CN108491821A - Vehicle insurance accident discrimination method, system and storage medium based on image procossing and deep learning - Google Patents
Vehicle insurance accident discrimination method, system and storage medium based on image procossing and deep learning Download PDFInfo
- Publication number
- CN108491821A CN108491821A CN201810283452.3A CN201810283452A CN108491821A CN 108491821 A CN108491821 A CN 108491821A CN 201810283452 A CN201810283452 A CN 201810283452A CN 108491821 A CN108491821 A CN 108491821A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- accident
- photo
- video
- danger
- 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
- 238000013135 deep learning Methods 0.000 title claims abstract description 55
- 238000012850 discrimination method Methods 0.000 title claims abstract description 31
- 238000003860 storage Methods 0.000 title claims abstract description 13
- 206010052428 Wound Diseases 0.000 claims abstract description 46
- 208000027418 Wounds and injury Diseases 0.000 claims abstract description 46
- 238000000034 method Methods 0.000 claims abstract description 24
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 238000011282 treatment Methods 0.000 claims abstract description 11
- 238000012545 processing Methods 0.000 claims description 16
- 238000012360 testing method Methods 0.000 claims description 13
- 230000027455 binding Effects 0.000 claims description 9
- 238000009739 binding Methods 0.000 claims description 9
- 230000002123 temporal effect Effects 0.000 claims description 8
- 238000013519 translation Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 5
- 238000012549 training Methods 0.000 description 18
- 238000010586 diagram Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 230000003190 augmentative effect Effects 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 210000004556 brain Anatomy 0.000 description 2
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000004069 differentiation Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000002265 prevention Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 240000007594 Oryza sativa Species 0.000 description 1
- 235000007164 Oryza sativa Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000001771 impaired effect Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000008450 motivation Effects 0.000 description 1
- 235000009566 rice Nutrition 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/18—Legal services; Handling legal documents
-
- G06Q50/40—
Abstract
The invention discloses a kind of vehicle insurance accident discrimination method, system and storage medium based on image procossing and deep learning, this approach includes the following steps：When being connected to vehicle insurance and reporting a case to the security authorities, the photo and/or video of be in danger field accident vehicle and associated vehicle are obtained；Photo and/or video to the field accident vehicle that is in danger carry out information pre-processing, obtain the relevant information of accident vehicle；If relevant information meets preset condition, photo to be in danger field accident vehicle and associated vehicle and/or video carry out image procossing, obtain accident vehicle and the associated vehicle surface of a wound treated photo and/or video；Using deep learning method by after accident vehicle and associated vehicle Wound treatment photo and/or video match；If matching result reaches preset value, judges that this accident is bumped against by accident vehicle and associated vehicle and generate.Invention achieves a kind of efficient, quick, intelligentized vehicle insurance field antifraud effects.
Description
Technical field
The present invention relates to car accident processing technology field more particularly to a kind of vehicles based on image procossing and deep learning
Dangerous accident discrimination method, system and storage medium.
Background technology
Currently, vehicle insurance fraud case is just showing the features such as modus operandi is various, concealment is strong, intelligent, group is formd
The development trends such as body, specialization and professionalism, vehicle insurance fraud seriously affected insurance company's own service benign development and
The level of profitability, while the business risk of insurance company is also increased, finally compromise the legitimate rights and interests of numerous insurers.For vehicle
The anti-fraud in danger, substantially takes including the multidisciplinary joint coordination such as public security, traffic police, insurance company, insurance association, establishes vehicle insurance and take advantage of
Prevention system is cheated, the means such as strick precaution, investigation, evidence obtaining, prevention and strike are passed through, it is established that by all kinds of means, multi-faceted, multi-angle anti-
Insurance Fraud network promotes the case-solving rate of fraud case to contain spreading unchecked for vehicle insurance case of victimization.
Method is taken precautions against in insurance fraud during settling a claim for current vehicle insurance, has its reasonability and validity, but also face manpower
The shortcomings such as cost is excessive, Sectors cooperation is improper, efficiency is unsatisfactory.
Invention content
It is a primary object of the present invention to propose a kind of vehicle insurance accident discrimination method based on image procossing and deep learning,
System and storage medium, it is intended to improve the efficiency and accuracy of vehicle insurance accident fraud discriminating, reduce human cost.
To achieve the above object, the present invention provides a kind of vehicle insurance accident discriminating side using image procossing and deep learning
Method the described method comprises the following steps：
When being connected to vehicle insurance and reporting a case to the security authorities, the photo and/or video of be in danger field accident vehicle and associated vehicle are obtained；
Photo and/or video to the field accident vehicle that is in danger carry out information pre-processing, obtain the phase of accident vehicle
Close information；
If the relevant information meets preset condition, to the photo of be in danger field accident vehicle and the associated vehicle
And/or video carries out image procossing, obtains the accident vehicle and the associated vehicle surface of a wound treated photo and/or video；
Using deep learning method by after the accident vehicle and associated vehicle Wound treatment photo and/or video carry out
Matching；
If matching result reaches preset value, judges that this accident is bumped against by the accident vehicle and associated vehicle and generate.
Preferably, if the relevant information meets preset condition, to field accident vehicle and the associated vehicle of being in danger
Photo and/or video carry out image procossing the step of include：
If the relevant information meets preset condition, to the photo of be in danger field accident vehicle and the associated vehicle
And/or video carries out data prediction, obtains the photograph of be in danger field accident vehicle and the associated vehicle after data prediction
Piece and/or video, the data prediction include defogging, go the one or several kinds in reflective, rotation or translation；
It is described using deep learning method by the photo and/or video after the accident vehicle and associated vehicle Wound treatment
The step of being matched include：
The photo and/or video of be in danger described in after the data prediction field accident vehicle and associated vehicle are put into
Classification processing is carried out in the sorter network model trained, identifies the damaged part of the accident vehicle and associated vehicle；
By the field accident vehicle that is in danger described in after the damaged part of the accident vehicle and associated vehicle, data prediction
Input with the photo and/or video of associated vehicle as the surface of a wound Matching Model trained, obtains the accident vehicle and phase
It cut-offs a matching degree for surface of a wound photo.
