
Add the first set of man and wife's photo pair;
Because male 6 female 6 were effectively matched duration more than 20 years, more than preset matching duration 16 years, male 6 photos, 6 photo of female
Add the first set of man and wife's photo pair;
Because male 7 female 7 were effectively matched duration more than 10 years, can not judge whether to be more than or less than preset matching duration 16
Year, the first set of man and wife's photo pair had both been added without, has also been added without the second set of man and wife's photo pair.
Finally, first set is { male 1 photo, 1 photo of female;Male 3 photos, 3 photo of female;Male 4 photos, 4 photo of female;Man 5
Photo, 5 photo of female;Male 6 photos, 6 photo of female }, second set is { male 2 photos, 2 photo of female }.
For example, when preset matching during a length of all one's life,
Male 1 female 1 is effectively matched duration 20 years, and less than preset matching duration all one's life, male 1 photo, 1 photo of female add man and wife
The second set of photo pair;
Because male 2 female 2 are effectively matched duration 10 years, less than preset matching duration all one's life, male 2 photos, 2 photo of female add
The second set of man and wife's photo pair;
Because male 3 female 3 are effectively matched duration all one's life, equal to preset matching duration all one's life, male 3 photos, 3 photo of female add
The first set of man and wife's photo pair;
Because male 4 female 4 are effectively matched duration all one's life, equal to preset matching duration all one's life, male 4 photos, 4 photo of female add
The first set of man and wife's photo pair;
Because male 5 female 5 were effectively matched duration more than 60 years, can not judge whether to be more than or less than preset matching duration one
Raw, male 5 photos, 5 photo of female be both added without the first set of man and wife's photo pair, were also added without the second set of man and wife's photo pair;
Because male 6 female 6 were effectively matched duration more than 20 years, can not judge whether to be more than or less than preset matching duration one
Raw, male 6 photos, 6 photo of female be both added without the first set of man and wife's photo pair, were also added without the second set of man and wife's photo pair;
Because male 7 female 7 were effectively matched duration more than 10 years, can not judge whether to be more than or less than preset matching duration one
Raw, male 7 photos, 7 photo of female be both added without the first set of man and wife's photo pair, were also added without the second set of man and wife's photo pair.
Finally, first set is { male 3 photos, 3 photo of female;Male 4 photos, 4 photo of female }, second set for male 1 photo,
1 photo of female;Male 2 photos, 2 photo of female }.
5th, deep neural network is initialized so that the receptible data format of input layer of the deep neural network is photograph
Piece, 2 default labels for exporting node layer correspond in preset matching duration without divorce in divorce, preset matching duration respectively
(can be represented with value 0 and 1).
For example, initialization deep learning neutral net so that the receptible number of input layer of the deep learning neutral net
Form according to form for photo in the set of man and wife's photo pair;When the output label for exporting node layer is 1 interval scale preset matching
Do not divorce in length, output label is divorce in 0 interval scale preset matching duration
6th, it is right using each man and wife's photo obtained from marriage big data to the input as the deep neural network
Deep learning neutral net carries out unsupervised training.
Using male 1 photo, 1 photo of female as the input of deep learning neutral net, nothing is carried out to deep learning neutral net
Supervised training;
Using male 2 photos, 2 photo of female as the input of deep learning neutral net, nothing is carried out to deep learning neutral net
Supervised training;
Until matching, each of image pair inputs photo deep learning neutral net, to deep learning neutral net
Carry out unsupervised training.
7th, the input using each man and wife's photo in the first set of man and wife's photo pair as deep neural network, will be default
Match in duration without corresponding default label of divorcing, such as 1, the output of output node is corresponded to as the deep neural network
Value, Training is carried out to the deep neural network.
Such as, when preset matching at a length of 16 years,
Using male 1 photo, 1 photo of female as deep learning neutral net input, using 1 as deep learning neutral net
Corresponding anticipated output, Training is carried out to deep learning neutral net.
