CN110069988A - AI based on multidimensional data drives risk analysis method, server and storage medium - Google Patents
AI based on multidimensional data drives risk analysis method, server and storage medium Download PDFInfo
- Publication number
- CN110069988A CN110069988A CN201910195210.3A CN201910195210A CN110069988A CN 110069988 A CN110069988 A CN 110069988A CN 201910195210 A CN201910195210 A CN 201910195210A CN 110069988 A CN110069988 A CN 110069988A
- Authority
- CN
- China
- Prior art keywords
- sample
- risk analysis
- model
- preset
- risk
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012502 risk assessment Methods 0.000 title claims abstract description 251
- 238000000034 method Methods 0.000 title claims abstract description 118
- 238000004458 analytical method Methods 0.000 claims abstract description 88
- 238000012549 training Methods 0.000 claims description 114
- 230000015654 memory Effects 0.000 claims description 34
- 230000004927 fusion Effects 0.000 claims description 23
- 239000000284 extract Substances 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 5
- 238000003556 assay Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 3
- 230000001360 synchronised effect Effects 0.000 claims description 2
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 238000005516 engineering process Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 14
- 230000008569 process Effects 0.000 description 14
- 230000005540 biological transmission Effects 0.000 description 8
- 238000012545 processing Methods 0.000 description 7
- 238000011160 research Methods 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 238000000605 extraction Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 230000004438 eyesight Effects 0.000 description 4
- 239000011521 glass Substances 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 3
- 235000013399 edible fruits Nutrition 0.000 description 3
- 239000007787 solid Substances 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 210000004709 eyebrow Anatomy 0.000 description 2
- 230000001815 facial effect Effects 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 210000004209 hair Anatomy 0.000 description 1
- 230000007257 malfunction Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/251—Fusion techniques of input or preprocessed data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G06Q50/40—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/59—Context or environment of the image inside of a vehicle, e.g. relating to seat occupancy, driver state or inner lighting conditions
- G06V20/597—Recognising the driver's state or behaviour, e.g. attention or drowsiness
-
- 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
- G06V40/168—Feature extraction; Face representation
-
- 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
Abstract
The present invention relates to artificial intelligence technologys, a kind of AI driving risk analysis method based on multidimensional data are disclosed, this method comprises: Analysis server obtains the picture to be analyzed for the target user that terminal is sent；One or more human face recognition models trained in advance are called to identify the picture to be analyzed, to identify the face feature of risk of preset kind；User data corresponding with target user is extracted from predetermined database, and preset kind structured features data are extracted from the user data；And the face feature of risk of the preset kind and the preset kind structured features data are spliced into the integrated risk feature of multidimensional, and the driving risk analysis model that integrated risk feature input is trained in advance, with output integrated risk analysis value.The present invention is also disclosed that a kind of Analysis server and computer storage medium.Using the present invention, the accuracy and objectivity for driving risk analysis can be improved.
Description
Technical field
The present invention relates to field of artificial intelligence more particularly to a kind of AI based on multidimensional data to drive risk analysis side
Method, Analysis server and computer readable storage medium.
Background technique
In recent years, China's transportation industry flourishes, resident's car ownership substantial increase, and the following road is handed over
Interpreter's event frequently occurs, rear-end impact, rollover of vehicle etc., these traffic accidents are all largely the danger due to driver
Caused by dangerous driving behavior, such as speed is excessively high, fatigue driving.
Carry out that accurate, objective analysis significance is great to the driving risk of driver in advance, for example, vehicle insurance company can root
According to the accurate assay value for driving risk, the risk class that vehicle insurance is insured is determined；Traffic control department can be according to the accurate of driving risk
Assay value, the duration etc. when determining the safe driving training session of driver.
Although occurring the scheme that the driving risk to driver is analyzed on the market, currently existing scheme is usually
What the inertial thinking based on those skilled in the art was developed, and then currently existing scheme is caused to be normally based on year of driver
The structured features such as age, gender, former years loss ratio and situation of being in danger carry out driving risk analysis, analysis result it is accurate
Property it is usually lower, often malfunction.
Summary of the invention
In view of the foregoing, the present invention provides a kind of AI driving risk analysis method, Analysis Service based on multidimensional data
Device and computer readable storage medium, main purpose are to greatly improve the accuracy and objectivity for driving risk analysis.
To achieve the above object, the present invention provides a kind of AI driving risk analysis method based on multidimensional data, is applied to
Analysis server, the Analysis server are connected with terminal, this method comprises:
S1, the picture to be analyzed for receiving the target user that the terminal is sent, or in the needle for receiving the terminal transmission
After the driving risk analysis of target user request, control the terminal shoot the photo of the target user as it is described to
Analysis picture, or the user identifier of the target user according to the driving risk analysis request, from predetermined number
According to the picture to be analyzed for extracting the target user in library；
S2, one or more human face recognition models trained in advance are called to identify the picture to be analyzed, to know
Not Chu preset kind face feature of risk；
S3, user data corresponding with the target user is extracted from predetermined database, from the user
Preset kind structured features data are extracted in data；And
S4, the face feature of risk of the preset kind and the preset kind structured features data are spliced into multidimensional
Integrated risk feature, and the driving risk analysis model that integrated risk feature input is trained in advance, with output integrated
Risk analysis value.
Optionally, this method further include:
If the integrated risk analysis value of output is greater than the first preset threshold, the first preset format is sent to the terminal
Drive Risk-warning information；And/or
If the integrated risk analysis value of output is less than or equal to the first preset threshold, and is greater than the second preset threshold, then
The driving Risk-warning information of the second preset format is sent to the terminal；And/or
If the integrated risk analysis value of output is less than or equal to the second preset threshold, it is pre- that third is sent to the terminal
If the result feedback information of format.
Optionally, this method further include:
According to the mapping relations of predetermined integrated risk analysis value range and preset kind vehicle insurance property parameters, determine
The corresponding preset kind vehicle insurance property parameters of the integrated risk analysis value, and the preset kind vehicle insurance property parameters are sent
To the terminal.
In addition, the present invention also provides a kind of Analysis server, which includes: memory, processor, described to deposit
It is stored with can run on the processor first on reservoir and drives Risk analysis procedure, described first drives risk analysis quilt
, it can be achieved that the AI based on multidimensional data drives the arbitrary steps in risk analysis method as described above when the processor executes.
In addition, including in the computer readable storage medium the present invention also provides a kind of computer readable storage medium
First drives Risk analysis procedure, it can be achieved that base as described above when the first driving Risk analysis procedure is executed by processor
The arbitrary steps in risk analysis method are driven in the AI of multidimensional data.
In addition, to achieve the above object, the present invention also provides a kind of, and the AI based on multidimensional data drives risk analysis method,
Applied to Analysis server, the Analysis server is connected with terminal, this method comprises:
A1, the picture to be analyzed for receiving the target user that the terminal is sent, or in the needle for receiving the terminal transmission
After the driving risk analysis of target user request, control the terminal shoot the photo of the target user as it is described to
Analysis picture, or the user identifier of the target user according to the driving risk analysis request, from predetermined number
According to the picture to be analyzed for extracting the target user in library；
A2, user data corresponding with the target user is extracted from predetermined database, from the user
Preset kind structured features data are extracted in data；And
A3, the picture to be analyzed and the preset kind structured features data are inputted into integrated risk trained in advance
In analysis model, output integrated risk analysis value.
In addition, the present invention also provides a kind of Analysis server, which includes: memory, processor, described to deposit
It is stored with can run on the processor second on reservoir and drives Risk analysis procedure, described second drives risk analysis quilt
, it can be achieved that the AI based on multidimensional data drives the arbitrary steps in risk analysis method as described above when the processor executes.
In addition, including in the computer readable storage medium the present invention also provides a kind of computer readable storage medium
Second drives Risk analysis procedure, it can be achieved that base as described above when the second driving Risk analysis procedure is executed by processor
The arbitrary steps in risk analysis method are driven in the AI of multidimensional data.
The inertial thinking of those skilled in the art is normally based on age of driver, gender, former years loss ratio and goes out
The structured features such as dangerous situation condition carry out driving risk analysis, and especially in vehicle insurance business scope, this inertial thinking is even more root
The deep base of a fruit is solid.Compared to the prior art, the AI proposed by the present invention based on multidimensional data drives risk analysis method, Analysis server
And computer readable storage medium, effectively break this inertial thinking, creative has found face characteristic and driving risk
Objective connection, maximization played effect of the face characteristic in terms of driving risk objective analysis, effectively prevent structuring
The accuracy and objectivity for driving risk analysis greatly improved in unilateral influence of the characteristic on model.
Detailed description of the invention
Fig. 1 is the process that the AI based on multidimensional data that first embodiment of the invention proposes drives risk analysis method
Figure；
Fig. 2 is the ResBlock structure operation schematic diagram of face characteristic identification model；
Fig. 3 is the model structure schematic diagram of face characteristic identification model；
Fig. 4 a is the schematic diagram that face characteristic identification model extracts face feature of risk；
Fig. 4 b is the model structure schematic diagram for driving risk analysis model；
Fig. 5 is the model structure schematic diagram for driving the end-to-end various features Fusion training of support of risk analysis model；
Fig. 6 is the schematic diagram for the Analysis server that second embodiment of the invention proposes；
Fig. 7 is the first module diagram for driving Risk analysis procedure in Fig. 6；
Fig. 8 is the flow chart that the AI based on multidimensional data that fourth embodiment of the invention proposes drives risk analysis method；
Fig. 9 is the schematic diagram for the Analysis server that fifth embodiment of the invention proposes；
Figure 10 is the second module diagram for driving Risk analysis procedure in Fig. 9.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
First embodiment of the invention proposes that a kind of AI based on multidimensional data drives risk analysis method.Shown in referring to Fig.1,
For the flow chart for the AI driving risk analysis method based on multidimensional data that first embodiment of the invention proposes.This method can be by
One device executes, which can be by software and or hardware realization.