Preferably, the relevant information of the accident vehicle includes：The number-plate number information of accident vehicle；
The photo and/or video to the field accident vehicle that is in danger carries out information pre-processing, obtains accident vehicle
Relevant information the step of after include：
According to the vehicle license plate number information with personnel's ID bindings of reporting a case to the security authorities in personnel's ID inquiry databases of reporting a case to the security authorities；
By the vehicle with the vehicle license plate number information and the accident vehicle of personnel's ID bindings of reporting a case to the security authorities in the database
Board number information compared to pair；
If comparison result is matching, executes the photo to be in danger field accident vehicle and the associated vehicle and/or regard
Frequency carries out the step of data prediction；
If comparison result is to mismatch, in background management system, prompt operating personnel carry out manpower intervention processing.
Preferably, the relevant information of the accident vehicle further includes：Shooting is in danger field accident vehicle and associated vehicle
The temporal information and location information of photo and/or video；
The photo and/or video to the field accident vehicle that is in danger carries out information pre-processing, obtains accident vehicle
Relevant information the step of after further include：
The be in danger photo of field accident vehicle and associated vehicle and/or the temporal information of video and place letter are shot by described
Breath, compared with obtaining the temporal information and location information of photo and/or video of be in danger field accident vehicle and associated vehicle pair；
If comparison result is matching, and in the database with the vehicle license plate number information of the personnel ID binding of reporting a case to the security authorities with
The number-plate number information of the accident vehicle compare to comparison result be matching, then execute to the field accident vehicle that is in danger
The step of data prediction being carried out with the photo and/or video of associated vehicle；
If comparison result is to mismatch, in background management system, prompt operating personnel carry out manpower intervention processing.
Preferably, when being connected to vehicle insurance and reporting a case to the security authorities, by field accident vehicle and the associated vehicle of being in danger described in personnel's shooting of reporting a case to the security authorities
Photo and/or video,
It is described when being connected to vehicle insurance and reporting a case to the security authorities, obtain the photo and/or video of be in danger field accident vehicle and associated vehicle
Step includes：
In the photo and/or video of be in danger field accident vehicle and the associated vehicle of the personnel that report a case to the security authorities described in identification shooting
Accident vehicle the number-plate number；
If identification is unsuccessful, the personnel that report a case to the security authorities described in prompt re-shoot the photograph of be in danger field accident vehicle and associated vehicle
Piece and/or video.
Preferably, described when being connected to vehicle insurance and reporting a case to the security authorities, obtain be in danger field accident vehicle and associated vehicle photo and/or
The step of video further includes：
In the photo and/or video of be in danger field accident vehicle and the associated vehicle of the personnel that report a case to the security authorities described in detection shooting
Whether there is the profile of the accident vehicle and associated vehicle；
If testing result is in danger for described in only one in the photo and/or video of field accident vehicle and associated vehicle
Vehicle, the then personnel that report a case to the security authorities described in prompt re-shoot the photo and/or video of be in danger field accident vehicle and associated vehicle.
Preferably, described when being connected to vehicle insurance and reporting a case to the security authorities, obtain be in danger field accident vehicle and associated vehicle photo and/or
The step of video further includes：
Obtain the distance between interface surface of a wound position and the camera of the accident vehicle and associated vehicle；
If the surface of a wound of the accident vehicle and associated vehicle is completely covered in coverage, and the distance is more than preset value,
The personnel that then report a case to the security authorities described in prompt re-shoot.
Preferably, described to obtain between the accident vehicle and the interface surface of a wound position and camera position of associated vehicle
The calculation formula of distance is：
D=√ [(x1-x2) ^2+ (y1-y2) ^2+ (z1-z2) ^2]；
Wherein, the interface surface of a wound position of the accident vehicle and associated vehicle is A (x1, y1, z1), the camera position
For B (x2, y2, z2).
To achieve the above object, the present invention also proposes that a kind of differentiate based on the vehicle insurance accident of image procossing and deep learning is
System, the system comprises：Memory, processor and be stored on the memory and can run on the processor based on
The vehicle insurance accident evaluator of image procossing and deep learning, it is described to be differentiated based on the vehicle insurance accident of image procossing and deep learning
The vehicle insurance accident discrimination method based on image procossing and deep learning as described above is realized when program is executed by the processor
The step of.
To achieve the above object, the present invention also proposes a kind of computer readable storage medium, the computer-readable storage
The vehicle insurance accident evaluator based on image procossing and deep learning is stored on medium, it is described to be based on image procossing and depth
The vehicle insurance based on image procossing and deep learning as described above is realized when the vehicle insurance accident evaluator of habit is executed by processor
The step of accident discrimination method.
The present invention is based on the beneficial effects of image procossing and the vehicle insurance accident discrimination method of deep learning, system and storage medium
Fruit is：The present invention obtains be in danger field accident vehicle and associated vehicle through the above technical solutions, when being connected to vehicle insurance and reporting a case to the security authorities
Photo and/or video；Photo and/or video to the field accident vehicle that is in danger carry out information pre-processing, obtain accident vehicle
Relevant information；If the relevant information meets preset condition, to be in danger field accident vehicle and the associated vehicle
Photo and/or video carry out image procossing, obtain the accident vehicle and the associated vehicle surface of a wound treated photo and/or regard
Frequently；Using deep learning method by after the accident vehicle and associated vehicle Wound treatment photo and/or video match；
If matching result reaches preset value, judges that this accident is bumped against by the accident vehicle and associated vehicle and generate, improve vehicle
The efficiency and accuracy that dangerous accident fraud differentiates, reduce human cost.