Using male 3 photos, 3 photo of female as deep learning neutral net input, using 1 as deep learning neutral net
Corresponding anticipated output, Training is carried out to the deep learning neutral net;
Deep learning neutral net is all inputted to photo until each in first set, to deep learning neutral net into
Row Training.
8th, the input using each man and wife's photo in the second set of man and wife's photo pair as deep neural network, will be default
The corresponding default label of divorce in duration, such as 0 are matched, the output valve of output node is corresponded to as the deep neural network, it is right
The deep neural network carries out Training.
Such as, when preset matching at a length of 16 years,
Using male 2 photos, 2 photo of female as deep learning neutral net input, using 0 as deep learning neutral net
Corresponding anticipated output, Training is carried out to deep learning neutral net.
Deep learning neutral net is all inputted to photo until each in second set, to deep learning neutral net into
Row Training.
9th, trained deep learning neutral net is finally obtained.
10th, the first object images A to be matched is obtained, object images B1, B2, B3, B4, B5 to be recommended are obtained, according to depth
Spend input format image B1, B2, B3, B4, B5 to the first object images A and each object to be recommended of learning neural network
Pre-processed, specifically included:
Judge image B1, B2, B3, B4, B5 and preset data form of the first object images A and each object to be recommended
It is whether consistent, if it is inconsistent, being preset data form by the format conversion of the image.
Preset matching duration is consistent with preset matching duration during training neutral net, by the first object images A and each
Input of the image of object to be recommended as deep learning neutral net, obtains deep learning neutral net and corresponds to output node
Output label, if output label with preset matching duration without the corresponding default label of divorce, to the first object correspondence
Terminal push object information to be recommended, if output label is identical with corresponding default label of divorcing in preset matching duration,
This object information to be recommended is not pushed to the corresponding terminal of the first object.
Preset duration during such as deep learning neural metwork training is 16 years;
Using corresponding two photos of two of the marriage certificate that waits for claimant objects to be matched as the defeated of deep learning neutral net
Enter, the output label that deep learning neutral net corresponds to output node is obtained, if output label in preset duration with not divorcing
Corresponding default label 1 is identical, if then predicting that friend men and women gets married, will not divorce in 16 years in preset duration, if defeated
Outgoing label is identical with the corresponding default label 0 of divorce in preset duration, if then predicting that friend men and women gets married, default
Duration can divorce in 16 years.
Preset duration during such as deep learning neural metwork training is all one's life;
Using corresponding two photos of two of the marriage certificate that waits for claimant objects to be matched as the defeated of deep learning neutral net
Enter, obtain the output label that the deep learning neutral net corresponds to output node, if output label in preset duration with not having
Corresponding default label 1 of divorcing is identical, if then predicting that friend men and women gets married, will not divorce within preset duration all one's life,
If output label is identical with the corresponding default label 0 of divorce in preset duration, if predicting that friend men and women gets married,
It can divorce in preset duration all one's life.
In one embodiment, as shown in Figure 7, there is provided one kind pushes object information device to be recommended, including:
Acquisition module 510, for obtaining the first object image data and the second object image data to be matched.
Matching result acquisition module 520, for the first object image data and the input of the second object image data to have been instructed
Experienced deep learning neutral net, the parameter of deep learning neutral net is big with the corresponding matching image of married matching object
Data are as input, and by the married matching, object is corresponding is effectively matched duration and deep learning neural network training model
The corresponding label that relatively obtains of preset matching duration as anticipated output, input deep learning neural network training model
Training obtains, the corresponding matching result of deep learning neutral net trained output described in acquisition.
Matching status determining module 530, for determining the first object and the second object in preset matching according to matching result
Matching status in the range of duration.