In the first embodiment, the AI based on multidimensional data drives risk analysis method applied to a kind of Analysis Service
Device, the Analysis server are connected with terminal, this method comprises: step S1-S4.
S1, the picture to be analyzed for receiving the target user that the terminal is sent, or in the needle for receiving the terminal transmission
After the driving risk analysis of target user request, control the terminal shoot the photo of the target user as it is described to
Analysis picture, or the user identifier of the target user according to the driving risk analysis request, from predetermined number
According to the picture to be analyzed for extracting the target user in library.
In the present embodiment, the picture to be analyzed for the target user that terminal is sent is the figure that terminal is uploaded from local photograph album
Piece.The user identifier of target user can be ID card No., the passport No. etc. of user.Predetermined database can be vehicle
Dangerous service database, banking business data, collage-credit data library etc..
S2, one or more human face recognition models trained in advance are called to identify the picture to be analyzed, to know
Not Chu preset kind face feature of risk.
For example, the face feature of risk of the preset kind can be the face feature of risk of 8 dimensions, advantage is to ensure foot
While enough accuracy of identification, guarantees the speed of operation, system operations resource is effectively reduced, in the present embodiment, people of 8 dimension
Face feature of risk comprehensive description multinomial face characteristic relevant with risk is driven, including and it is not limited to glasses, picture quality, mouth
The features such as lip thickness, eyes opening and closing degree, gender, age, and/or the shape of face.Wherein, shape of face feature includes: round face or long face,
Wide chin or narrow chin etc..
Technical research theoretical basis of the invention: pass through the analysis and research to a large amount of vehicle insurance Claims Resolution data, it was found that visitor
The characteristic rule of sight: the difference of the unstructured feature such as facial image feature, GPS track feature is reflected in the duty in car accident
Difference is appointed to seem highly significant.For example, the first research achievement is: human eye is when observing object, if having defective vision, can narrow as far as possible
Eye, the opening degree of eyes are usually less than normal value, and caused by many car accidents are the influence because of eyesight, especially can
In the lower situation of degree of opinion；Second of research achievement is: the severity that car accident occurs for the driver to wear glasses is generally big
In the driver not worn glasses, objective reason may include eyesight do not correct in place, eyesight continuous decrease replace eye not in time
Mirror, glasses, which are easy to be atomized, leads to short time blind view etc.；The third research achievement is: car accident occurs for the good driver of phychology
Probability be much smaller than the driver of phychology difference, and the people that phychology is good, some of face are generally characterized by diastole, natural；The
Four kinds of research achievements are: the probability that car accident occurs for female driver is generally greater than male driver；5th kind of research achievement
Be: the probability that car accident occurs for advanced age driver is generally greater than young and middle-aged driver.Therefore, the invention by face
Feature is introduced into the analysis for driving risk, accuracy that is objective, effectively improving driving risk analysis.
S3, user data corresponding with the target user is extracted from predetermined database, from the user
Preset kind structured features data are extracted in data.
Wherein, predetermined database can be vehicle insurance service database, banking business data, collage-credit data library etc..
The preset kind structured features data may include the structured features of 20 dimensions, the structured features of 20 dimension
It may include age, gender, occupation, annual income, native place, information of not repaying, credit card debt information, values information, sign
The dimensional characteristics such as letter information, financial fraud information, vehicle insurance Claims Resolution information.Such as wherein, values information includes: the compassion to society
Sight/optimism degree etc., the values information can be obtained in client's transacting business by allowing client to fill in evaluation questionnaire analysis
Out；Reference information includes: with the presence or absence of situation etc. of breaking one's promise, and the reference information can be from predetermined collage-credit data library " example
Such as, the personal collage-credit data library of People's Bank of China " obtains；Financial fraud information includes: whether once to carry out financial fraud；
Vehicle insurance Claims Resolution information includes: the number of generation car accident in preset time, great car accident occurs in preset time
Number etc..
S4, the face feature of risk of the preset kind and the preset kind structured features data are spliced into multidimensional
Integrated risk feature, and the driving risk analysis model that integrated risk feature input is trained in advance, with output integrated
Risk analysis value.
In the present embodiment, integrated risk feature is the integrated risk feature of 28 dimensions, including the 8 non-structured face wind of dimension
The structured features of dangerous feature and 20 dimensions.
The driving risk analysis model of training can be any suitable intelligent decision making model in advance, for example, GLM
(general linear model, general linear model), DNN (deep neural network, deep neural network),
GBDT (gradient boost descision tree, gradient promote decision tree) etc..
Integrated risk analysis value be it is any suitable, can it is objective represent drive risk preset kind mathematical value, for example,
Risk score value, vehicle insurance Claims Resolution rate etc..
The inertial thinking of those skilled in the art is normally based on age of driver, gender, former years loss ratio and goes out
The structured features such as dangerous situation condition carry out driving risk analysis, and especially in vehicle insurance business scope, this inertial thinking is even more root
The deep base of a fruit is solid, and the AI based on multidimensional data that the present embodiment proposes drives risk analysis method, has effectively broken this inertial thinking,
The creative objective connection for having found face characteristic and driving risk, it is objective in driving risk that maximization has played face characteristic
The effect for analyzing aspect, effectively prevents unilateral influence of the structured features data on model, greatly improved and drives risk point
The accuracy and objectivity of analysis.
Optionally, it is further comprising the steps of to drive risk analysis method by the AI based on multidimensional data:
If the integrated risk analysis value of output is greater than the first preset threshold, the first preset format is sent to the terminal
Drive Risk-warning information；And/or
If the integrated risk analysis value of output is less than or equal to the first preset threshold, and is greater than the second preset threshold, then
The driving Risk-warning information of the second preset format is sent to the terminal；And/or
If the integrated risk analysis value of output is less than or equal to the second preset threshold, it is pre- that third is sent to the terminal
If the result feedback information of format.
Wherein, the driving Risk-warning information of the first preset format can illustrate are as follows: this modal analysis results is that " height is driven
Sailing danger ", do be noted that；The driving Risk-warning information of second preset format can illustrate are as follows: this model analysis
The result is that " middle driving risk ", is please suitably noted that；The driving Risk-warning information of third preset format can illustrate are as follows: this
Secondary modal analysis results are " low driving risks ", please be known.First preset threshold, the second preset threshold, third predetermined threshold value can
It is adjusted according to actual needs.
Optionally, it is further comprising the steps of to drive risk analysis method by the AI based on multidimensional data:
According to the mapping relations of predetermined integrated risk analysis value range and preset kind vehicle insurance property parameters, determine
The corresponding preset kind vehicle insurance property parameters of the integrated risk analysis value, and the preset kind vehicle insurance property parameters are sent
To the terminal.
Wherein, preset kind vehicle insurance property parameters include: vehicle insurance risk class, vehicle insurance discount rate etc..
It is optionally, one in the present embodiment in order to improve accuracy, stability and the speed of face characteristic identification
Or multiple human face recognition models trained in advance include human face region identification model and face characteristic identification model, step S2 packet
It includes:
S21, human face region identification model trained in advance is called to identify user's picture to be analyzed, with identification
It human face region and is calibrated out；And
S22, face characteristic identification model trained in advance is called to carry out feature identification to the human face region identified,
To identify the face feature of risk of preset kind.
In the present embodiment, human face region identification model trained in advance is Multi-task Cascaded
Convolutional Networks, multitask depth convolutional neural networks model；Face characteristic identification model is
FacialRiskNet model；The face feature of risk of preset kind is the face feature of risk of 8 dimensions.
Optionally, the human face region identification model is (for example, Multi-task Cascaded Convolutional
Networks, multitask depth convolutional neural networks model) individually training, the training process packet of the human face region identification model
It includes:
Obtain user's picture sample of the preset quantity sample of users in preset time；For example, obtaining in nearest 1 year
User's picture sample of 1000000 sample of users；
Respectively each user's picture sample label human face region and the key point position coordinates of human face five-sense-organ；Example
Such as, human face region can be marked using square and/or rectangular pre-set color wire frame, the key point position of human face five-sense-organ is sat
Mark includes: two intraocular outer angle points, nose, place between the eyebrows point, 2 corners of the mouth points；
User's picture sample after label is divided into the verifying collection of the training set of the first ratio, the second ratio；For example, first
Ratio is 70%, and the second ratio is 30%；
It is trained to obtain using user's picture sample training human face region identification model in the training set
Human face region identification model verifies trained human face region identification model using user's picture sample that the verifying is concentrated
Accuracy rate；And
If accuracy rate is more than or equal to preset threshold, training terminates, alternatively, if accuracy rate is less than preset threshold,
Increase the quantity of user's picture sample, and re-executes above steps.