Description of the drawings
Fig. 1 is that the present invention is based on the flows of the vehicle insurance accident discrimination method preferred embodiment of image procossing and deep learning to show
It is intended to；
Fig. 2 is that the present invention is based on training nets in the vehicle insurance accident discrimination method preferred embodiment of image procossing and deep learning
The schematic diagram of network model；
Fig. 3 is that the present invention is based on photograph videos before reporting a case to the security authorities in image procossing and the vehicle insurance accident discrimination method of deep learning to believe
Cease the related procedure schematic diagram of acquisition；
Fig. 4 is that the present invention is based on photograph video information in image procossing and the vehicle insurance accident discrimination method of deep learning to locate in advance
Manage related procedure schematic diagram；
Fig. 5 is that the present invention is based on two licence piece video matchings of image procossing and the vehicle insurance accident discrimination method pair of deep learning
The related procedure schematic diagram of the differentiation of degree.
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 implementation mode
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.
Primary solutions the present invention is based on image procossing and the vehicle insurance accident discrimination method embodiment of deep learning are：
When being connected to vehicle insurance and reporting a case to the security authorities, the photo and/or video of be in danger field accident vehicle and associated vehicle are obtained；To the scene of being in danger
The photo and/or video of accident vehicle carry out information pre-processing, obtain the relevant information of accident vehicle；If the relevant information symbol
Preset condition is closed, then to photo and/or video the progress image procossing of be in danger field accident vehicle and the associated vehicle, is obtained
The accident vehicle and the associated vehicle surface of a wound treated photo and/or video；Using deep learning method by the accident
Photo and/or video after vehicle and associated vehicle Wound treatment are matched；If matching result reaches preset value, this is judged
Secondary accident is bumped against by the accident vehicle and associated vehicle to be generated.
Efficiency and accuracy that vehicle insurance accident fraud differentiates are improved as a result, reduce human cost.
Specifically, Fig. 1 is please referred to, Fig. 1 is the vehicle insurance accident discrimination method the present invention is based on image procossing and deep learning
The flow diagram of preferred embodiment.
As shown in Figure 1, the present invention is based on the vehicle insurance accident discrimination methods of image procossing and deep learning to include the following steps：
Step S10 obtains the photo of be in danger field accident vehicle and associated vehicle and/or regards when being connected to vehicle insurance and reporting a case to the security authorities
Frequently.
It is understood that in this step, when being connected to vehicle insurance and reporting a case to the security authorities, by the field accident that is in danger described in personnel's shooting of reporting a case to the security authorities
The photo and/or video of vehicle and associated vehicle.Wherein, it is described report a case to the security authorities personnel can use be equipped with vehicle insurance accident differentiate APP,
And with camera mobile phone or other intelligent terminals shot.
After the personnel that report a case to the security authorities take the photo and/or video of be in danger field accident vehicle and the associated vehicle, meeting
It is compressed locally, is then transmitted to cloud server again and is stored.
As an implementation, above-mentioned steps S10 may comprise steps of：
Step S101, the photo of be in danger field accident vehicle and the associated vehicle of personnel's shooting of reporting a case to the security authorities described in identification
And/or the number-plate number of the accident vehicle in video.
Step S102, if identification is unsuccessful, the personnel that report a case to the security authorities described in prompt re-shoot be in danger field accident vehicle and phase
The photo and/or video to cut-off.
It can also include the following steps as a kind of embodiment, above-mentioned steps S10 again：
Step S103, the photo of be in danger field accident vehicle and the associated vehicle of personnel's shooting of reporting a case to the security authorities described in detection
And/or whether there is the profile of the accident vehicle and associated vehicle in video.
Step S104, if testing result is in danger for described in the photo and/or video of field accident vehicle and associated vehicle
An only vehicle, the then personnel that report a case to the security authorities described in prompt re-shoot the photo of be in danger field accident vehicle and associated vehicle and/or regard
Frequently.
It can also include the following steps as a kind of embodiment, above-mentioned steps S10 again：
Step S105 obtains the distance between interface surface of a wound position and camera of the accident vehicle and associated vehicle.
Step S106, if the surface of a wound of the accident vehicle and associated vehicle is completely covered in coverage, and the distance is big
In preset value, then the personnel that report a case to the security authorities described in prompt re-shoot.
Wherein, it is described obtain between the accident vehicle and the interface surface of a wound position and camera position of associated vehicle away from
From calculation formula be：
D=√ [(x1-x2) ^2+ (y1-y2) ^2+ (z1-z2) ^2].
Wherein, the interface surface of a wound position of the accident vehicle and associated vehicle is A (x1, y1, z1), the camera position
For B (x2, y2, z2).
Specifically, the personnel that report a case to the security authorities need to carry out closely the interface surface of a wound of the accident vehicle and associated vehicle with mobile phone
Distance shooting.When it is implemented, centered on the mobile phone screen page, two-dimensional coordinate is first obtained, is passed to SceneKit, is used
After ARKIT detects object, three-dimensional coordinate is obtained, then pass through above-mentioned formula D=√ [(x1-x2) ^2+ (y1-y2) ^2+ (z1-
Z2) ^2] position of object distance mobile phone camera real world can be obtained.Guarantee to the accident vehicle with mutually cut-off
The surface of a wound position covering it is complete under the premise of, if obtained distance D be more than preset value (such as 50 centimetres), prompt described in
The personnel that report a case to the security authorities adjust mobile phone operation and are re-shoot to appropriate location.
It should be noted that ARKit, that is, AR (Augmented Reality), i.e. augmented reality, that is, in camera
A virtual 3D model is shown in the image of the real world captured.ARKit is a series of Software Development Tools of apple
Title, it is that iOS device develops augmented reality application to enable developer.In iOS11 systems, ARKit formally becomes iOS system frame
The iOS device user of frame, configuration apple A9 or A10 chip can be applied using ARKit.
SceneKit：This is the frame that apple one can be used for building 3D scenes, it is established on the basis of OpenGL,
Such as illumination, model, material, the advanced engine characteristics of video camera are contained, and can be with Core Image, Core
The existing graphics frame such as Animation, Sprite Kit is formed integral with one another and cooperates, due to its powerful ease for use and versatility
So that this 3D frame also has significant progress potentiality in non-gaming application.
Furthermore, it is necessary to explanation, every intelligent terminal with ARKit, SceneKit function is all in the guarantor of the present invention
Within the scope of shield, and it is not limited to apple intelligent terminal.