In one embodiment, as shown in figure 8, device further includes:
Training module 540, for obtaining training image big data, training image big data includes married matching object pair
The matching image big data answered, extracts the corresponding matching figure of corresponding each married matching object from matching image big data
As right, the married matching object of acquisition is corresponding to be effectively matched duration, obtains default of deep learning neural network training model
With duration, corresponding matching is schemed according to the corresponding relation for being effectively matched duration and preset matching duration of married matching object
For picture to being included into corresponding goal set, different goal sets corresponds to different labels, using each matching image to as depth
The input of learning neural network, carries out unsupervised training to deep learning neutral net, obtains successively in each goal set
Object matching image pair label corresponding with goal set forms matching relationship, successively by the target in each goal set
With image to the input as deep learning neutral net, the label of matching relationship is will be provided with as deep learning neutral net
Anticipated output, to deep learning neutral net carry out Training, the deep learning neutral net trained.
In one embodiment, training module 540 is additionally operable to obtain the matching initial time of married matching object and current
Matching status, if current matching state is non-matching, obtains matching and terminates the time, and the time is terminated with matching according to matching
Initial time obtains being effectively matched duration, if current matching state is matching, obtains married matching object both sides' state,
If married matching object both sides' state is normal, current time is obtained, is had according to current time with matching initial time
Effect matching duration scope, if at least one party's abnormal state in married matching object both sides, will be effectively matched duration and be determined as
Preset duration.
In one embodiment, if training module 540 is additionally operable to be effectively matched duration when being greater than or equal to preset matching
It is long, then the corresponding matching image of married matching object is corresponded into the first label to being included into first set, first set, if
Effect matching duration is less than preset matching duration, then described by the corresponding matching image of married matching object to being included into second set
Second set corresponds to the second label, if the relation for being effectively matched duration and preset matching duration does not determine, by married matching
The corresponding matching image of object is to being classified as invalid data
In one embodiment, if it is the first label that matching status determining module 530, which is additionally operable to matching result, really
Fixed first object and the second object are successful match state in the range of the preset matching duration, if matching result is
Second label, it is determined that the first object and the second object are it fails to match state in the range of preset matching duration.
In one embodiment, acquisition module 510, which is additionally operable to receive, asks the matching of matching object user terminal transmission please
Ask, matching request includes request matching object image data, using request matching object image data as first object diagram
As data, object image data to be recommended is obtained as second object image data.
As shown in figure 9, device further includes:
Sending module 550, if for asking matching object and object to be recommended in the scope of the preset matching duration
Interior matching status is successful match state, then it is corresponding to send object to be recommended to the request matching object user terminal
User information.
In one embodiment it is proposed that a kind of computer equipment, including memory and processor, memory storage have
Computer program, when computer program is executed by processor so that processor performs following steps:Obtain to be matched first
Object image data and the second object image data;First object image data and the input of the second object image data have been instructed
Experienced deep learning neutral net, the parameter of deep learning neutral net is big with the corresponding matching image of married matching object
Data are effectively matched the pre- of duration and deep learning neural network training model as input, by married matching object is corresponding
If the corresponding label that matching duration relatively obtains is as anticipated output, the neural network training model training of input deep learning
Obtain, obtain the corresponding matching result of trained deep learning neutral net output, first pair is determined according to matching result
As the matching status with the second object in the range of the preset matching duration.
In one embodiment, the processor is performing acquisition the first object image data to be matched and second pair
Before the step of view data, it is additionally operable to perform following steps:Obtain training image big data, the big number of training image
According to including the corresponding matching image big data of married matching object, extracted from matching image big data corresponding each married
The corresponding matching image pair of object is matched, the married matching object of acquisition is corresponding to be effectively matched duration, obtains deep learning nerve
The preset matching duration of network training model, according to married matching, object is corresponding is effectively matched duration and preset matching duration
Relation by corresponding matching image to being included into corresponding goal set, different goal sets corresponds to different labels, will be respectively
A matching image carries out unsupervised training, successively to the input as deep learning neutral net to deep learning neutral net
The object matching image pair label corresponding with goal set obtained in each goal set forms matching relationship, successively will be each
Object matching image in a goal set will be provided with the label of matching relationship to the input as deep learning neutral net
As the anticipated output of the deep learning neutral net, Training is carried out to deep learning neutral net, has been instructed
Experienced deep learning neutral net.