Optionally, the face characteristic identification model (for example, FacialRiskNet model) individually training, the face
The training process of feature identification model includes:
The human face region picture sample of the preset quantity sample of users in preset time is obtained, and is each human face region
Picture sample determines corresponding integrated risk analysis value, each face administrative division map piece sample and its corresponding integrated risk point
Analysis value constitutes the face characteristic sample of a face characteristic identification model；For example, 1,000,000 samples obtained in nearest 1 year are used
The human face region picture sample at family, each corresponding integrated risk analysis value of face administrative division map piece sample can be according to history vehicle
Danger Claims Resolution data are determined；
The face characteristic sample that all face administrative division map piece samples and its corresponding integrated risk analysis value are constituted is divided into
The verifying collection of the training set of first ratio, the second ratio；For example, the first ratio is 70%, the second ratio is 30%；
It is trained to obtain using face characteristic identification model described in the face characteristic sample training in the training set
Face characteristic identification model verifies trained face characteristic identification model using the face characteristic sample that the verifying is concentrated
Accuracy rate；
If accuracy rate is more than or equal to preset threshold, training terminates, alternatively, if accuracy rate is less than preset threshold,
Increase the quantity of the human face region picture sample, and re-executes above steps.
Optionally, the driving risk analysis model is individually trained, wherein the driving risk analysis model, which can be, appoints
The applicable intelligent decision making model of meaning, for example, GLM (general linear model, general linear model), DNN (deep
Neural network, deep neural network), (gradient boost descision tree, gradient promote decision to GBDT
Tree etc., training process includes:
Obtain user's picture sample of the preset quantity sample of users in preset time and the data sample of user data；Example
Such as, user's picture sample of 1,000,000 sample of users in nearest 1 year and the data sample of user data are obtained；
For each user's picture sample, one or more human face recognition models trained in advance is called to be identified,
To identify the face feature of risk of preset kind；Wherein, face feature of risk is the face feature of risk of 8 dimensions；
For each data sample, preset kind structured features data are extracted；For example, data sample i's is default
Type structure characteristic can be { Xi1, Xi2, Xi3 ..., Xim }, and this feature set includes m different types of spies
Sign, m can be equal to 20；
Respectively by the face feature of risk of the corresponding preset kind of each sample of users and preset kind structured features number
It according to the integrated risk feature for being spliced into multidimensional, and is the corresponding driving value-at-risk of each integrated risk signature；Wherein, more
The integrated risk feature of dimension is the integrated risk feature of 28 dimensions, the knot including 8 dimensions non-structured face feature of risk and 20 dimensions
Structure feature；
All integrated risk features of splicing are divided into the verifying collection of the training set of the first ratio, the second ratio；For example, the
One ratio is 70%, and the second ratio is 30%；
It is trained to obtain using the integrated risk feature training driving risk analysis model in the training set
Risk analysis model is driven, verifies trained driving risk analysis model using the key feature information that the verifying is concentrated
Accuracy rate；And
If accuracy rate is more than or equal to preset threshold, training terminates, alternatively, if accuracy rate is less than preset threshold,
Increase the quantity of user's picture sample and data sample, and re-executes above steps.
It is required when further, in order to which the driving risk analysis model and the training of face characteristic identification model is effectively reduced
Training samples number, guarantee face characteristic identification model, drive the training harmony of risk analysis model, precision consistency and
The uniformity of training effect is optionally melted for the face characteristic identification model and driving risk analysis model using end-to-end
Training is closed, the process of the end-to-end Fusion training includes:
Obtain the human face region picture sample of the preset quantity sample of users in preset time and the data sample of user data
This, and corresponding integrated risk analysis value is determined for each face administrative division map piece sample；For example, obtaining in nearest 1 year
The human face region picture sample of 1000000 sample of users, each corresponding integrated risk analysis value of face administrative division map piece sample
The data that can be settled a claim according to history vehicle insurance are determined；
For each data sample, preset kind structured features data are extracted, each sample of users is corresponding
Human face region picture sample, preset kind structured features data and integrated risk analysis value constitute an end-to-end Fusion training
Sample；For example, the preset kind structured features data of data sample i can be { Xi1, Xi2, Xi3 ..., Xim }, the spy
It includes m different types of features that collection, which is closed, and m can be equal to 20；
All sample of users corresponding end-to-end Fusion training samples is divided into the training set of the first ratio, the second ratio
Verifying collection；For example, the first ratio is 70%, the second ratio is 30%；
It is special that the corresponding human face region picture sample of each sample of users in the training set is inputted into the face respectively
Identification model is levied, respectively by the structured features data link of each output of face characteristic identification model and corresponding sample of users
To the input for driving risk analysis model, so that the driving risk analysis model and face characteristic identification model be constituted
The model of one entirety is trained, and realizes the Integral synchronous optimization for driving risk analysis model and face characteristic identification model；
It is special that the corresponding human face region picture sample of each sample of users in the training set is inputted into the face respectively
Identification model is levied, respectively by the structured features data link of each output of face characteristic identification model and corresponding sample of users
To the input for driving risk analysis model, so that the trained driving risk analysis model and face characteristic be identified
The model that model constitutes an entirety is verified；
If accuracy rate is more than or equal to preset threshold, end-to-end Fusion training terminates, alternatively, if accuracy rate is less than in advance
If threshold value, then increase the quantity of the human face region picture sample and data sample, and re-executes above steps.
In the present embodiment, the face characteristic identification model is constructed on Facial RiskNet network.The people
The structure of face feature identification model is as follows:
Wherein, Input size indicates the size of input picture, and block config indicates each layer of basic composition knot
Structure, Conv indicate that the convolutional layer of model, BN indicate batch standardization of model, and PRelu indicates that activation primitive, Linear indicate linear
Transformation, Kernel Size indicate that the size of convolution kernel is 3x3；Stride indicates the moving step length of convolution kernel, that is, finishes a secondary volume
The distance of next convolution position is moved to after product；Padding indicates the size to the image completion among current network layer,
The ResBlock structure operation schematic diagram of face characteristic identification model is referring to figure 2..
The creativeness of above-described embodiment is: it is directly right to joined a full articulamentum after FacialRiskNet network
It drives risk and is analyzed (detailed process is referring to figure 3.).Pass through the analysis end to end to risk is driven, realization pair
FacialRiskNet network model parameter realizes Automatic Optimal, and 8 dimensional features for exporting it have contained face and driven risk
Various features；Technical benefits using the model structure of above-mentioned face characteristic identification model are: 1) realize in face with drive
Nearly relevant key feature carries out fast and accurately automatic identification extraction for sailing；2) automatically extracting by feature of risk is realized
By key feature dimensionality reduction to 8 dimensions, computing resource is greatly reduced；3) output each dimension of feature is made to contain a variety of driving risks letters
Breath, so that the Fusion training of the unstructured feature of face risk and the structured features of user data is possibly realized and (please join
See Fig. 4 a and Fig. 4 b).
In order to further enhance the risk profile ability of model, optionally, the embodiment of the present invention is further drawn in a model
The structural data of access customer, wherein the structural data of user is made of discrete features and continuous feature.Face characteristic identification
Model and driving risk analysis model are individually trained, face characteristic identification model and the independent instruction for driving risk analysis model
Experienced step includes:
1) by the model structure training pattern of Fig. 3 signal, FacialRiskNet network parameter is obtained,
2) Fig. 4 a is pressed, extracts 8 dimension people from each user's face picture data by trained FacialRiskNet network
Face feature of risk；
3) Fig. 4 b is pressed, respectively by the structure of each output of trained FacialRiskNet network and corresponding sample of users
Change characteristic and is linked to input (that is: the described merging features for driving risk analysis model for driving risk analysis model
Layer input), in which: joined in the driving risk analysis model merging features layer (concat layers), effect be by
The continuous feature and discrete features of structuring and non-structured face characteristic carry out multidimensional characteristic anastomosing and splicing；Embedding
For embeding layer, FC is full articulamentum, and BN is standardization, and PReLU is activation primitive.
The problem of training process of model has individually been divided into multiple steps by Fig. 4 a and Fig. 4 b, is brought is independent by Fig. 3
Trained optimal model parameters cannot be guaranteed that the face feature of risk of FacialRiskNet network abstraction is also optimal in fig. 4b
, in order to solve this problem, the embodiment of the present invention it is creative to face characteristic identification model and drive risk analysis model
Using end-to-end Fusion training, face characteristic identification model and the step of drive the end-to-end Fusion training of risk analysis model
Include:
The model of the model of Fig. 3 signal and Fig. 4 b signal is merged and (refers to Fig. 5)；When training, family is mixed the sample with
Human face region picture input FacialRiskNet network, each output of FacialRiskNet network is directly linked to institute
State the input (that is: the input of the described merging features layer for driving risk analysis model) for driving risk analysis model；Directly by sample
The structured features data (continuous feature and discrete features) of this user are linked to the input for driving risk analysis model
(that is: the input of the described merging features layer for driving risk analysis model).
Through the above steps, can simultaneously to face characteristic identification model and drive risk analysis model all parameters into
The disposable end-to-end Fusion training optimization of row, effectively improves the accuracy rate of model, and reduce the consumption to system resource.