Step S20, photo and/or video to the field accident vehicle that is in danger carry out information pre-processing, obtain accident
The relevant information of vehicle.
As an implementation, the relevant information of the accident vehicle for example may include the number-plate number of accident vehicle
Information shoots be in danger field accident vehicle and the photo of associated vehicle and/or the temporal information and location information of video.
Step S30, if the relevant information meets preset condition, to field accident vehicle and the associated vehicle of being in danger
Photo and/or video carry out image procossing, obtain the accident vehicle and the associated vehicle surface of a wound treated photo and/or
Video.
If specifically, the relevant information meets preset condition, to field accident vehicle and the associated vehicle of being in danger
Photo and/or video carry out data prediction, obtain after data prediction it is described be in danger field accident vehicle with mutually cut-off
Photo and/or video, the data prediction includes defogging, it is reflective to go, a kind of or several in rotation or translation
Kind.
It is understood that image procossing is analyzed image with computer, to reach the technology of required result.Figure
As processing refers generally to Digital Image Processing.Digital picture refers to equipment such as industrial camera, video camera, scanners by shooting
The big two-dimensional array arrived, the element of the array are known as pixel, and value is known as gray value.Image processing techniques generally comprises
Compression of images, enhancing and recovery, three matching, description and identification parts.The common method of image procossing has image transformation, image
Coding compression, image enhancement and recovery, image segmentation, iamge description and image classification (identification).
As an implementation, above-mentioned steps S20, photo and/or video to the field accident vehicle that is in danger into
Row information pre-processes, and includes after the step of obtaining the relevant information of accident vehicle：
Step S201, according to the vehicle license plate number with personnel's ID bindings of reporting a case to the security authorities in personnel's ID inquiry databases of reporting a case to the security authorities
Information.
Step S202, by the database with the vehicle license plate number information of the personnel ID binding of reporting a case to the security authorities and the accident
The number-plate number information of vehicle compared to pair.
Step S203 executes the photograph to be in danger field accident vehicle and the associated vehicle if comparison result is matching
The step of piece and/or video carry out data prediction；
If comparison result is to mismatch, in background management system, prompt operating personnel carry out manpower intervention processing.
As another embodiment, above-mentioned steps S20, to the photo and/or video of the field accident vehicle that is in danger
Further include after the step of carrying out information pre-processing, obtaining the relevant information of accident vehicle：
Step S204 believes the time of the photo for shooting be in danger field accident vehicle and associated vehicle and/or video
Breath and location information are in danger field accident vehicle and the photo of associated vehicle and/or temporal information and the place of video with obtaining
Information compared to pair.
Step S205, if comparison result is matching, and the vehicle vehicle in the database with personnel's ID bindings of reporting a case to the security authorities
Trade mark information compared with the number-plate number information of the accident vehicle to comparison result be matching, then execute to it is described be in danger it is existing
The photo and/or video of field accident vehicle and associated vehicle carry out the step of data prediction；
If comparison result is to mismatch, in background management system, prompt operating personnel carry out manpower intervention processing.
Step S40, using deep learning method by after the accident vehicle and associated vehicle Wound treatment photo and/or
Video is matched.
As an implementation, when it is implemented, the step step S40 may comprise steps of：
Step S401, photo to be in danger described in after the data prediction field accident vehicle and associated vehicle and/
Or video is put into the sorter network model trained and carries out classification processing, identifies the impaired of the accident vehicle and associated vehicle
Position.
Wherein, as shown in Fig. 2, specifically training network model uses ResNet-50 depth convolutional networks, specific steps packet
Include training stage and test phase：
Training stage：Data set includes about 100,000 images (9.5 ten thousand training images, 5000 authentication images), training
ResNet-50 depth convolutional networks, loss layers use Softmax.
Test phase：A kind of test pictures are inputted, the corresponding damaged part of this pictures is correctly provided.
Step S402, it is existing by being in danger described in after the damaged part of the accident vehicle and associated vehicle, data prediction
Input of the photo and/or video of field accident vehicle and associated vehicle as the surface of a wound Matching Model trained, obtains the thing
Therefore the matching degree of vehicle and associated vehicle surface of a wound photo.
Referring once again to Fig. 2, it includes instruction that specific training pattern, which uses ResNet-50 depth convolutional networks, specific steps,
Practice stage and test phase：
Training stage：Data set includes about 50,000 pairs of images (4.75 ten thousand pairs of training images, 2500 pairs of authentication images), training
ResNet-50 depth convolutional networks, loss layers use Euclidean distance.
Test phase：A pair of of test pictures of input, provide this matching degree to picture, and matching degree is higher than setting value then result
Correctly.
It is understood that the concept of deep learning is derived from the research of artificial neural network, by Hinton et al. in 2006
It proposes in year, is a kind of based on the method for carrying out representative learning to data in machine learning, is that one in machine learning research is new
Field, motivation is to establish, simulation human brain carries out the neural network of analytic learning, and the mechanism that it imitates human brain explains number
According to, such as image, sound and text.Point of depth machine learning method also supervised learning and unsupervised learning.Different
It is very different to practise the learning model established under frame.
Step S50 judges this accident by the accident vehicle and associated vehicle phase if matching result reaches preset value
Hit generation.
If matching result reaches preset value, judges that this accident is bumped against by the accident vehicle and associated vehicle and generates,
That is, this, which is reported a case to the security authorities, belongs to true accident, fraud is not present.
In conclusion the present embodiment through the above technical solutions, acquired in the scene of the accident using the personnel that report a case to the security authorities photo, regard
The data files such as frequency are extracted and are compared to elements such as position, times in picture, video, and pass through image procossing, people
The technological means such as work intelligence, analyze the surface of a wound of accident vehicle and associated vehicle, to determine whether vehicle insurance fraud,
It has been finally reached a kind of efficient, quick, intelligentized vehicle insurance field antifraud effect.