In one embodiment, obtaining married matching object corresponding the step of being effectively matched duration includes:Obtain married
The matching initial time and current matching state of object are matched, if current matching state is non-matching, matching is obtained and terminates
Time, obtains being effectively matched duration according to the matching termination time with matching initial time, if current matching state is matching,
Married matching object both sides' state is then obtained, if married matching object both sides' state is normal, obtains current time, according to
Current time obtains being effectively matched duration scope with matching initial time, if at least one party's shape in married matching object both sides
State is abnormal, then will be effectively matched duration and be determined as preset duration.
In one embodiment, according to the corresponding pass for being effectively matched duration and preset matching duration of married matching object
It is that the step of different goal sets corresponds to different labels is wrapped by corresponding matching image to being included into corresponding goal set
Include:If being effectively matched duration is greater than or equal to preset matching duration, by the corresponding matching image of married matching object to returning
Enter first set, the first set corresponds to the first label, if being effectively matched duration is less than preset matching duration, by
The corresponding matching image of wedding matching object is to being included into second set, and the second set corresponds to the second label, if be effectively matched
The relation of duration and preset matching duration does not determine, then by the corresponding matching image of married matching object to being classified as invalid data.
In one embodiment, determine the first object and the second object in the preset matching according to the matching result
The step of matching status in the range of duration, includes:If matching result is the first label, it is determined that first object with
Second object is successful match state in the range of preset matching duration, if matching result is the second label, it is determined that the
An object and the second object are it fails to match state in the range of preset matching duration.
In one embodiment, the step of matching degree of the first object and the second object is determined according to the matching result
Including:If the matching result is identical with preset matching label, it is determined that the matching degree of first object and the second object
For successful match;If the matching result is identical with presetting non-matching label, it is determined that first object and the second object
Matching degree for it fails to match.
In one embodiment, the step of obtaining the first object image data and the second object image data to be matched
Including:The matching request that request matching object user terminal is sent is received, the matching request includes request matching object images
Data, match object image data as first object image data using the request, obtain object images number to be recommended
According to as second object image data.
The processor determines the model of the first object and the second object in preset matching duration in execution according to matching result
After the step of enclosing interior matching status, it is additionally operable to perform following steps:If request matching object is with object to be recommended pre-
If the matching status in the range of matching duration is successful match state, is then sent to request matching object user's terminal and wait to push away
Recommend the corresponding user information of object.
In one embodiment it is proposed that a kind of computer-readable recording medium, is stored with computer program, the meter
When calculation machine program is executed by processor so that the processor performs following steps:Obtain the first object images number to be matched
According to the second object image data;By the trained depth of the first object image data and the input of the second object image data
Practise neutral net, the parameter of deep learning neutral net is using the corresponding matching image big data of married matching object as defeated
Enter, by the corresponding preset matching duration for being effectively matched duration and deep learning neural network training model of married matching object
The corresponding label compared is trained to obtain, obtained as anticipated output, input deep learning neural network training model
The corresponding matching result of trained deep learning neutral net output, the first object and second pair are determined according to matching result
As the matching status in the range of preset matching duration.
In one embodiment, the processor is performing acquisition the first object image data to be matched and second pair
Before the step of view data, it is additionally operable to perform following steps:Obtain training image big data, the big number of training image
According to including the corresponding matching image big data of married matching object, extracted from matching image big data corresponding each married
The corresponding matching image pair of object is matched, the married matching object of acquisition is corresponding to be effectively matched duration, obtains deep learning nerve
The preset matching duration of network training model, according to married matching, object is corresponding is effectively matched duration and preset matching duration
Relation by corresponding matching image to being included into corresponding goal set, different goal sets corresponds to different labels, will be respectively
A matching image carries out unsupervised training, successively to the input as deep learning neutral net to deep learning neutral net
The object matching image pair label corresponding with goal set obtained in each goal set forms matching relationship, successively will be each
Object matching image in a goal set will be provided with the label of matching relationship to the input as deep learning neutral net
As the anticipated output of the deep learning neutral net, Training is carried out to deep learning neutral net, has been instructed
Experienced deep learning neutral net.