Second embodiment of the invention proposes a kind of Analysis server 1.Referring to shown in Fig. 6, mentioned for second embodiment of the invention
The schematic diagram of Analysis server 1 out.
In the present embodiment, Analysis server 1 can be rack-mount server, blade server, tower server or
Cabinet-type server.
The Analysis server 1 includes first memory 11, first processor 12 and first network interface 13.
Wherein, first memory 11 includes at least a type of readable storage medium storing program for executing, and the readable storage medium storing program for executing includes
Flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory etc.), magnetic storage, disk, CD etc..The
One memory 11 can be the internal storage unit of the Analysis server 1, such as the Analysis server 1 in some embodiments
Hard disk.First memory 11 is also possible to the External memory equipment of the Analysis server 1 in further embodiments, such as
The plug-in type hard disk being equipped on the Analysis server 1, intelligent memory card (Smart Media Card, SMC), secure digital
(Secure Digital, SD) card, flash card (Flash Card) etc..Further, first memory 11 can also both include
The internal storage unit of the Analysis server 1 also includes External memory equipment.
First memory 11 can be not only used for the application software and Various types of data that storage is installed on the Analysis server 1,
For example, first drives Risk analysis procedure 10 etc., can be also used for temporarily storing the data that has exported or will export.
First processor 12 can be in some embodiments a central processing unit (Central Processing Unit,
CPU), controller, microcontroller, microprocessor or other data processing chips store in first memory 11 for running
Program code or processing data, for example, first drives Risk analysis procedure 10 etc..
First network interface 13 optionally may include standard wireline interface and wireless interface (such as WI-FI interface), usually
For establishing communication connection between the Analysis server 1 and other electronic equipments.For example, Analysis server 1 can in advance really
Fixed terminal (not shown) carries out data transmission.
Fig. 6 illustrates only the Analysis server 1 with component 11-13, it will be appreciated by persons skilled in the art that Fig. 6
The structure shown does not constitute the restriction to Analysis server 1, may include than illustrating less perhaps more components or group
Close certain components or different component layouts.
Optionally, which can also include user interface, and user interface may include display
(Display), input unit such as keyboard (Keyboard), optional user interface can also include standard wireline interface,
Wireless interface.
Optionally, in some embodiments, display can be light-emitting diode display, liquid crystal display, touch control type LCD and show
Device and Organic Light Emitting Diode (Organic Light-Emitting Diode, OLED) touch device etc..Wherein, display
It is properly termed as display screen or display unit, for being shown in the information handled in Analysis server 1 and for showing visualization
User interface.
In 1 embodiment of Analysis server shown in Fig. 6, as being stored in a kind of memory 11 of computer storage medium
First drives the program code of Risk analysis procedure 10, and first processor 12 executes the first program for driving Risk analysis procedure 10
When code, realize that the AI in above-mentioned first embodiment based on multidimensional data such as drives the arbitrary steps of risk analysis method.
Optionally, the first driving Risk analysis procedure 10 can also be divided into one or more module, one or
Multiple modules are stored in first memory 11, and by one or more processors (the present embodiment is first processor 12) institute
It executes, to complete the present invention, the so-called module of the present invention is the series of computation machine program instruction for referring to complete specific function
Section.It is the first module diagram for driving Risk analysis procedure 10 in Fig. 6 for example, referring to shown in Fig. 7, in the embodiment, first
The first receiving module 110, identification module 120, the first extraction module 130, the can be divided by driving Risk analysis procedure 10
One analysis module 140 and the first warning module 150, the functions or operations step that the module 110-150 is realized is and above
Similar, and will not be described here in detail, illustratively, such as wherein:
First receiving module 110, the picture to be analyzed of the target user for receiving the terminal transmission, or receiving
After the driving risk analysis request for the target user that the terminal is sent, controls the terminal and shoot the target user
Photo as the picture to be analyzed, or according to it is described driving risk analysis request described in target user user mark
Know, the picture to be analyzed of the target user is extracted from predetermined database；
Identification module 120, for calling one or more human face recognition models trained in advance to the picture to be analyzed
It is identified, to identify the face feature of risk of preset kind；
First extraction module 130, for extracting use corresponding with the target user from predetermined database
User data extracts preset kind structured features data from the user data；
First analysis module 140, for by the face feature of risk of the preset kind and the preset kind structuring
Characteristic is spliced into the integrated risk feature of multidimensional, and the driving risk point that integrated risk feature input is trained in advance
Model is analysed, with output integrated risk analysis value；
First warning module 150, for sending preset format to predetermined terminal according to integrated risk analysis value
Drive Risk-warning information.
In addition, third embodiment of the invention proposes a kind of computer readable storage medium, the computer-readable storage medium
It include the first driving Risk analysis procedure 10 in matter, the first driving Risk analysis procedure 10 is realized such as when being executed by processor
Lower operation:
It receives the picture to be analyzed for the target user that the terminal is sent, or is receiving that the terminal sends for institute
After the driving risk analysis request for stating target user, controls the terminal and shoot the photo of the target user as described to be analyzed
Picture, or the user identifier of the target user according to the driving risk analysis request, from predetermined database
The middle picture to be analyzed for extracting the target user；
One or more human face recognition models trained in advance are called to identify the picture to be analyzed, to identify
The face feature of risk of preset kind；
User data corresponding with the target user is extracted from predetermined database, from the user data
In extract preset kind structured features data；And
The face feature of risk of the preset kind and the preset kind structured features data are spliced into multidimensional
Integrated risk feature, and the driving risk analysis model that integrated risk feature input is trained in advance, with output integrated wind
Dangerous assay value.
Multidimensional number is based in the specific embodiment of the computer readable storage medium of the present invention and above-mentioned first embodiment
According to AI drive risk analysis method specific embodiment it is roughly the same, details are not described herein.
Fourth embodiment of the invention proposes that a kind of AI based on multidimensional data drives risk analysis method.Referring to shown in Fig. 8,
For the flow chart for the AI driving risk analysis method based on multidimensional data that fourth embodiment of the invention proposes.
In the fourth embodiment, the AI driving risk analysis method based on multidimensional data is applied to Analysis server,
The Analysis server is connected with terminal, this method comprises: step A1-A3.
A1, the picture to be analyzed for receiving the target user that the terminal is sent, or in the needle for receiving the terminal transmission
After the driving risk analysis of target user request, control the terminal shoot the photo of the target user as it is described to
Analysis picture, or the user identifier of the target user according to the driving risk analysis request, from predetermined number
According to the picture to be analyzed for extracting the target user in library.
In the present embodiment, the user identifier of target user can be ID card No., the passport No. etc. of user.In advance really
Fixed database can be vehicle insurance service database, banking business data, collage-credit data library etc..
A2, user data corresponding with the target user is extracted from predetermined database, from the user
Preset kind structured features data are extracted in data.
Wherein, predetermined database can be vehicle insurance service database, banking business data, collage-credit data library etc..
The preset kind structured features data may include the structured features of 20 dimensions, the structured features of 20 dimension
It may include age, gender, occupation, annual income, native place, information of not repaying, credit card debt information, values information, sign
The dimensional characteristics such as letter information, financial fraud information, vehicle insurance Claims Resolution information.Such as wherein, values information includes: the compassion to society
Sight/optimism degree etc., the values information can be obtained in client's transacting business by allowing client to fill in evaluation questionnaire analysis
Out；Reference information includes: with the presence or absence of situation etc. of breaking one's promise, and the reference information can be from predetermined collage-credit data library " example
Such as, the personal collage-credit data library of People's Bank of China " obtains；Financial fraud information includes: whether once to carry out financial fraud；
Vehicle insurance Claims Resolution information includes: the number of generation car accident in preset time, great car accident occurs in preset time
Number etc..
A3, the picture to be analyzed and the preset kind structured features data are inputted into integrated risk trained in advance
In analysis model, output integrated risk analysis value.
The inertial thinking of those skilled in the art is normally based on age of driver, gender, former years loss ratio and goes out
The structured features such as dangerous situation condition carry out driving risk analysis, and especially in vehicle insurance business scope, this inertial thinking is even more root
The deep base of a fruit is solid, and the AI based on multidimensional data that the present embodiment proposes drives risk analysis method, has effectively broken this inertial thinking,
The creative objective connection for having found face characteristic and driving risk, it is objective in driving risk that maximization has played face characteristic
The effect for analyzing aspect, effectively prevents unilateral influence of the structured features data on model, greatly improved and drives risk point
The accuracy and objectivity of analysis.
Optionally, it is further comprising the steps of to drive risk analysis method by the AI based on multidimensional data:
If the integrated risk analysis value of output is greater than the first preset threshold, the first preset format is sent to the terminal
Drive Risk-warning information；And/or
If the integrated risk analysis value of output is less than or equal to the first preset threshold, and is greater than the second preset threshold, then
The driving Risk-warning information of the second preset format is sent to the terminal；And/or
If the integrated risk analysis value of output is less than or equal to the second preset threshold, it is pre- that third is sent to the terminal
If the result feedback information of format.
Wherein, the driving Risk-warning information of the first preset format can illustrate are as follows: this modal analysis results is that " height is driven
Sailing danger ", do be noted that；The driving Risk-warning information of second preset format can illustrate are as follows: this model analysis
The result is that " middle driving risk ", is please suitably noted that；The driving Risk-warning information of third preset format can illustrate are as follows: this
Secondary modal analysis results are " low driving risks ", please be known.First preset threshold, the second preset threshold, third predetermined threshold value can
It is adjusted according to actual needs.