Below to the present invention is based on the vehicle insurance accident discrimination methods of image procossing and deep learning to be further elaborated：
1) picture video information collection before reporting a case to the security authorities
At the scene of being in danger, car owner opens the relevant APP of the present invention, starts flow of reporting a case to the security authorities.In gatherer process, APP can draw
It leads user suitably to be operated, and picture to acquisition or video data carry out local pretreatment, relative points are as follows：
A) for the video of typing, its number-plate number can be extracted and be identified.If identification is unsuccessful, can give notice
To user interface layer, car owner is prompted to re-start video typing.
B whether) for the video of typing, can check wherein has the profile of two cars.If it find that wherein only occurring one
Vehicle then can also give notice to user interface layer, car owner is prompted to re-start video typing.
C) car owner needs the interface surface of a wound with two vehicle of mobile phone pair to carry out shooting at close range.With mobile phone screen page center, first
Two-dimensional coordinate is obtained, SceneKit is passed to, after detecting object using ARKit, gets three-dimensional coordinate, passes through following calculating public affairs
Formula can obtain the position of object distance camera real world：
D=√ [(x1-x2) ^2+ (y1-y2) ^2+ (z1-z2) ^2]
Wherein interface surface of a wound position is A (x1, y1, z1), and mobile phone camera real time position is B (x2, y2, z2).
Under the premise of guaranteeing complete to the covering of surface of a wound range, if obtained distance D is more than (such as 50 lis of predetermined value
Rice), then car owner can be prompted to adjust mobile phone operation and shot to correct position.
D after) video and picture data all acquire completion, it can be compressed locally, then be transmitted to cloud service again
Device is stored.
Related procedure is as shown in Figure 3.
2) server photograph video information pre-processing
Car owner acquires the photo and video of accident by APP, uploads to cloud server, and locate in advance into row information to it
Reason；Cloud service program can inquire the relevant information (such as number-plate number) in database according to the car owner ID that reports a case to the security authorities simultaneously, with photograph
The number-plate number identified in piece/video is compared, and if there is mismatch case, then can be prompted in background management system
Operator carries out manpower intervention processing.
Later, cloud service program can extract its shooting time and location information from picture data, then with Cloud Server
Uplink time and location information when the car owner's upload pictures/video data received compare, if in time or place
It is less than setting value with degree, then operator can be proposed to warn in background management system, it is reminded to do more artificial treatments.
Related procedure is as shown in Figure 4.
3) to the differentiation of two vehicle surface of a wound picture match degree
The correlation step of two vehicle surface of a wound picture match degree judgement systems is as follows：
A) data processing：Two vehicle surface of a wound photos are subjected to data prediction (mainly removing noise effect), the data are pre-
Processing procedure includes：Defogging goes reflective, rotation, translation etc..
B damaged part) is identified：Pretreated photo is put into the sorter network model trained and is carried out at classification
Reason, determines damaged parts title.Specific training network model (using ResNet-50 depth convolutional networks, as shown in Figure 2) step
It is as follows：
Training stage：Data set includes about 100,000 images (9.5 ten thousand training images, 5000 authentication images), training
ResNet-50 depth convolutional networks, loss layers use Softmax.
Test phase：A test pictures are inputted, the corresponding damaged part of this pictures is correctly provided.
C matching degree) is calculated：Using the loss position of two vehicle surface of a wound photos, pretreated photo as the surface of a wound trained
The input of Matching Model obtains the matching degree of two vehicle surface of a wound photos, can be on backstage if the matching degree obtained is less than setting value
Operator is proposed to warn in management system, it is reminded to do more artificial treatments.Specific training network model (uses
ResNet-50 depth convolutional networks, as shown in Figure 2) steps are as follows：
Training stage：Data set includes about 50,000 pairs of images (4.75 ten thousand pairs of training images, 2500 pairs of authentication images), training
ResNet-50 depth convolutional networks, loss layers use Euclidean distance.
Test phase：A pair of of test pictures of input, provide this matching degree to picture, and matching degree is higher than setting value then result
Correctly.
Related procedure is as shown in Figure 5.
4) conclusion obtains
According to the handling result of step 2) and step 3), you can obtain the relatively high conclusion of accuracy, i.e. this vehicle insurance report
Whether case has car to car impact generation, and then provides the operator of background management system with suitable treatment advice.
In conclusion the present invention is based on the vehicle insurance accident discrimination method of image procossing and deep learning, pass through above-mentioned technology
Scheme, the data files such as the photo acquired in the scene of the accident using the personnel that report a case to the security authorities, video, to the position in picture, video, time
Equal elements are extracted and are compared, and by technological means such as image procossing, artificial intelligence, to accident vehicle and associated vehicle
The surface of a wound is analyzed, and to determine whether vehicle insurance fraud, has been finally reached a kind of efficient, quick, intelligentized vehicle insurance
Field antifraud effect.
In addition, the present invention also proposes a kind of vehicle insurance accident identification system based on image procossing and deep learning, the system
System includes：Memory, processor and be stored on the memory and can run on the processor based on image procossing
With the vehicle insurance accident evaluator of deep learning, the vehicle insurance accident evaluator based on image procossing and deep learning is by institute
It states and realizes the vehicle insurance accident discrimination method based on image procossing and deep learning described in embodiment as above when processor executes
Step, which is not described herein again.
In addition, the present invention also proposes a kind of computer readable storage medium, stored on the computer readable storage medium
There are the vehicle insurance accident evaluator based on image procossing and deep learning, the vehicle insurance thing based on image procossing and deep learning
Therefore evaluator realizes the vehicle insurance accident based on image procossing and deep learning described in embodiment as above when being executed by processor
The step of discrimination method, which is not described herein again.
In conclusion the present invention is based on the image procossing and vehicle insurance accident discrimination method of deep learning, system and storages to be situated between
Matter is using data files such as report a case to the security authorities photo, videos that personnel acquire in the scene of the accident, to position, the time etc. in picture, video
Element is extracted and is compared, and by technological means such as image procossing, artificial intelligence, to the wound of accident vehicle and associated vehicle
Face is analyzed, and to determine whether vehicle insurance fraud, has been finally reached a kind of efficient, quick, intelligentized vehicle insurance neck
Domain antifraud effect.