In one embodiment, obtaining married matching object corresponding the step of being effectively matched duration includes:Obtain married
The matching initial time and current matching state of object are matched, if current matching state is non-matching, matching is obtained and terminates
Time, obtains being effectively matched duration according to the matching termination time with matching initial time, if current matching state is matching,
Married matching object both sides' state is then obtained, if married matching object both sides' state is normal, obtains current time, according to
Current time obtains being effectively matched duration scope with matching initial time, if at least one party's shape in married matching object both sides
State is abnormal, then will be effectively matched duration and be determined as preset duration.
In one embodiment, according to the corresponding pass for being effectively matched duration and preset matching duration of married matching object
It is that the step of different goal sets corresponds to different labels is wrapped by corresponding matching image to being included into corresponding goal set
Include:If being effectively matched duration is greater than or equal to preset matching duration, by the corresponding matching image of married matching object to returning
Enter first set, the first set corresponds to the first label, if being effectively matched duration is less than preset matching duration, by
The corresponding matching image of wedding matching object is to being included into second set, and the second set corresponds to the second label, if be effectively matched
The relation of duration and preset matching duration does not determine, then by the corresponding matching image of married matching object to being classified as invalid data.
In one embodiment, determine the first object and the second object in the preset matching according to the matching result
The step of matching status in the range of duration, includes:If matching result is the first label, it is determined that first object with
Second object is successful match state in the range of preset matching duration, if matching result is the second label, it is determined that the
An object and the second object are it fails to match state in the range of preset matching duration.
In one embodiment, the step of matching degree of the first object and the second object is determined according to the matching result
Including:If the matching result is identical with preset matching label, it is determined that the matching degree of first object and the second object
For successful match;If the matching result is identical with presetting non-matching label, it is determined that first object and the second object
Matching degree for it fails to match.
In one embodiment, the step of obtaining the first object image data and the second object image data to be matched
Including:The matching request that request matching object user terminal is sent is received, the matching request includes request matching object images
Data, match object image data as first object image data using the request, obtain object images number to be recommended
According to as second object image data.
The processor determines the model of the first object and the second object in preset matching duration in execution according to matching result
After the step of enclosing interior matching status, it is additionally operable to perform following steps:If request matching object is with object to be recommended pre-
If the matching status in the range of matching duration is successful match state, is then sent to request matching object user's terminal and wait to push away
Recommend the corresponding user information of object.
One of ordinary skill in the art will appreciate that realizing all or part of flow in above-described embodiment method, being can
To instruct relevant hardware to complete by computer program, described program can be stored in a computer-readable storage and be situated between
In matter, in the embodiment of the present invention, which can be stored in the storage medium of computer system, and by the computer system
In at least one processor perform, with realize include as above-mentioned each method embodiment flow.Wherein, the storage is situated between
Matter can be magnetic disc, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random
Access Memory, RAM) etc..
Each technical characteristic of embodiment described above can be combined arbitrarily, to make description succinct, not to above-mentioned reality
Apply all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, the scope that this specification is recorded all is considered to be.
Embodiment described above only expresses the several embodiments of the present invention, its description is more specific and detailed, but simultaneously
Cannot therefore it be construed as limiting the scope of the patent.It should be pointed out that come for those 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 the guarantor of the present invention
Protect scope.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (10)
1. a kind of love and marriage object matching data processing method, the described method includes:
Obtain the first object image data and the second object image data to be matched;
It is described by the trained deep learning neutral net of first object image data and the input of the second object image data
The parameter of deep learning neutral net be using it is married matching object it is corresponding matching image big data as input, will be described married
The corresponding preset matching duration for being effectively matched duration and deep learning neural network training model of matching object relatively obtains
Corresponding label trains to obtain as anticipated output, input deep learning neural network training model;
The corresponding matching result of deep learning neutral net trained output described in obtaining;
The matching shape of first object and the second object in the range of the preset matching duration is determined according to the matching result
State.