Optionally, it is further comprising the steps of to drive risk analysis method by the AI based on multidimensional data:
According to the mapping relations of predetermined integrated risk analysis value range and preset kind vehicle insurance property parameters, determine
The corresponding preset kind vehicle insurance property parameters of the integrated risk analysis value, and the preset kind vehicle insurance property parameters are sent
To the terminal.
Wherein, preset kind vehicle insurance property parameters include: vehicle insurance risk class, vehicle insurance discount rate etc..
Optionally, the integrated risk analysis model include three submodels, be respectively the human face region identification model,
Face characteristic identification model and driving risk analysis model, the corresponding structured features data of target user and the input of user's picture
After the integrated risk analysis model, user's picture inputs the human face region identification model, the human face region
The output of identification model is linked to the input of the face characteristic identification model, the corresponding structured features data of target user and
The output of the face characteristic identification model is linked to the output for driving risk analysis model.
Optionally, the human face region identification model is (for example, Multi-task Cascaded Convolutional
Networks, multitask depth convolutional neural networks model) individually training, the training process packet of the human face region identification model
It includes:
Obtain user's picture sample of the preset quantity sample of users in preset time；For example, obtaining in nearest 1 year
User's picture sample of 1000000 sample of users；
Respectively each user's picture sample label human face region and the key point position coordinates of human face five-sense-organ；Example
Such as, human face region can be marked using square and/or rectangular pre-set color wire frame, the key point position of human face five-sense-organ is sat
Mark includes: two intraocular outer angle points, nose, place between the eyebrows point, 2 corners of the mouth points；
User's picture sample after label is divided into the verifying collection of the training set of the first ratio, the second ratio；For example, first
Ratio is 70%, and the second ratio is 30%；
It is trained to obtain using user's picture sample training human face region identification model in the training set
Human face region identification model verifies trained human face region identification model using user's picture sample that the verifying is concentrated
Accuracy rate；
If accuracy rate is more than or equal to preset threshold, training terminates, alternatively, if accuracy rate is less than preset threshold,
Increase the quantity of user's picture sample, and re-executes above steps.
Optionally, the face characteristic identification model (for example, FacialRiskNet model) individually training, the face
The training process of feature identification model includes:
The human face region picture sample of the preset quantity sample of users in preset time is obtained, and is each human face region
Picture sample determines corresponding integrated risk analysis value, each face administrative division map piece sample and its corresponding integrated risk point
Analysis value constitutes the face characteristic sample of a face characteristic identification model；For example, 1,000,000 samples obtained in nearest 1 year are used
The human face region picture sample at family, each corresponding integrated risk analysis value of face administrative division map piece sample can be according to history vehicle
Danger Claims Resolution data are determined；
The face characteristic sample that all face administrative division map piece samples and its corresponding integrated risk analysis value are constituted is divided into
The verifying collection of the training set of first ratio, the second ratio；For example, the first ratio is 70%, the second ratio is 30%；
It is trained to obtain using face characteristic identification model described in the face characteristic sample training in the training set
Face characteristic identification model verifies trained face characteristic identification model using the face characteristic sample that the verifying is concentrated
Accuracy rate；
If accuracy rate is more than or equal to preset threshold, training terminates, alternatively, if accuracy rate is less than preset threshold,
Increase the quantity of the human face region picture sample, and re-executes above steps.
Optionally, the driving risk analysis model is individually trained, wherein the driving risk analysis model, which can be, appoints
The applicable intelligent decision making model of meaning, for example, GLM (general linear model, general linear model), DNN (deep
Neural network, deep neural network), (gradient boost descision tree, gradient promote decision to GBDT
Tree etc., training process includes:
Obtain user's picture sample of the preset quantity sample of users in preset time and the data sample of user data；Example
Such as, user's picture sample of 1,000,000 sample of users in nearest 1 year and the data sample of user data are obtained；
For each user's picture sample, one or more human face recognition models trained in advance is called to be identified,
To identify the face feature of risk of preset kind；Wherein, face feature of risk is the face feature of risk of 8 dimensions；
For each data sample, preset kind structured features data are extracted；For example, data sample i's is default
Type structure characteristic can be { Xi1, Xi2, Xi3 ..., Xim }, and this feature set includes m different types of spies
Sign, m can be equal to 20；
Respectively by the face feature of risk of the corresponding preset kind of each sample of users and preset kind structured features number
It according to the integrated risk feature for being spliced into multidimensional, and is the corresponding driving value-at-risk of each integrated risk signature；Wherein, more
The integrated risk feature of dimension is the integrated risk feature of 28 dimensions, the knot including 8 dimensions non-structured face feature of risk and 20 dimensions
Structure feature；
All integrated risk features of splicing are divided into the verifying collection of the training set of the first ratio, the second ratio；For example, the
One ratio is 70%, and the second ratio is 30%；
It is trained to obtain using the integrated risk feature training driving risk analysis model in the training set
Risk analysis model is driven, verifies trained driving risk analysis model using the key feature information that the verifying is concentrated
Accuracy rate；And
If accuracy rate is more than or equal to preset threshold, training terminates, alternatively, if accuracy rate is less than preset threshold,
Increase the quantity of user's picture sample and data sample, and re-executes above steps.
It is required when further, in order to which the driving risk analysis model and the training of face characteristic identification model is effectively reduced
Training samples number, guarantee face characteristic identification model, drive the training harmony of risk analysis model, precision consistency and
The uniformity of training effect is optionally melted for the face characteristic identification model and driving risk analysis model using end-to-end
Training is closed, the process of the end-to-end Fusion training includes:
Obtain the human face region picture sample of the preset quantity sample of users in preset time and the data sample of user data
This, and corresponding integrated risk analysis value is determined for each face administrative division map piece sample；For example, obtaining in nearest 1 year
The human face region picture sample of 1000000 sample of users, each corresponding integrated risk analysis value of face administrative division map piece sample
The data that can be settled a claim according to history vehicle insurance are determined；
For each data sample, preset kind structured features data are extracted, each sample of users is corresponding
Human face region picture sample, preset kind structured features data and integrated risk analysis value constitute an end-to-end Fusion training
Sample；For example, the preset kind structured features data of data sample i can be { Xi1, Xi2, Xi3 ..., Xim }, the spy
It includes m different types of features that collection, which is closed, and m can be equal to 20；
All sample of users corresponding end-to-end Fusion training samples is divided into the training set of the first ratio, the second ratio
Verifying collection；For example, the first ratio is 70%, the second ratio is 30%；
It is special that the corresponding human face region picture sample of each sample of users in the training set is inputted into the face respectively
Identification model is levied, respectively by the structured features data link of each output of face characteristic identification model and corresponding sample of users
To the input for driving risk analysis model, so that the driving risk analysis model and face characteristic identification model be constituted
The model of one entirety is trained, to obtain the trained driving risk analysis model and face characteristic identification model；
It is special that the corresponding human face region picture sample of each sample of users in the training set is inputted into the face respectively
Identification model is levied, respectively by the structured features data link of each output of face characteristic identification model and corresponding sample of users
To the input for driving risk analysis model, so that the trained driving risk analysis model and face characteristic be identified
The model that model constitutes an entirety is verified；
If accuracy rate is more than or equal to preset threshold, end-to-end Fusion training terminates, alternatively, if accuracy rate is less than in advance
If threshold value, then increase the quantity of the human face region picture sample and data sample, and re-executes above steps.
Fifth embodiment of the invention proposes a kind of Analysis server 2.Referring to shown in Fig. 9, mentioned for fifth embodiment of the invention
The schematic diagram of Analysis server 2 out.
In the present embodiment, Analysis server 2 can be rack-mount server, blade server, tower server or
Cabinet-type server.
The Analysis server 2 includes second memory 21, second processor 22 and the second network interface 23.
Wherein, second memory 21 includes at least a type of readable storage medium storing program for executing, and the readable storage medium storing program for executing includes
Flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory etc.), magnetic storage, disk, CD etc..The
Two memories 21 can be the internal storage unit of the Analysis server 2, such as the Analysis server 2 in some embodiments
Hard disk.Second memory 21 is also possible to the External memory equipment of the Analysis server 2 in further embodiments, such as
The plug-in type hard disk being equipped on the Analysis server 2, intelligent memory card (Smart Media Card, SMC), secure digital
(Secure Digital, SD) card, flash card (Flash Card) etc..Further, second memory 21 can also both include
The internal storage unit of the Analysis server 2 also includes External memory equipment.
Second memory 21 can be not only used for the application software and Various types of data that storage is installed on the Analysis server 2,
For example, second drives Risk analysis procedure 20 etc., can be also used for temporarily storing the data that has exported or will export.
Second processor 22 can be in some embodiments a central processing unit (Central ProcessingUnit,
CPU), controller, microcontroller, microprocessor or other data processing chips store in second memory 21 for running
Program code or processing data, for example, second drives Risk analysis procedure 20 etc..
Second network interface 23 optionally may include standard wireline interface and wireless interface (such as WI-FI interface), usually
For establishing communication connection between the Analysis server 2 and other electronic equipments.For example, Analysis server 2 can in advance really
Fixed terminal (not shown) carries out data transmission.