It these are only the preferred embodiment of the present invention, be not intended to limit the scope of the invention, it is every 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 (10)
1. a kind of vehicle insurance accident discrimination method based on image procossing and deep learning, which is characterized in that the method includes with
Lower step：
When being connected to vehicle insurance and reporting a case to the security authorities, the photo and/or video of be in danger field accident vehicle and associated vehicle are obtained；
Photo and/or video to the field accident vehicle that is in danger carry out information pre-processing, obtain the related letter of accident vehicle
Breath；
If the relevant information meets preset condition, photo to be in danger field accident vehicle and the associated vehicle and/or
Video carries out image procossing, obtains the accident vehicle and the associated vehicle surface of a wound treated photo and/or video；
Using deep learning method by after the accident vehicle and associated vehicle Wound treatment photo and/or video carry out
Match；
If matching result reaches preset value, judges that this accident is bumped against by the accident vehicle and associated vehicle and generate.
2. the vehicle insurance accident discrimination method according to claim 1 based on image procossing and deep learning, which is characterized in that
If the relevant information meets preset condition, photo to be in danger field accident vehicle and the associated vehicle and/or
Video carry out image procossing the step of include：
If the relevant information meets preset condition, photo to be in danger field accident vehicle and the associated vehicle and/or
Video carries out data prediction, the photo of be in danger field accident vehicle and the associated vehicle that obtain after data prediction and/
Or video, the data prediction include defogging, go the one or several kinds in reflective, rotation or translation；
It is described using deep learning method by after the accident vehicle and associated vehicle Wound treatment photo and/or video carry out
The step of matching includes：
The photo and/or video of be in danger described in after the data prediction field accident vehicle and associated vehicle are put into and instructed
Classification processing is carried out in experienced sorter network model, identifies the damaged part of the accident vehicle and associated vehicle；
By be in danger described in after the damaged part of the accident vehicle and associated vehicle, data prediction field accident vehicle and phase
Input of the photo and/or video to cut-off as the surface of a wound Matching Model trained, obtains the accident vehicle and mutually cut-offs
The matching degree of surface of a wound photo.
3. the vehicle insurance accident discrimination method according to claim 2 based on image procossing and deep learning, which is characterized in that
The relevant information of the accident vehicle includes：The number-plate number information of accident vehicle；
The photo and/or video to the field accident vehicle that is in danger carries out information pre-processing, obtains the phase of accident vehicle
Include after the step of closing information：
According to the vehicle license plate number information with personnel's ID bindings of reporting a case to the security authorities in personnel's ID inquiry databases of reporting a case to the security authorities；
By the license plate number with the vehicle license plate number information and the accident vehicle of personnel's ID bindings of reporting a case to the security authorities in the database
Code information compared to pair；
If comparison result is matching, execute to the photo of be in danger field accident vehicle and the associated vehicle and/or video into
The step of line number Data preprocess；
If comparison result is to mismatch, in background management system, prompt operating personnel carry out manpower intervention processing.
4. the vehicle insurance accident discrimination method according to claim 3 based on image procossing and deep learning, which is characterized in that
The relevant information of the accident vehicle further includes：Shooting is in danger the photo and/or video of field accident vehicle and associated vehicle
Temporal information and location information；
The photo and/or video to the field accident vehicle that is in danger carries out information pre-processing, obtains the phase of accident vehicle
Further include after the step of closing information：
Be in danger field accident vehicle and the photo of associated vehicle and/or the temporal information and location information of video are shot by described,
Compared with obtaining the temporal information and location information of photo and/or video of be in danger field accident vehicle and associated vehicle pair；
If comparison result is matching, and in the database with the vehicle license plate number information of the personnel ID binding of reporting a case to the security authorities with it is described
The number-plate number information of accident vehicle compare to comparison result be matching, then execute to field accident vehicle and the phase of being in danger
The step of photo and/or video to cut-off carries out data prediction；
If comparison result is to mismatch, in background management system, prompt operating personnel carry out manpower intervention processing.
5. the vehicle insurance accident discrimination method according to claim 1 based on image procossing and deep learning, which is characterized in that
When being connected to vehicle insurance and reporting a case to the security authorities, by the photo and/or video of field accident vehicle and associated vehicle of being in danger described in personnel's shooting of reporting a case to the security authorities,
It is described when being connected to vehicle insurance and reporting a case to the security authorities, the step of obtaining the photo and/or video of be in danger field accident vehicle and associated vehicle
Including：
The photo of be in danger field accident vehicle and the associated vehicle of the personnel that report a case to the security authorities described in identification shooting and/or the thing in video
Therefore the number-plate number of vehicle；
If identification is unsuccessful, the personnel that report a case to the security authorities described in prompt re-shoot the photo of be in danger field accident vehicle and associated vehicle
And/or video.
6. the vehicle insurance accident discrimination method according to claim 5 based on image procossing and deep learning, which is characterized in that
It is described when being connected to vehicle insurance and reporting a case to the security authorities, the step of photo and/or video for obtaining be in danger field accident vehicle and associated vehicle, also wraps
It includes：
In the photo and/or video of be in danger field accident vehicle and the associated vehicle of the personnel that report a case to the security authorities described in detection shooting whether
Profile with the accident vehicle and associated vehicle；
An only vehicle in the photo and/or video of field accident vehicle and associated vehicle if testing result is in danger for described in,
The personnel that report a case to the security authorities described in prompt re-shoot the photo and/or video of be in danger field accident vehicle and associated vehicle.
7. the vehicle insurance accident discrimination method according to claim 5 based on image procossing and deep learning, which is characterized in that
It is described when being connected to vehicle insurance and reporting a case to the security authorities, the step of photo and/or video for obtaining be in danger field accident vehicle and associated vehicle, also wraps
It includes：
Obtain the distance between interface surface of a wound position and the camera of the accident vehicle and associated vehicle；
If the surface of a wound of the accident vehicle and associated vehicle is completely covered in coverage, and the distance is more than preset value, then carries
Show that the personnel that report a case to the security authorities re-shoot.