2. according to the method described in claim 1, it is characterized in that, described obtain the first object image data to be matched and the
Before the step of two object image datas, further include:
Training image big data is obtained, the training image big data includes the corresponding matching big number of image of married matching object
According to, from it is described matching image big data in extract it is corresponding it is each it is married matching object it is corresponding matching image pair;
Obtain the corresponding preset matching for being effectively matched duration, obtaining deep learning neural network training model of married matching object
Duration;
Image pair is matched by corresponding according to the corresponding relation for being effectively matched duration and preset matching duration of married matching object
Corresponding goal set is included into, different goal sets corresponds to different labels;
Using each matching image to the input as the deep learning neutral net, the deep learning neutral net is carried out
Unsupervised training;
The object matching image pair label corresponding with goal set obtained successively in each goal set forms matching relationship, according to
The secondary object matching image using in each goal set will be provided with matching and close to the input as the deep learning neutral net
Anticipated output of the label of system as the deep learning neutral net, has carried out the deep learning neutral net supervision instruction
Practice;
Obtain the trained deep learning neutral net.
3. according to the method described in claim 2, it is characterized in that, described obtain married matching object be corresponding when being effectively matched
Long step includes:
Obtain the matching initial time and current matching state of married matching object;
If current matching state is non-matching, obtains matching and terminate the time, when terminating the time with matching starting according to matching
Between obtain being effectively matched duration;
If current matching state is matching, married matching object both sides' state is obtained, if married matching object both sides' shape
State is normal, then obtains current time, obtains being effectively matched duration scope with matching initial time according to current time;
If at least one party's abnormal state in married matching object both sides, will be effectively matched duration and be determined as preset duration.
4. according to the method described in claim 2, it is characterized in that, it is described according to it is married matching object is corresponding be effectively matched when
Long and preset matching duration relation corresponds to corresponding matching image to being included into corresponding goal set, different goal sets
The step of different label, includes:
If being effectively matched duration is greater than or equal to preset matching duration, by the corresponding matching image of the married matching object
To being included into first set, the first set corresponds to the first label;
If being effectively matched duration is less than preset matching duration, by the married corresponding matching image of object that matches to being included into
Second set, the second set correspond to the second label;
If the relation for being effectively matched duration and preset matching duration does not determine, by the corresponding matching of the married matching object
Image is to being classified as invalid data.
5. according to the method described in claim 1, it is characterized in that, described determine the first object and according to the matching result
Two objects include in the step of matching status in the range of the preset matching duration:
If the matching result is the first label, it is determined that first object and the second object are in the preset matching duration
In the range of be successful match state;
If the matching result is the second label, it is determined that first object and the second object are in the preset matching duration
In the range of for it fails to match state.
6. according to the method described in claim 1, it is characterized in that, described obtain the first object image data to be matched and the
The step of two object image datas, includes:
The matching request that request matching object user terminal is sent is received, the matching request includes request matching object images number
According to, using it is described request match object image data as first object image data, obtain object image data to be recommended
As second object image data;
It is described that of first object and the second object in the range of the preset matching duration is determined according to the matching result
After the step of state, further include:
If matching status of the request matching object with object to be recommended in the range of the preset matching duration is to match into
Work(state, then send the corresponding user information of object to be recommended to the request matching object user terminal.
7. a kind of love and marriage object matching data processing equipment, it is characterised in that described device includes:
Acquisition module, for obtaining the first object image data and the second object image data to be matched;
Matching result acquisition module, for trained first object image data and the input of the second object image data
Deep learning neutral net, the parameter of the deep learning neutral net is with the corresponding matching big number of image of married matching object
According to as input, the pre- of duration and deep learning neural network training model is effectively matched by the married matching object is corresponding
If the corresponding label that relatively obtains of matching duration is used as anticipated output, it is trained to input deep learning neural network training model
Arrive, the corresponding matching result of deep learning neutral net trained output described in acquisition;
Matching status determining module, for determining the first object and the second object in the preset matching according to the matching result
Matching status in the range of duration.