Fig. 9 illustrates only the Analysis server 2 with component 21-23, it will be appreciated by persons skilled in the art that Fig. 9
The structure shown does not constitute the restriction to Analysis server 2, may include than illustrating less perhaps more components or group
Close certain components or different component layouts.
Optionally, which can also include user interface, and user interface may include display
(Display), input unit such as keyboard (Keyboard), optional user interface can also include standard wireline interface,
Wireless interface.
Optionally, in some embodiments, display can be light-emitting diode display, liquid crystal display, touch control type LCD and show
Device and Organic Light Emitting Diode (Organic Light-Emitting Diode, OLED) touch device etc..Wherein, display
It is properly termed as display screen or display unit, for being shown in the information handled in Analysis server 2 and for showing visualization
User interface.
In 2 embodiment of Analysis server shown in Fig. 9, as in a kind of second memory 21 of computer storage medium
Storage second drives the program code of Risk analysis procedure 20, and second processor 22 executes second and drives Risk analysis procedure 20
When program code, realize that the AI in above-mentioned fourth embodiment based on multidimensional data such as drives the arbitrary steps of risk analysis method.
Optionally, the second driving Risk analysis procedure 20 can also be divided into one or more module, one or
Multiple modules are stored in second memory 21, and by one or more processors (the present embodiment is second processor 22) institute
It executes, to complete the present invention, the so-called module of the present invention is the series of computation machine program instruction for referring to complete specific function
Section.It is the second module diagram for driving Risk analysis procedure 20 in Fig. 9 for example, referring to shown in Figure 10, in the embodiment, the
Two driving Risk analysis procedures 20 can be divided into the second receiving module 210, the second extraction module 220, the second analysis module
230 and second warning module 240, the functions or operations step that the module 210-240 is realized is similar as above, herein not
It is described in detail again, illustratively, such as wherein:
Second receiving module 210, the picture to be analyzed of the target user for receiving the terminal transmission, or receiving
After the driving risk analysis request for the target user that the terminal is sent, controls the terminal and shoot the target user
Photo as the picture to be analyzed, or according to it is described driving risk analysis request described in target user user mark
Know, the picture to be analyzed of the target user is extracted from predetermined database；
Second extraction module 220, for extracting use corresponding with the target user from predetermined database
User data extracts preset kind structured features data from the user data；
Second analysis module 230, for inputting the picture to be analyzed and the preset kind structured features data
In advance in trained integrated risk analysis model, output integrated risk analysis value；And
Second warning module 240, for sending preset format to predetermined terminal according to integrated risk analysis value
Drive Risk-warning information.
In addition, sixth embodiment of the invention proposes a kind of computer readable storage medium, the computer-readable storage medium
It include the second driving Risk analysis procedure 20 in matter, the second driving Risk analysis procedure 20 is realized such as when being executed by processor
Lower operation:
It receives the picture to be analyzed for the target user that the terminal is sent, or is receiving that the terminal sends for institute
After the driving risk analysis request for stating target user, controls the terminal and shoot the photo of the target user as described to be analyzed
Picture, or the user identifier of the target user according to the driving risk analysis request, from predetermined database
The middle picture to be analyzed for extracting the target user；
User data corresponding with the target user is extracted from predetermined database, from the user data
In extract preset kind structured features data；And
The picture to be analyzed and the preset kind structured features data are inputted into integrated risk trained in advance point
It analyses in model, output integrated risk analysis value.
Multidimensional number is based in the specific embodiment of the computer readable storage medium of the present invention and above-mentioned fourth embodiment
According to AI drive risk analysis method specific embodiment it is roughly the same, details are not described herein.
The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.
It should be noted that, in this document, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that the process, device, article or the method that include a series of elements not only include those elements, and
And further include other elements that are not explicitly listed, or further include for this process, device, article or method institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including being somebody's turn to do
There is also other identical elements in the process, device of element, article or method.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical solution of the present invention substantially in other words does the prior art
The part contributed out can be embodied in the form of software products, which is stored in one as described above
In storage medium (such as ROM/RAM, magnetic disk, CD), including some instructions are used so that terminal device (it can be mobile phone,
Computer, server or network equipment etc.) execute method described in each embodiment of the present invention.
The above is only a preferred embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (22)
1. a kind of AI based on multidimensional data drives risk analysis method, be applied to Analysis server, the Analysis server with
Terminal is connected, which is characterized in that the described method includes:
S1, the picture to be analyzed for receiving the target user that the terminal is sent, or receiving that the terminal sends for institute
After the driving risk analysis request for stating target user, controls the terminal and shoot the photo of the target user as described to be analyzed
Picture, or the user identifier of the target user according to the driving risk analysis request, from predetermined database
It is middle to extract the picture to be analyzed；
S2, one or more human face recognition models trained in advance are called to identify the picture to be analyzed, to identify
The face feature of risk of preset kind；
S3, user data corresponding with the target user is extracted from predetermined database, from the user data
In extract preset kind structured features data；And
S4, the face feature of risk of the preset kind and the preset kind structured features data are spliced into the comprehensive of multidimensional
Feature of risk, and the driving risk analysis model that integrated risk feature input is trained in advance are closed, with output integrated risk
Assay value.
2. the AI according to claim 1 based on multidimensional data drives risk analysis method, which is characterized in that this method is also
Include:
If the integrated risk analysis value of output is greater than the first preset threshold, the driving of the first preset format is sent to the terminal
Risk-warning information；And/or
If the integrated risk analysis value of output is less than or equal to the first preset threshold, and is greater than the second preset threshold, then to institute
State the driving Risk-warning information that terminal sends the second preset format；And/or
If the integrated risk analysis value of output is less than or equal to the second preset threshold, third is sent to the terminal and presets lattice
The result feedback information of formula.
3. the AI according to claim 1 based on multidimensional data drives risk analysis method, which is characterized in that this method is also
Include:
According to the mapping relations of predetermined integrated risk analysis value range and preset kind vehicle insurance property parameters, determine described in
The corresponding preset kind vehicle insurance property parameters of integrated risk analysis value, and the preset kind vehicle insurance property parameters are sent to institute
State terminal.
4. the AI according to claim 1 based on multidimensional data drives risk analysis method, which is characterized in that the driving
Risk analysis model includes a merging features layer, and this feature splicing layer is used to receive the face feature of risk of preset kind and pre-
If the link of type structure characteristic inputs, by the face feature of risk and preset kind knot of the preset kind of link input
Structure characteristic carries out multidimensional characteristic anastomosing and splicing.
5. the AI as claimed in any of claims 1 to 4 based on multidimensional data drives risk analysis method, feature
It is, one or more of human face recognition models trained in advance include human face region identification model and face characteristic identification mould
Type；
The step S2 includes:
S21, human face region identification model trained in advance is called to identify the picture to be analyzed, to identify face area
It is simultaneously calibrated in domain；And
S22, face characteristic identification model trained in advance is called to carry out feature identification to the human face region identified, to know
Not Chu preset kind face feature of risk.
6. the AI according to claim 5 based on multidimensional data drives risk analysis method, which is characterized in that the face
Region recognition model is individually trained, and the training step of the human face region identification model includes:
Obtain user's picture sample of the preset quantity sample of users in preset time；
Respectively each user's picture sample label human face region and the key point position coordinates of human face five-sense-organ；
User's picture sample after label is divided into the verifying collection of the training set of the first ratio, the second ratio；
Using user's picture sample training human face region identification model in the training set, to obtain trained face
Region recognition model verifies the accurate of trained human face region identification model using user's picture sample that the verifying is concentrated
Rate；And
If accuracy rate is more than or equal to preset threshold, training terminates, alternatively, increasing if accuracy rate is less than preset threshold
The quantity of user's picture sample, and re-execute above steps.
7. the AI according to claim 5 based on multidimensional data drives risk analysis method, which is characterized in that the face
Feature identification model is individually trained, and the training step of the face characteristic identification model includes:
The human face region picture sample of the preset quantity sample of users in preset time is obtained, and is each face administrative division map piece
Sample determines corresponding integrated risk analysis value, each face administrative division map piece sample and its corresponding integrated risk analysis value
Constitute the face characteristic sample of a face characteristic identification model；
The face characteristic sample that all face administrative division map piece samples and its corresponding integrated risk analysis value are constituted is divided into first
The verifying collection of the training set of ratio, the second ratio；
Using face characteristic identification model described in the face characteristic sample training in the training set, to obtain trained face
Feature identification model verifies the accurate of trained face characteristic identification model using the face characteristic sample that the verifying is concentrated
Rate；And
If accuracy rate is more than or equal to preset threshold, training terminates, alternatively, increasing if accuracy rate is less than preset threshold
The quantity of the human face region picture sample, and re-execute above steps.
8. the AI according to claim 5 based on multidimensional data drives risk analysis method, which is characterized in that the driving
Risk analysis model is individually trained, and the training step for driving risk analysis model includes:
Obtain user's picture sample of the preset quantity sample of users in preset time and the data sample of user data；
For each user's picture sample, one or more human face recognition models trained in advance are called to be identified, to know
Not Chu preset kind face feature of risk；
For each data sample, preset kind structured features data are extracted；
The face feature of risk of the corresponding preset kind of each sample of users and preset kind structured features data are spelled respectively
It is connected into the integrated risk feature of multidimensional, and is the corresponding driving value-at-risk of each integrated risk signature；
All integrated risk features of splicing are divided into the verifying collection of the training set of the first ratio, the second ratio；
Using the integrated risk feature training driving risk analysis model in the training set, to obtain trained driving
Risk analysis model verifies the accurate of trained driving risk analysis model using the key feature information that the verifying is concentrated
Rate；And
If accuracy rate is more than or equal to preset threshold, training terminates, alternatively, increasing if accuracy rate is less than preset threshold
The quantity of user's picture sample and data sample, and re-execute above steps.
9. the AI according to claim 5 based on multidimensional data drives risk analysis method, which is characterized in that be the people
It face feature identification model and drives risk analysis model and uses end-to-end Fusion training, the step of end-to-end Fusion training wraps
It includes:
The human face region picture sample of the preset quantity sample of users in preset time and the data sample of user data are obtained, and
Corresponding integrated risk analysis value is determined for each face administrative division map piece sample；
For each data sample, preset kind structured features data, the corresponding face of each sample of users are extracted
Region picture sample, preset kind structured features data and integrated risk analysis value constitute an end-to-end Fusion training sample
This；
The corresponding end-to-end Fusion training sample of all sample of users is divided into the verifying of the training set of the first ratio, the second ratio
Collection；
The corresponding human face region picture sample of each sample of users in the training set face characteristic is inputted respectively to know
Other model, respectively by the structured features data link of each output of face characteristic identification model and corresponding sample of users to institute
The input for driving risk analysis model is stated, so that the driving risk analysis model and face characteristic identification model are constituted one
Whole model is trained, and realizes the Integral synchronous optimization for driving risk analysis model and face characteristic identification model；
The corresponding human face region picture sample of each sample of users in the training set face characteristic is inputted respectively to know
Other model, respectively by the structured features data link of each output of face characteristic identification model and corresponding sample of users to institute
The input for driving risk analysis model is stated, thus by the trained driving risk analysis model and face characteristic identification model
The model for constituting an entirety is verified；And
If accuracy rate is more than or equal to preset threshold, end-to-end Fusion training terminates, alternatively, if accuracy rate is less than default threshold
Value, then increase the quantity of the human face region picture sample and data sample, and re-execute above steps.
10. a kind of Analysis server, which is characterized in that the Analysis server includes: memory, processor, on the memory
It is stored with can run on the processor first and drives Risk analysis procedure, described first drives Risk analysis procedure by institute
It states when processor executes, it can be achieved that the AI based on multidimensional data drives risk point as in one of claimed in any of claims 1 to 9
The step of analysis method.
11. a kind of computer readable storage medium, which is characterized in that include the first driving in the computer readable storage medium
Risk analysis procedure, when the first driving Risk analysis procedure is executed by processor, it can be achieved that as appointed in claim 1 to 9
The step of AI described in meaning one based on multidimensional data drives risk analysis method.
12. a kind of AI based on multidimensional data drives risk analysis method, be applied to Analysis server, the Analysis server with
Terminal is connected, which is characterized in that the described method includes:
A1, the picture to be analyzed for receiving the target user that the terminal is sent, or receiving that the terminal sends for institute
After the driving risk analysis request for stating target user, controls the terminal and shoot the photo of the target user as described to be analyzed
Picture, or the user identifier of the target user according to the driving risk analysis request, from predetermined database
The middle picture to be analyzed for extracting the target user；
A2, user data corresponding with the target user is extracted from predetermined database, from the user data
In extract preset kind structured features data；And
A3, the picture to be analyzed and the preset kind structured features data are inputted into integrated risk analysis trained in advance
In model, output integrated risk analysis value.
13. the AI according to claim 12 based on multidimensional data drives risk analysis method, which is characterized in that this method
Further include:
If the integrated risk analysis value of output is greater than the first preset threshold, the driving of the first preset format is sent to the terminal
Risk-warning information；And/or
If the integrated risk analysis value of output is less than or equal to the first preset threshold, and is greater than the second preset threshold, then to institute
State the driving Risk-warning information that terminal sends the second preset format；And/or
If the integrated risk analysis value of output is less than or equal to the second preset threshold, third is sent to the terminal and presets lattice
The result feedback information of formula.
14. the AI according to claim 12 based on multidimensional data drives risk analysis method, which is characterized in that this method
Further include:
According to the mapping relations of predetermined integrated risk analysis value range and preset kind vehicle insurance property parameters, determine described in
The corresponding preset kind vehicle insurance property parameters of integrated risk analysis value, and determining preset kind vehicle insurance property parameters are sent to
The terminal.
15. the AI described in any one of 2 to 14 based on multidimensional data drives risk analysis method according to claim 1,
It is characterized in that, the integrated risk analysis model includes: human face region identification model, face characteristic identification model and driving risk
Analysis model.
16. the AI according to claim 15 based on multidimensional data drives risk analysis method, which is characterized in that described to drive
Sail risk analysis model include a merging features layer, this feature splicing layer be used for receive preset kind face feature of risk and
The link of preset kind structured features data inputs, by the face feature of risk and preset kind of the preset kind of link input
Structured features data carry out multidimensional characteristic anastomosing and splicing.
17. the AI according to claim 15 based on multidimensional data drives risk analysis method, which is characterized in that the people
Face region recognition model is individually trained, and the training step of the human face region identification model includes:
Obtain user's picture sample of the preset quantity sample of users in preset time；
Respectively each user's picture sample label human face region and the key point position coordinates of human face five-sense-organ；
User's picture sample after label is divided into the verifying collection of the training set of the first ratio, the second ratio；
Using user's picture sample training human face region identification model in the training set, to obtain trained face
Region recognition model verifies the accurate of trained human face region identification model using user's picture sample that the verifying is concentrated
Rate；And
If accuracy rate is more than or equal to preset threshold, training terminates, alternatively, increasing if accuracy rate is less than preset threshold
The quantity of user's picture sample, and re-execute above steps.
18. the AI according to claim 15 based on multidimensional data drives risk analysis method, which is characterized in that the people
Face feature identification model is individually trained, and the training step of the face characteristic identification model includes:
The human face region picture sample of the preset quantity sample of users in preset time is obtained, and is each face administrative division map piece
Sample determines corresponding integrated risk analysis value, each face administrative division map piece sample and its corresponding integrated risk analysis value
Constitute the face characteristic sample of a face characteristic identification model；
The face characteristic sample that all face administrative division map piece samples and its corresponding integrated risk analysis value are constituted is divided into first
The verifying collection of the training set of ratio, the second ratio；
Using face characteristic identification model described in the face characteristic sample training in the training set, to obtain trained face
Feature identification model verifies the accurate of trained face characteristic identification model using the face characteristic sample that the verifying is concentrated
Rate；And
If accuracy rate is more than or equal to preset threshold, training terminates, alternatively, increasing if accuracy rate is less than preset threshold
The quantity of the human face region picture sample, and re-execute above steps.
19. the AI according to claim 15 based on multidimensional data drives risk analysis method, which is characterized in that described to drive
It sails risk analysis model individually to train, the training step for driving risk analysis model includes:
Obtain user's picture sample of the preset quantity sample of users in preset time and the data sample of user data；
For each user's picture sample, one or more human face recognition models trained in advance are called to be identified, to know
Not Chu preset kind face feature of risk, the human face recognition model include human face region identification model and face characteristic identification
Model；
For each data sample, preset kind structured features data are extracted；
The face feature of risk of the corresponding preset kind of each sample of users and preset kind structured features data are spelled respectively
It is connected into the integrated risk feature of multidimensional, and is the corresponding driving value-at-risk of each integrated risk signature；
All integrated risk features of splicing are divided into the verifying collection of the training set of the first ratio, the second ratio；
Using the integrated risk feature training driving risk analysis model in the training set, to obtain trained driving
Risk analysis model verifies the accurate of trained driving risk analysis model using the key feature information that the verifying is concentrated
Rate；And
If accuracy rate is more than or equal to preset threshold, training terminates, alternatively, increasing if accuracy rate is less than preset threshold
The quantity of user's picture sample and data sample, and re-execute above steps.
20. the AI according to claim 15 based on multidimensional data drives risk analysis method, which is characterized in that be described
It face characteristic identification model and drives risk analysis model and uses end-to-end Fusion training, the step of end-to-end Fusion training wraps
It includes:
The human face region picture sample of the preset quantity sample of users in preset time and the data sample of user data are obtained, and
Corresponding integrated risk analysis value is determined for each face administrative division map piece sample；
For each data sample, preset kind structured features data, the corresponding face of each sample of users are extracted
Region picture sample, preset kind structured features data and integrated risk analysis value constitute an end-to-end Fusion training sample
This；
The corresponding end-to-end Fusion training sample of all sample of users is divided into the verifying of the training set of the first ratio, the second ratio
Collection,
The corresponding human face region picture sample of each sample of users in the training set face characteristic is inputted respectively to know
Other model, respectively by the structured features data link of each output of face characteristic identification model and corresponding sample of users to institute
The input for driving risk analysis model is stated, so that the driving risk analysis model and face characteristic identification model are constituted one
Whole model is trained, to obtain the trained driving risk analysis model and face characteristic identification model；
The corresponding human face region picture sample of each sample of users in the training set face characteristic is inputted respectively to know
Other model, respectively by the structured features data link of each output of face characteristic identification model and corresponding sample of users to institute
The input for driving risk analysis model is stated, thus by the trained driving risk analysis model and face characteristic identification model
The model for constituting an entirety is verified；And
If accuracy rate is more than or equal to preset threshold, end-to-end Fusion training terminates, alternatively, if accuracy rate is less than default threshold
Value, then increase the quantity of the human face region picture sample and data sample, and re-execute above steps.
21. a kind of Analysis server, which is characterized in that the Analysis server includes: memory, processor, on the memory
It is stored with can run on the processor second and drives Risk analysis procedure, described second drives Risk analysis procedure by institute
It states when processor executes, it can be achieved that the AI based on multidimensional data as described in any one of claim 12 to 20 drives risk
The step of analysis method.
22. a kind of computer readable storage medium, which is characterized in that include the second driving in the computer readable storage medium
Risk analysis procedure, it can be achieved that as in claim 12 to 20 when the second driving Risk analysis procedure is executed by processor
The step of AI described in any one based on multidimensional data drives risk analysis method.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2019101010164 | 2019-01-31 | ||
| CN201910101016 | 2019-01-31 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110069988A true CN110069988A (en) | 2019-07-30 |
Family
ID=67366233
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910195210.3A Pending CN110069988A (en) | 2019-01-31 | 2019-03-14 | AI based on multidimensional data drives risk analysis method, server and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110069988A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110717377A (en) * | 2019-08-26 | 2020-01-21 | 平安科技（深圳）有限公司 | Face driving risk prediction model training and prediction method thereof and related equipment |
| CN110837783A (en) * | 2019-10-22 | 2020-02-25 | 恒大智慧科技有限公司 | Automatic warning method, device and storage medium for suspicious people coming from outside in intelligent community |
| CN111310562A (en) * | 2020-01-10 | 2020-06-19 | 中国平安财产保险股份有限公司 | Vehicle driving risk management and control method based on artificial intelligence and related equipment thereof |
| CN111414874A (en) * | 2020-03-26 | 2020-07-14 | 中国平安财产保险股份有限公司 | Driving risk prediction method, device and equipment based on human face and storage medium |
| CN111475532A (en) * | 2020-03-05 | 2020-07-31 | 拉扎斯网络科技（上海）有限公司 | Data processing optimization method and device, storage medium and terminal |
| CN111738415A (en) * | 2020-06-17 | 2020-10-02 | 北京字节跳动网络技术有限公司 | Model synchronous updating method and device and electronic equipment |
| CN112365107A (en) * | 2020-12-16 | 2021-02-12 | 北京易华录信息技术股份有限公司 | Myopia risk assessment method, device and system based on artificial intelligence |
| CN113221984A (en) * | 2021-04-29 | 2021-08-06 | 平安科技（深圳）有限公司 | Method, device and equipment for analyzing and predicting drunk driving behaviors of user and storage medium |
| CN115115843A (en) * | 2022-06-02 | 2022-09-27 | 马上消费金融股份有限公司 | Data processing method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060206724A1 (en) * | 2005-02-16 | 2006-09-14 | David Schaufele | Biometric-based systems and methods for identity verification |
| US20160189303A1 (en) * | 2014-03-21 | 2016-06-30 | Gil Emanuel Fuchs | Risk Based Automotive Insurance Rating System |
| CN107161104A (en) * | 2017-05-15 | 2017-09-15 | 成都中技智慧企业管理咨询有限公司 | A kind of vehicle drive assist system |
| CN108537198A (en) * | 2018-04-18 | 2018-09-14 | 济南浪潮高新科技投资发展有限公司 | A kind of analysis method of the driving habit based on artificial intelligence |
-
2019
- 2019-03-14 CN CN201910195210.3A patent/CN110069988A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060206724A1 (en) * | 2005-02-16 | 2006-09-14 | David Schaufele | Biometric-based systems and methods for identity verification |
| US20160189303A1 (en) * | 2014-03-21 | 2016-06-30 | Gil Emanuel Fuchs | Risk Based Automotive Insurance Rating System |
| CN107161104A (en) * | 2017-05-15 | 2017-09-15 | 成都中技智慧企业管理咨询有限公司 | A kind of vehicle drive assist system |
| CN108537198A (en) * | 2018-04-18 | 2018-09-14 | 济南浪潮高新科技投资发展有限公司 | A kind of analysis method of the driving habit based on artificial intelligence |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110717377A (en) * | 2019-08-26 | 2020-01-21 | 平安科技（深圳）有限公司 | Face driving risk prediction model training and prediction method thereof and related equipment |
| WO2021035983A1 (en) * | 2019-08-26 | 2021-03-04 | 平安科技（深圳）有限公司 | Method for training face-based driving risk prediction model, driving risk prediction method based on face, and related devices |
| CN110837783A (en) * | 2019-10-22 | 2020-02-25 | 恒大智慧科技有限公司 | Automatic warning method, device and storage medium for suspicious people coming from outside in intelligent community |
| CN111310562A (en) * | 2020-01-10 | 2020-06-19 | 中国平安财产保险股份有限公司 | Vehicle driving risk management and control method based on artificial intelligence and related equipment thereof |
| CN111475532B (en) * | 2020-03-05 | 2023-11-03 | 拉扎斯网络科技（上海）有限公司 | Data processing optimization method and device, storage medium and terminal |
| CN111475532A (en) * | 2020-03-05 | 2020-07-31 | 拉扎斯网络科技（上海）有限公司 | Data processing optimization method and device, storage medium and terminal |
| CN111414874B (en) * | 2020-03-26 | 2020-10-30 | 中国平安财产保险股份有限公司 | Driving risk prediction method, device and equipment based on human face and storage medium |
| CN111414874A (en) * | 2020-03-26 | 2020-07-14 | 中国平安财产保险股份有限公司 | Driving risk prediction method, device and equipment based on human face and storage medium |
| CN111738415A (en) * | 2020-06-17 | 2020-10-02 | 北京字节跳动网络技术有限公司 | Model synchronous updating method and device and electronic equipment |
| CN111738415B (en) * | 2020-06-17 | 2023-07-04 | 北京字节跳动网络技术有限公司 | Model synchronous updating method and device and electronic equipment |
| CN112365107A (en) * | 2020-12-16 | 2021-02-12 | 北京易华录信息技术股份有限公司 | Myopia risk assessment method, device and system based on artificial intelligence |
| CN112365107B (en) * | 2020-12-16 | 2024-01-23 | 北京易华录信息技术股份有限公司 | Myopia risk assessment method, device and system based on artificial intelligence |
| CN113221984A (en) * | 2021-04-29 | 2021-08-06 | 平安科技（深圳）有限公司 | Method, device and equipment for analyzing and predicting drunk driving behaviors of user and storage medium |
| CN115115843A (en) * | 2022-06-02 | 2022-09-27 | 马上消费金融股份有限公司 | Data processing method and device |
| CN115115843B (en) * | 2022-06-02 | 2023-08-22 | 马上消费金融股份有限公司 | Data processing method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110069988A (en) | AI based on multidimensional data drives risk analysis method, server and storage medium | |
| US20210358050A1 (en) | Using machine learning to classify insurance card information | |
| US11157830B2 (en) | Automated customized web portal template generation systems and methods | |
| CN108446697B (en) | Image processing method, electronic device and storage medium | |
| US20230401828A1 (en) | Method for training image recognition model, electronic device and storage medium | |
| CN110689459B (en) | Contract processing method, contract processing device, contract processing equipment and storage medium | |
| CN105095827A (en) | Facial expression recognition device and facial expression recognition method | |
| WO2021051611A1 (en) | Face visibility-based face recognition method, system, device, and storage medium | |
| CN106295591A (en) | Gender identification method based on facial image and device | |
| CN109858362A (en) | A kind of mobile terminal method for detecting human face based on inversion residual error structure and angle associated losses function | |
| CN110827236B (en) | Brain tissue layering method, device and computer equipment based on neural network | |
| CN105930771A (en) | Driving behavior grading method and device | |
| CN111898550B (en) | Expression recognition model building method and device, computer equipment and storage medium | |
| CN108509833A (en) | A kind of face identification method, device and equipment based on structured analysis dictionary | |
| CN109460725A (en) | Receipt consumption details content mergence and extracting method | |
| CN109409196A (en) | Personality prediction technique based on face, device, electronic equipment | |
| EP3869472A1 (en) | Detecting identification tampering using ultra-violet imaging | |
| CN108021863A (en) | Electronic device, the character classification by age method based on image and storage medium | |
| CN112784751A (en) | Training method, device, equipment and medium of image recognition model | |
| WO2023273259A1 (en) | Method and apparatus for generating direction recognition model, and device, medium and program product | |
| US20230063311A1 (en) | Information processing apparatus, information processing method, and program | |
| CN113869256A (en) | Driver fatigue detection method, system, device and storage medium | |
| CN113591603A (en) | Certificate verification method and device, electronic equipment and storage medium | |
| CN112634017A (en) | Remote card opening activation method and device, electronic equipment and computer storage medium | |
| CN109298783A (en) | Mark monitoring method, device and electronic equipment based on Expression Recognition |
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 |