8. the vehicle insurance accident discrimination method according to claim 7 based on image procossing and deep learning, which is characterized in that
The calculation formula of the interface surface of a wound position and the distance between camera position for obtaining the accident vehicle and associated vehicle
For：
D=√ [(x1-x2) ^2+ (y1-y2) ^2+ (z1-z2) ^2]；
Wherein, the interface surface of a wound position of the accident vehicle and associated vehicle is A (x1, y1, z1), and the camera position is B
(x2,y2,z2)。
9. a kind of vehicle insurance accident identification system based on image procossing and deep learning, which is characterized in that the system comprises：It deposits
Reservoir, processor and be stored on the memory and can run on the processor based on image procossing and deep learning
Vehicle insurance accident evaluator, the vehicle insurance accident evaluator based on image procossing and deep learning held by the processor
The vehicle insurance accident discrimination method based on image procossing and deep learning as described in any one of claim 1-8 is realized when row
The step of.
10. a kind of computer readable storage medium, which is characterized in that be stored on the computer readable storage medium based on figure
It is described that journey is differentiated based on the vehicle insurance accident of image procossing and deep learning as the vehicle insurance accident evaluator of processing and deep learning
The vehicle insurance thing based on image procossing and deep learning as described in any one of claim 1-8 is realized when sequence is executed by processor
Therefore the step of discrimination method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810283452.3A CN108491821A (en) | 2018-04-02 | 2018-04-02 | Vehicle insurance accident discrimination method, system and storage medium based on image procossing and deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810283452.3A CN108491821A (en) | 2018-04-02 | 2018-04-02 | Vehicle insurance accident discrimination method, system and storage medium based on image procossing and deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108491821A true CN108491821A (en) | 2018-09-04 |
Family
ID=63318164
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810283452.3A Pending CN108491821A (en) | 2018-04-02 | 2018-04-02 | Vehicle insurance accident discrimination method, system and storage medium based on image procossing and deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108491821A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109658260A (en) * | 2018-12-10 | 2019-04-19 | 泰康保险集团股份有限公司 | Method and device, medium and electronic equipment are determined based on the fraud of block chain |
| CN109784170A (en) * | 2018-12-13 | 2019-05-21 | 平安科技（深圳）有限公司 | Vehicle insurance damage identification method, device, equipment and storage medium based on image recognition |
| CN110942623A (en) * | 2018-09-21 | 2020-03-31 | 阿里巴巴集团控股有限公司 | Auxiliary traffic accident handling method and system |
| TWI690874B (en) * | 2018-10-24 | 2020-04-11 | 富邦產物保險股份有限公司 | Method and system for verifying remote accident scene image and identity |
| CN111402065A (en) * | 2019-12-23 | 2020-07-10 | 北京明略软件系统有限公司 | Method and device for mining vehicle insurance fraud groups and readable storage medium |
| CN111444901A (en) * | 2020-06-17 | 2020-07-24 | 爱保科技有限公司 | Processing method and device for distinguishing multiple vehicles from pictures in vehicle insurance case and electronic equipment |
| WO2020239089A1 (en) * | 2019-05-30 | 2020-12-03 | 深圳市聚蜂智能科技有限公司 | Insurance loss assessment method and apparatus, and computer device and storage medium |
| CN112242062A (en) * | 2020-10-15 | 2021-01-19 | 杭州宣迅电子科技有限公司 | Intelligent traffic real-time monitoring and processing system based on big data |
| CN113538293A (en) * | 2021-08-20 | 2021-10-22 | 爱保科技有限公司 | Method and device for enhancing vehicle damage image |
| CN115834621A (en) * | 2022-11-16 | 2023-03-21 | 山东新一代信息产业技术研究院有限公司 | Accident quick-place device and method based on artificial intelligence |
| US11615616B2 (en) | 2019-04-01 | 2023-03-28 | Jeff Jian Chen | User-guidance system based on augmented-reality and/or posture-detection techniques |
| US11966981B2 (en) | 2019-05-30 | 2024-04-23 | Shenzhen Ju Feng Technology Company | Method and apparatus for assessing an insured loss, computer device, and storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105389731A (en) * | 2015-11-09 | 2016-03-09 | 深圳市云智易联科技有限公司 | Image information processing method and related apparatus |
| CN105719188A (en) * | 2016-01-22 | 2016-06-29 | 平安科技（深圳）有限公司 | Method and server for achieving insurance claim anti-fraud based on consistency of multiple pictures |
| KR101656808B1 (en) * | 2015-03-20 | 2016-09-22 | 현대자동차주식회사 | Accident information manage apparatus, vehicle having the same and method for managing accident information |
| CN106127747A (en) * | 2016-06-17 | 2016-11-16 | 史方 | Car surface damage classifying method and device based on degree of depth study |
| CN106228449A (en) * | 2016-07-29 | 2016-12-14 | 深圳市永兴元科技有限公司 | Vehicle insurance Claims Resolution antifraud method and apparatus |
| CN106600422A (en) * | 2016-11-24 | 2017-04-26 | 中国平安财产保险股份有限公司 | Car insurance intelligent loss assessment method and system |
| CN106780048A (en) * | 2016-11-28 | 2017-05-31 | 中国平安财产保险股份有限公司 | A kind of self-service Claims Resolution method of intelligent vehicle insurance, self-service Claims Resolution apparatus and system |
| US20170337753A1 (en) * | 2016-05-17 | 2017-11-23 | International Business Machines Corporation | Vehicle accident reporting system |
| CN107862614A (en) * | 2017-11-08 | 2018-03-30 | 广东翼卡车联网服务有限公司 | A kind of method, storage device and car-mounted terminal for preventing vehicle insurance insurance fraud |
-
2018
- 2018-04-02 CN CN201810283452.3A patent/CN108491821A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101656808B1 (en) * | 2015-03-20 | 2016-09-22 | 현대자동차주식회사 | Accident information manage apparatus, vehicle having the same and method for managing accident information |
| CN105389731A (en) * | 2015-11-09 | 2016-03-09 | 深圳市云智易联科技有限公司 | Image information processing method and related apparatus |
| CN105719188A (en) * | 2016-01-22 | 2016-06-29 | 平安科技（深圳）有限公司 | Method and server for achieving insurance claim anti-fraud based on consistency of multiple pictures |
| US20170337753A1 (en) * | 2016-05-17 | 2017-11-23 | International Business Machines Corporation | Vehicle accident reporting system |
| CN106127747A (en) * | 2016-06-17 | 2016-11-16 | 史方 | Car surface damage classifying method and device based on degree of depth study |
| CN106228449A (en) * | 2016-07-29 | 2016-12-14 | 深圳市永兴元科技有限公司 | Vehicle insurance Claims Resolution antifraud method and apparatus |
| CN106600422A (en) * | 2016-11-24 | 2017-04-26 | 中国平安财产保险股份有限公司 | Car insurance intelligent loss assessment method and system |
| CN106780048A (en) * | 2016-11-28 | 2017-05-31 | 中国平安财产保险股份有限公司 | A kind of self-service Claims Resolution method of intelligent vehicle insurance, self-service Claims Resolution apparatus and system |
| CN107862614A (en) * | 2017-11-08 | 2018-03-30 | 广东翼卡车联网服务有限公司 | A kind of method, storage device and car-mounted terminal for preventing vehicle insurance insurance fraud |
Non-Patent Citations (1)
| Title |
|---|
| 胡伊等: "基于车载记录仪数据的车辆风险模型实证", 《计算机系统应用》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110942623A (en) * | 2018-09-21 | 2020-03-31 | 阿里巴巴集团控股有限公司 | Auxiliary traffic accident handling method and system |
| TWI690874B (en) * | 2018-10-24 | 2020-04-11 | 富邦產物保險股份有限公司 | Method and system for verifying remote accident scene image and identity |
| CN109658260A (en) * | 2018-12-10 | 2019-04-19 | 泰康保险集团股份有限公司 | Method and device, medium and electronic equipment are determined based on the fraud of block chain |
| CN109784170A (en) * | 2018-12-13 | 2019-05-21 | 平安科技（深圳）有限公司 | Vehicle insurance damage identification method, device, equipment and storage medium based on image recognition |
| US11615616B2 (en) | 2019-04-01 | 2023-03-28 | Jeff Jian Chen | User-guidance system based on augmented-reality and/or posture-detection techniques |
| WO2020239089A1 (en) * | 2019-05-30 | 2020-12-03 | 深圳市聚蜂智能科技有限公司 | Insurance loss assessment method and apparatus, and computer device and storage medium |
| US11966981B2 (en) | 2019-05-30 | 2024-04-23 | Shenzhen Ju Feng Technology Company | Method and apparatus for assessing an insured loss, computer device, and storage medium |
| CN111402065A (en) * | 2019-12-23 | 2020-07-10 | 北京明略软件系统有限公司 | Method and device for mining vehicle insurance fraud groups and readable storage medium |
| CN111444901A (en) * | 2020-06-17 | 2020-07-24 | 爱保科技有限公司 | Processing method and device for distinguishing multiple vehicles from pictures in vehicle insurance case and electronic equipment |
| CN112242062A (en) * | 2020-10-15 | 2021-01-19 | 杭州宣迅电子科技有限公司 | Intelligent traffic real-time monitoring and processing system based on big data |
| CN112242062B (en) * | 2020-10-15 | 2021-09-24 | 陕西天诚软件有限公司 | Intelligent traffic real-time monitoring and processing system based on big data |
| CN113538293A (en) * | 2021-08-20 | 2021-10-22 | 爱保科技有限公司 | Method and device for enhancing vehicle damage image |
| CN115834621A (en) * | 2022-11-16 | 2023-03-21 | 山东新一代信息产业技术研究院有限公司 | Accident quick-place device and method based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108491821A (en) | Vehicle insurance accident discrimination method, system and storage medium based on image procossing and deep learning | |
| CN111340008B (en) | Method and system for generation of counterpatch, training of detection model and defense of counterpatch | |
| CN103390153B (en) | For the method and system of the textural characteristics of biological characteristic validation | |
| CN106845414B (en) | Method and system for quality metrics for biometric verification | |
| Zhang et al. | Face morphing detection using Fourier spectrum of sensor pattern noise | |
| CN104537292B (en) | The method and system detected for the electronic deception of biological characteristic validation | |
| CN108229325A (en) | Method for detecting human face and system, electronic equipment, program and medium | |
| CN111881726B (en) | Living body detection method and device and storage medium | |
| CN106778525A (en) | Identity identifying method and device | |
| CN107438854A (en) | The system and method that the image captured using mobile device performs the user authentication based on fingerprint | |
| CN108549854A (en) | A kind of human face in-vivo detection method | |
| CN112507922B (en) | Face living body detection method and device, electronic equipment and storage medium | |
| CN109919754A (en) | A kind of data capture method, device, terminal and storage medium | |
| CN110705353A (en) | Method and device for identifying face to be shielded based on attention mechanism | |
| CN109977832A (en) | A kind of image processing method, device and storage medium | |
| CN110121109A (en) | Towards the real-time source tracing method of monitoring system digital video, city video monitoring system | |
| CN112132766A (en) | Image restoration method and device, storage medium and electronic device | |
| CN116541545A (en) | Method, device, equipment and storage medium for identifying flip image | |
| Barni et al. | Iris deidentification with high visual realism for privacy protection on websites and social networks | |
| Xu et al. | Facial depth forgery detection based on image gradient | |
| CN111241918A (en) | Vehicle anti-tracking method and system based on face recognition | |
| Jiang et al. | Haze relevant feature attention network for single image dehazing | |
| CN113989886A (en) | Crew identity verification method based on face recognition | |
| CN114065163A (en) | Display mainboard and terminal with face identification and identity verification functions | |
| CN114170470A (en) | Sample generation method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180904 |