8. device according to claim 7, it is characterised in that the acquisition module is additionally operable to receive request matching object use
The matching request that family terminal is sent, the matching request include request matching object image data, and the request is matched object
View data obtains object image data to be recommended as the second object images number as first object image data
According to;
Described device further includes:
Sending module, if for asking matching object and matching of the object to be recommended in the range of the preset matching duration
State is successful match state, then sends the corresponding user information of object to be recommended to the request matching object user terminal.
9. a kind of computer equipment, including memory and processor, the memory storage has computer program, the computer
When program is performed by the processor so that the processor performs the step of the method as any one of claim 1 to 6
Suddenly.
10. a kind of computer-readable recording medium, is stored with computer program, when the computer program is executed by processor,
So that the processor is performed as any one of claim 1 to 6 the step of method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711260402.5A CN108038496B (en) | 2017-12-04 | 2017-12-04 | Marriage and love object matching data processing method and device based on big data and deep learning, computer equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711260402.5A CN108038496B (en) | 2017-12-04 | 2017-12-04 | Marriage and love object matching data processing method and device based on big data and deep learning, computer equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108038496A true CN108038496A (en) | 2018-05-15 |
CN108038496B CN108038496B (en) | 2021-07-30 |
Family
ID=62094982
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711260402.5A Active CN108038496B (en) | 2017-12-04 | 2017-12-04 | Marriage and love object matching data processing method and device based on big data and deep learning, computer equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108038496B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109857944A (en) * | 2018-12-24 | 2019-06-07 | 深圳市珍爱捷云信息技术有限公司 | Marriage and making friend's recommended method, device, computer equipment and computer storage medium |
CN112365104A (en) * | 2020-12-07 | 2021-02-12 | 杭州师范大学 | Marital matching method for predicting maximum marital satisfaction |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1774716A (en) * | 2003-04-16 | 2006-05-17 | 戴维·艾克门 | An internet system for authenticating membership profile information |
CN101044504A (en) * | 2004-10-19 | 2007-09-26 | 雅虎公司 | System and method for location based social networking |
CN102073715A (en) * | 2011-01-04 | 2011-05-25 | 吴欣 | Intelligent information exchange method and system |
CN103678394A (en) * | 2012-09-21 | 2014-03-26 | 孟露芳 | Image matching degree based marriage dating recommendation method and system |
CN103984775A (en) * | 2014-06-05 | 2014-08-13 | 网易(杭州)网络有限公司 | Friend recommending method and equipment |
CN104657477A (en) * | 2015-02-16 | 2015-05-27 | 百度在线网络技术(北京)有限公司 | Social contact recommendation method and device |
CN105389548A (en) * | 2015-10-23 | 2016-03-09 | 南京邮电大学 | Love and marriage evaluation system and method based on face recognition |
CN105898420A (en) * | 2015-01-09 | 2016-08-24 | 阿里巴巴集团控股有限公司 | Video recommendation method and device, and electronic equipment |
WO2016195496A2 (en) * | 2015-06-05 | 2016-12-08 | Universiteit Van Amsterdam | Deep receptive field networks |
CN106504104A (en) * | 2016-10-27 | 2017-03-15 | 江西瓷肌电子商务有限公司 | A kind of method of social activity of being made friends based on face recognition |
CN106980880A (en) * | 2017-03-06 | 2017-07-25 | 北京小米移动软件有限公司 | The method and device of images match |
CN107169031A (en) * | 2017-04-17 | 2017-09-15 | 广东工业大学 | It is a kind of to recommend method based on the picture materials that depthmeter reaches |
-
2017
- 2017-12-04 CN CN201711260402.5A patent/CN108038496B/en active Active
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1774716A (en) * | 2003-04-16 | 2006-05-17 | 戴维·艾克门 | An internet system for authenticating membership profile information |
CN101044504A (en) * | 2004-10-19 | 2007-09-26 | 雅虎公司 | System and method for location based social networking |
CN102073715A (en) * | 2011-01-04 | 2011-05-25 | 吴欣 | Intelligent information exchange method and system |
CN103678394A (en) * | 2012-09-21 | 2014-03-26 | 孟露芳 | Image matching degree based marriage dating recommendation method and system |
CN103984775A (en) * | 2014-06-05 | 2014-08-13 | 网易(杭州)网络有限公司 | Friend recommending method and equipment |
CN105898420A (en) * | 2015-01-09 | 2016-08-24 | 阿里巴巴集团控股有限公司 | Video recommendation method and device, and electronic equipment |
CN104657477A (en) * | 2015-02-16 | 2015-05-27 | 百度在线网络技术(北京)有限公司 | Social contact recommendation method and device |
WO2016195496A2 (en) * | 2015-06-05 | 2016-12-08 | Universiteit Van Amsterdam | Deep receptive field networks |
CN105389548A (en) * | 2015-10-23 | 2016-03-09 | 南京邮电大学 | Love and marriage evaluation system and method based on face recognition |
CN106504104A (en) * | 2016-10-27 | 2017-03-15 | 江西瓷肌电子商务有限公司 | A kind of method of social activity of being made friends based on face recognition |
CN106980880A (en) * | 2017-03-06 | 2017-07-25 | 北京小米移动软件有限公司 | The method and device of images match |
CN107169031A (en) * | 2017-04-17 | 2017-09-15 | 广东工业大学 | It is a kind of to recommend method based on the picture materials that depthmeter reaches |
Non-Patent Citations (1)
Title |
---|
李钰等: "利用深度学习在人脸图像中识别亲缘关系", 《第十一届和谐人机环境联合会议论文集》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109857944A (en) * | 2018-12-24 | 2019-06-07 | 深圳市珍爱捷云信息技术有限公司 | Marriage and making friend's recommended method, device, computer equipment and computer storage medium |
CN112365104A (en) * | 2020-12-07 | 2021-02-12 | 杭州师范大学 | Marital matching method for predicting maximum marital satisfaction |
Also Published As
Publication number | Publication date |
---|---|
CN108038496B (en) | 2021-07-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111199550B (en) | Training method, segmentation method, device and storage medium of image segmentation network | |
CN108182389B (en) | User data processing method based on big data and deep learning and robot system | |
CN109902546A (en) | Face identification method, device and computer-readable medium | |
CN110084281A (en) | Image generating method, the compression method of neural network and relevant apparatus, equipment | |
CN108062780A (en) | Method for compressing image and device | |
CN112446302B (en) | Human body posture detection method, system, electronic equipment and storage medium | |
CN112418292B (en) | Image quality evaluation method, device, computer equipment and storage medium | |
CN111428771B (en) | Video scene classification method and device and computer-readable storage medium | |
CN110222717A (en) | Image processing method and device | |
CN111741330A (en) | Video content evaluation method and device, storage medium and computer equipment | |
CN110288513B (en) | Method, apparatus, device and storage medium for changing face attribute | |
CN112395979A (en) | Image-based health state identification method, device, equipment and storage medium | |
CN113052150B (en) | Living body detection method, living body detection device, electronic apparatus, and computer-readable storage medium | |
CN109918630A (en) | Document creation method, device, computer equipment and storage medium | |
CN112232325A (en) | Sample data processing method and device, storage medium and electronic equipment | |
CN113822953A (en) | Processing method of image generator, image generation method and device | |
CN111046213B (en) | Knowledge base construction method based on image recognition | |
CN112529149A (en) | Data processing method and related device | |
CN115910319A (en) | Otology inquiry assisting method and device, electronic equipment and storage medium | |
CN108038496A (en) | Love and marriage object matching data processing method, device, computer equipment and storage medium based on big data and deep learning | |
CN108154165A (en) | Love and marriage object matching data processing method, device, computer equipment and storage medium based on big data and deep learning | |
CN116701706B (en) | Data processing method, device, equipment and medium based on artificial intelligence | |
CN113362852A (en) | User attribute identification method and device | |
CN115731620A (en) | Method for detecting counter attack and method for training counter attack detection model | |
CN113822790B (en) | Image processing method, device, equipment and computer readable storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |