CN105405054A - Insurance claim antifraud implementation method based on claim photo deep learning and server - Google Patents
Insurance claim antifraud implementation method based on claim photo deep learning and server Download PDFInfo
- Publication number
- CN105405054A CN105405054A CN201510925404.6A CN201510925404A CN105405054A CN 105405054 A CN105405054 A CN 105405054A CN 201510925404 A CN201510925404 A CN 201510925404A CN 105405054 A CN105405054 A CN 105405054A
- Authority
- CN
- China
- Prior art keywords
- photo
- segment
- character
- resolution
- rule
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000013135 deep learning Methods 0.000 title abstract 2
- 238000012795 verification Methods 0.000 claims abstract description 16
- 238000004458 analytical method Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims description 34
- 238000006243 chemical reaction Methods 0.000 claims description 22
- 238000012545 processing Methods 0.000 claims description 21
- 230000008569 process Effects 0.000 claims description 19
- 230000011218 segmentation Effects 0.000 claims description 19
- 230000002123 temporal effect Effects 0.000 claims description 19
- 239000000284 extract Substances 0.000 claims description 12
- 238000013528 artificial neural network Methods 0.000 claims description 10
- 230000000694 effects Effects 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 8
- 238000001914 filtration Methods 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 8
- 238000012216 screening Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 5
- 230000001143 conditioned effect Effects 0.000 claims description 4
- 230000006870 function Effects 0.000 abstract description 10
- 230000006399 behavior Effects 0.000 abstract description 3
- 230000001131 transforming effect Effects 0.000 abstract 2
- 238000012706 support-vector machine Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 238000002203 pretreatment Methods 0.000 description 6
- 238000007689 inspection Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000015572 biosynthetic process Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000003786 synthesis reaction Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses an insurance claim antifraud implementation method based on a claim photo deep learning. The method comprises: when a claim photo submitted by a user is received, performing frequency-domain transforming on the acquired photo based on a two-dimensional discrete cosine transform function, and according to a preset analysis rule and based on a color value of the photo subjected to the frequency-domain transforming in each color channel, performing authenticity verification on the photo; if the acquired photo are unreal, generating reminding information to remind that a fraudulent conduct exists in a claim application corresponding to the acquired photo; if the acquired photo is real, identifying photographing time of the acquired photo; extracting claim event occurrence time filled in the claim application corresponding to the acquired photo; and when the extracted claim event occurrence time does not match with the identified photographing time, generating reminding information to remind that the fraudulent conduct exists in the claim application corresponding to the acquired photo; The present invention also provides a server applicable to the method. The method can automatically identify a fraudulent claim behavior.
Description
Technical field
The present invention relates to financial services technology field, particularly a kind ofly realize the anti-method of swindling of settlement of insurance claim and server based on the study of Claims Resolution picture depth.
Background technology
At present, along with the continuous rising of vehicle guaranteeding organic quantity, the imbalance between supply and demand of road becomes increasingly conspicuous, and congestion in road phenomenon is more and more serious, manyly particularly slips as some the traffic congestion phenomenon that car, the fender-bender such as to knock into the back cause by traffic hazard and is on the rise.
For the traffic jam issue that solving road traffic hazard causes, traffic takes road surface people's police on duty and uses the fender-bender of summary procedure fast processing.But many fender-bender, driver dare not withdraw scene, and a lot of people thinks, once move car, insurance company has various reason and do not settle a claim, so still first wait traffic police, Zai Deng insurance company, the artificial heavy congestion causing road.
In view of the foregoing, associating Bao Jian department of traffic control department, common research and development mobile phone A PP software, collected evidence fast by mobile phone, once there occurs traffic hazard, with this APP Software Forensics, the information of shooting can be shared by traffic control department and insurance company.Allow car owner is relieved moves car, allow the relieved Claims Resolution of insurance company.
But such evidence collecting method, while facilitating the insurer and insurance company, also can be with and serve drawback, as in order to insurance fraud, the insurer may upload false evidence-obtaining photograph, as the scene of the accident photo synthesized by PS technology or distort, or uploads the photo of the non-genuine scene of the accident.Therefore, insurance company, when auditing the Claims Resolution application of the insurer, needs authenticity and the validity of desk checking photo, waste time and energy, and efficiency is not high.
Summary of the invention
In view of above content, be necessary to provide a kind of and realize the anti-method of swindling of settlement of insurance claim and server based on the study of Claims Resolution picture depth, the Claims Resolution photo going out to distort with self-verifying, and whether the automatic decision photo that goes out to settle a claim mates with corresponding Claims Resolution application, thus automatically identify the Claims Resolution behavior of swindle.
Realize the anti-method of swindling of settlement of insurance claim based on the study of Claims Resolution picture depth, comprising:
When receiving the Claims Resolution photo that user submits to, based on two-dimension discrete cosine transform function, frequency domain conversion is carried out to the photo obtained, according to preset analysis rule and based on frequency domain conversion after the color value of photo on each Color Channel, authenticity verification is carried out to described photo;
If the picture obtained is false, then generates prompting message and there is fraud to remind Claims Resolution application corresponding to the picture of acquisition;
If the picture obtained is real, then according to predetermined time recognition rule, identify the shooting time in the photo of acquisition;
Extract the Claims Resolution Time To Event filled in Claims Resolution application corresponding to the photo of acquisition; And
When the Claims Resolution Time To Event extracted does not mate with the shooting time of identification, generate prompting message and there is fraud to remind Claims Resolution application corresponding to the picture of acquisition.
Preferably, described default analysis rule is: extract the color value amplitude of variation of photo on each RGB Color Channel after frequency domain conversion, if there is the color value amplitude of variation on Color Channel to be greater than predetermined threshold value, then judge that the photo obtained is false photo; If do not have the color value amplitude of variation on color channel to be greater than described predetermined threshold value, then judge that the photo obtained is real photo.
Preferably, described predetermined time recognition rule is: position the segment of predeterminated position in the picture obtained according to the segment locating rule preset; Identify in oriented segment whether comprise temporal information according to the supporting vector machine model generated in advance; According to the Character segmentation rule preset, character figure block comminute is carried out to the segment identified; The character data that each character segment going out to split according to the character recognition Model Identification generated in advance is corresponding.
Preferably, described default segment locating rule is: application Gaussian Blur method comparison film predeterminated position carries out pre-service, reduces the level of detail of photo predeterminated position; By pretreated predeterminated position image gray processing; The computing of Sobel rim detection is carried out to the image after gray processing, thus obtains the single order horizontal direction derivative of image; Convert the image of gray processing to bianry image; Determine all profile blocks in bianry image, and filter out profile block to be analyzed according to the screening rule preset, and the profile block generation minimum enclosed rectangle frame for filtering out; The profile block filtering of being surrounded by rectangle frame at angle of inclination will be there is.
Preferably, described default Character segmentation rule is: will comprise the segment gray processing of time; Otsu threshold method is adopted to do binary conversion treatment to gray processing segment; Adopt findContours function to binaryzation segment contouring, and obtain the minimum enclosed rectangle frame of all character segments; Separated one by one for the segment in the minimum enclosed rectangle frame obtained, to be divided into each monocase segment.
Preferably, the generative process of described supporting vector machine model comprises: the photo sample obtaining predetermined number, positions the segment of predeterminated position in each the picture sample obtained according to the segment locating rule preset; Pre-service is carried out to each segment oriented, does not meet pre-conditioned segment to filter out; The segment comprising temporal information and the segment that do not comprise temporal information are distributed in two different files; From two files, the segment of each extraction the first preset ratio is as training data, and to carry out the training of supporting vector machine model, under two files, the segment of each the second remaining preset ratio is as test data, in order to the Classification and Identification effect of assessment models; Utilize the tile data of the first preset ratio extracted to carry out supporting vector machine model training to generate corresponding supporting vector machine model, utilize the tile data of the second remaining preset ratio to carry out Accuracy Verification to the supporting vector machine model generated; If train the supporting vector machine model recognition accuracy obtained to be less than default accuracy rate, then increase the figure number of blocks of training dataset, repeat the generative process of above-mentioned supporting vector machine model, until the supporting vector machine model accuracy rate generated is more than or equal to default accuracy rate.
Preferably, the generative process of described character recognition model comprises: the segment sample comprising temporal information obtaining predetermined number, carries out character figure block comminute according to above-mentioned default Character segmentation rule to each segment sample; Classify according to character types to all character segments of segmentation, so that the character segment of same character types is divided into a class, the character segment of kinds of characters type is divided into inhomogeneity; From each class, the character segment of each extraction the first preset ratio is as training data, and to carry out the training of multilayer perceptron model, under each class, the character segment of each the second remaining preset ratio is as test data, in order to the Classification and Identification effect of assessment models; Utilize the character tile data of the first preset ratio extracted to carry out artificial neural network training to generate corresponding multilayer perceptron model, utilize the character tile data of the second remaining preset ratio to carry out Accuracy Verification to the multilayer perceptron model generated; If the multilayer perceptron model accuracy rate generated is less than default accuracy rate, then increases the figure number of blocks of training dataset, repeat the generative process of above-mentioned multilayer perceptron model, until the multilayer perceptron model accuracy rate generated is more than or equal to default accuracy rate.
Be applicable to a server for said method, this server comprises memory device and processor, wherein:
Described storage unit, stores one and realizes the anti-system of swindling of settlement of insurance claim based on the study of Claims Resolution picture depth;
Described processor, described realizes the anti-system of swindling of settlement of insurance claim, to perform following steps based on the study of Claims Resolution picture depth for calling and performing:
When receiving the Claims Resolution photo that user submits to, based on two-dimension discrete cosine transform function, frequency domain conversion is carried out to the photo obtained, according to preset analysis rule and based on frequency domain conversion after the color value of photo on each Color Channel, authenticity verification is carried out to described photo;
If the picture obtained is false, then generates prompting message and there is fraud to remind Claims Resolution application corresponding to the picture of acquisition;
If the picture obtained is real, then according to predetermined time recognition rule, identify the shooting time in the photo of acquisition;
Extract the Claims Resolution Time To Event filled in Claims Resolution application corresponding to the photo of acquisition; And
When the Claims Resolution Time To Event extracted does not mate with the shooting time of identification, generate prompting message and there is fraud to remind Claims Resolution application corresponding to the picture of acquisition.
Preferably, described default analysis rule is: extract the color value amplitude of variation of photo on each RGB Color Channel after frequency domain conversion, if there is the color value amplitude of variation on Color Channel to be greater than predetermined threshold value, then judge that the photo obtained is false photo; If do not have the color value amplitude of variation on color channel to be greater than described predetermined threshold value, then judge that the photo obtained is real photo.
Preferably, described predetermined time recognition rule is: position the segment of predeterminated position in the picture obtained according to the segment locating rule preset; Identify in oriented segment whether comprise temporal information according to the supporting vector machine model generated in advance; According to the Character segmentation rule preset, character figure block comminute is carried out to the segment identified; The character data that each character segment going out to split according to the character recognition Model Identification generated in advance is corresponding.
Of the present inventionly realize settlement of insurance claim anti-method, system of swindling based on the study of Claims Resolution picture depth and be applicable to the server of said system, can the self-verifying Claims Resolution photo that goes out to distort, and whether the automatic decision photo that goes out to settle a claim mates with corresponding Claims Resolution application, thus automatically identify the Claims Resolution behavior of swindle.
Accompanying drawing explanation
Fig. 1 the present invention is based on the hardware environment figure that the study of Claims Resolution picture depth realizes the anti-system preferred embodiment of swindling of settlement of insurance claim.
Fig. 2 the present invention is based on the functional block diagram that the study of Claims Resolution picture depth realizes the anti-system preferred embodiment of swindling of settlement of insurance claim.
Fig. 3 the present invention is based on the method implementing procedure figure that the study of Claims Resolution picture depth realizes the anti-method preferred embodiment of swindling of settlement of insurance claim.
Fig. 4 is the detailed implementing procedure figure of one of them step in Fig. 3.
Embodiment
Consulting shown in Fig. 1, is the present invention is based on the hardware environment figure that the study of Claims Resolution picture depth realizes the anti-system preferred embodiment of swindling of settlement of insurance claim.
Realize the anti-system (hereinafter referred to as " the anti-fake system of settlement of insurance claim ") 10 of swindling of settlement of insurance claim based on the study of Claims Resolution picture depth described in the present embodiment can install and run in a server 1.This server 1 can be a Claims Resolution server.Described Claims Resolution server 1 can pass through communication module (not shown) and be connected with the communication of at least one station terminal 2, the Claims Resolution application submitted to the user of receiving terminal 2.Described Claims Resolution application can be a traffic hazard Claims Resolution application, and it can comprise Claims Resolution application form and relevant evidence, as the scene photograph etc. of traffic hazard.
Described terminal 3 can be the equipment such as PC, smart mobile phone, panel computer.
Described server 1 can include memory device and processor (not shown).Described memory device can be one or more non-volatile memory cells, as ROM, EPROM or FlashMemory (flash memory cell) etc.Described memory device can be built-in or be external in server 1.Described processor is arithmetic core (CoreUnit) and the control core (ControlUnit) of server 1, for the data in interpretive machine instruction and process computer software.
In the present embodiment, the anti-fake system of described settlement of insurance claim 10 can be a kind of computer software, it comprises the executable program code of computing machine, this program code can be stored in described memory device, under the execution of processor, realize following function: after the Claims Resolution application receiving user's submission, obtain the photo in this Claims Resolution application, whether utilize picture PS detection technique to detect photo itself has through artificial amendment, and utilize the shooting time information on degree of depth learning method comparison film surface to carry out extracting and identify, by the temporal information identified with described settle a claim apply in compared with the Claims Resolution Time To Event filled in, whether mate with described Claims Resolution Time To Event with the time identified described in judging, thus make whether the decision of claims rejected or Claims Resolution being applied for described Claims Resolution.
Consulting shown in Fig. 2, is the functional block diagram of the anti-fake system preferred embodiment of this settlement of insurance claim.
The function that the program code of the anti-fake system of settlement of insurance claim of the present invention 10 is different according to it, can be divided into multiple functional module.In present pre-ferred embodiments, the anti-fake system 10 of described settlement of insurance claim can comprise communication module 100, photo inspection module 101, alarm modules 102, time identification module 103, matching module 104 and decision module 105.
Described communication module 100 for realizing the information interaction with terminal 2, the Claims Resolution application that the user of such as receiving terminal 2 submits to, and transmit decision that this Claims Resolution is applied for the user of described terminal 2.
Whether described photo inspection module 101 for obtaining the Claims Resolution photo in described Claims Resolution application, and carries out authenticity verification to described photo, have through artificial amendment to detect photo itself.In the present embodiment, described photo inspection module 101 based on two-dimension discrete cosine transform function to obtain photo carry out frequency domain conversion, according to preset analysis rule and based on frequency domain conversion after the color value of photo on each Color Channel authenticity verification is carried out to described photo.
In present pre-ferred embodiments, described default analysis rule is: extract the color value amplitude of variation of photo on each RGB Color Channel after frequency domain conversion; If there is the color value amplitude of variation on Color Channel to be greater than predetermined threshold value, then judge that the photo obtained is false photo " such as, photo, the photo distorted of synthesis "; If do not have the color value amplitude of variation on color channel to be greater than described predetermined threshold value, then judge that the photo obtained is real photo.In the present embodiment, described predetermined threshold value can be 0.1.
Described alarm modules 102 for when the picture determining described acquisition is false picture, to system manager, as Claims Resolution application auditor reminds Claims Resolution application corresponding to the photo of described acquisition to there is fraud.
Described time identification module 103, for when determining that described photo is real photo, according to predetermined time recognition rule, identifies the shooting time in the photo of acquisition.
In present pre-ferred embodiments, described predetermined time recognition rule is: according to the segment locating rule preset to predeterminated position in the picture obtained, such as, the segment in the lower right corner positions, with in the picture of photo, find and determine the particular location at the date-time information place in photo; According to the classification forecast model generated in advance, as support vector machine (SupportVectorMachine, SVM) model, identify in oriented segment and whether comprise temporal information (such as, comprising shooting date and time point); According to the Character segmentation rule preset, character figure block comminute is carried out to the segment identified; Character data corresponding to each character segment of segmentation is identified according to the character recognition model generated in advance (such as, by the MLP model of artificial neural network training generation).
In present pre-ferred embodiments, described default segment locating rule is: application Gaussian Blur method comparison film predeterminated position carries out pre-service, (to the level of detail of reduction photo predeterminated position the object of pre-treatment step be that subsequent calculations removes noise, this pre-treatment step of preferred employing, in other embodiments of the invention, also this pre-treatment step can not be comprised); By pretreated predeterminated position image gray processing; The computing of Sobel rim detection is carried out to the image after gray processing, thus obtains the single order horizontal direction derivative (now shooting time information also will obviously be distinguished) of image; The image of gray processing is converted to bianry image (after this step, shooting time region will be connected to a rectangular-shaped region); Determine all profile blocks in bianry image, and filter out profile block to be analyzed according to the screening rule preset, and the profile block generation minimum enclosed rectangle frame for filtering out; The profile block filtering of being surrounded by rectangle frame at angle of inclination will be there is.In the present embodiment, described default screening rule is the minimum enclosed rectangle verifying every block profile according to the display size estimated in advance, filters out finally alternative rectangle frame.
In present pre-ferred embodiments, the generative process of described supporting vector machine model comprises: the photo sample obtaining predetermined number, according to described default segment locating rule, the segment of predeterminated position in each the picture sample obtained is positioned, determine the particular location at the date-time information place in photo; Pre-service is carried out to each segment oriented, to filter out the segment not meeting pre-conditioned (such as, angle of inclination is greater than predetermined angle extreme value, area is greater than preset area extreme value); The segment comprising temporal information and the segment that do not comprise temporal information are distributed in two different files; From two files, each extraction the first preset ratio (such as, 70%) segment is as training data, to carry out support vector machine (SupportVectorMachine, referred to as SVM) training of model, under two files, each the second remaining preset ratio (such as, 30%) segment as test data, in order to the Classification and Identification effect of assessment models; Utilize the tile data of the first preset ratio extracted to carry out SVM model training to generate corresponding SVM model, utilize the tile data of the second remaining preset ratio to carry out Accuracy Verification to the SVM model generated; If train the SVM Model Identification accuracy rate obtained to be less than default accuracy rate (such as, 99%), then increase the figure number of blocks of training dataset, repeat the generative process of above-mentioned SVM model, until the SVM model accuracy rate generated is more than or equal to default accuracy rate (such as, 99%).
In present pre-ferred embodiments, described default Character segmentation rule is: will comprise the segment gray processing of time; Otsu threshold method is adopted to do binary conversion treatment to gray processing segment; Adopt findContours function to binaryzation segment contouring, and obtain the minimum enclosed rectangle frame of all character segments; Separated one by one for the segment in the minimum enclosed rectangle frame obtained, to be divided into each monocase segment.
In present pre-ferred embodiments, the generative process of the described character recognition model generated in advance comprises: the segment sample comprising temporal information obtaining predetermined number, carries out character figure block comminute according to above-mentioned default Character segmentation rule to each segment sample; All character segments of segmentation are classified according to character types, so that the character segment of same character types is divided into a class, the character segment of kinds of characters type is divided into inhomogeneity (such as to comprise numeral, ten classes are divided into by 0-9, comprise symbol, be divided into other a few class by "/", "-", ": "); From each class, each extraction the first preset ratio (such as, 70%) character segment is as training data, to carry out artificial neural network (ArtificialNeuralNetworks, be abbreviated as ANN) in multilayer perceptron (multi-layerperceptrons, referred to as MLP) training of model, under each class, each the second remaining preset ratio (such as, 30%) character segment as test data, in order to the Classification and Identification effect of assessment models; Utilize the character tile data of the first preset ratio extracted to carry out artificial neural network training to generate corresponding MLP model, utilize the character tile data of the second remaining preset ratio to carry out Accuracy Verification to the MLP model generated; If the MLP model accuracy rate generated is less than default accuracy rate (such as, 99%), then increase the figure number of blocks of training dataset, repeat the generative process of above-mentioned MLP model, until the MLP model accuracy rate generated is more than or equal to default accuracy rate (such as, 99%).
Described matching module 104, for the Claims Resolution Time To Event filled in Claims Resolution application corresponding to the photo that extracts acquisition, judges whether the shooting time in photo matches with the Claims Resolution Time To Event filled in.Such as, whether the shooting time in described photo is identical with filled in Claims Resolution Time To Event.
The decision of described decision module 105 for applying for described Claims Resolution according to the generation of described matching result, as claims rejected or Claims Resolution.
Further, when the shooting time in photo does not mate with filled in Claims Resolution Time To Event, described alarm modules 102 also can to system manager, as Claims Resolution application auditor reminds Claims Resolution application corresponding to the photo of described acquisition to there is fraud.
Consulting shown in Fig. 3, is the present invention is based on the method implementing procedure figure that the study of Claims Resolution picture depth realizes the anti-method preferred embodiment of swindling of settlement of insurance claim.Realize the anti-method of swindling of settlement of insurance claim be not limited to step shown in process flow diagram based on the study of Claims Resolution picture depth described in the present embodiment, in addition in step shown in process flow diagram, some step can be omitted, order between step can change.
Step S10, communication module 100 judges whether the Claims Resolution application receiving user's submission from terminal 2.In the present embodiment, described Claims Resolution application can be a traffic hazard Claims Resolution application, and it can comprise Claims Resolution application form and relevant evidence, as the scene photograph etc. of traffic hazard.
Step S11, photo inspection module 101 obtains the photo in described Claims Resolution application, based on two-dimension discrete cosine transform function, frequency domain conversion is carried out to the photo obtained, according to the analysis rule preset and based on the color value of photo on each Color Channel after frequency domain conversion, authenticity verification is carried out to described photo, whether has through artificial amendment to detect photo itself.
In present pre-ferred embodiments, described default analysis rule is: extract the color value amplitude of variation of photo on each RGB Color Channel after frequency domain conversion, if there is the color value amplitude of variation on Color Channel to be greater than predetermined threshold value, then judge that the photo obtained is false photo " such as, photo, the photo distorted of synthesis "; If do not have the color value amplitude of variation on color channel to be greater than described predetermined threshold value, then judge that the photo obtained is real photo.In the present embodiment, described predetermined threshold value can be 0.1.
Step S12, described photo inspection module 101, according to above-mentioned validity check result, judges whether described photo is false photo, distorts as passed through.If described photo is through distorting, then flow performing step S17, alarm modules 102 reminds Claims Resolution application corresponding to the photo of acquisition to there is fraud, and in step S18, decision module 105 generates claims rejected and determines, to carry out claims rejected to described Claims Resolution application.
Otherwise if described photo is not through distorting, then perform step S13, time identification module 103, according to predetermined time recognition rule, identifies the shooting time in the photo of acquisition.In step S13, predetermined time recognition rule refers to the description in following Fig. 4.
Step S14, matching module 104 extracts the Claims Resolution Time To Event filled in Claims Resolution application corresponding to the photo of acquisition.
Step S15, matching module 104 judges whether the shooting time in photo matches with the Claims Resolution Time To Event filled in further.Such as, whether the shooting time in described photo is identical with filled in Claims Resolution Time To Event.If the shooting time in photo and the Claims Resolution Time To Event filled in match, then the step S16 that flow performing is following.Or, if whether the shooting time in photo does not match with the Claims Resolution Time To Event filled in, then the step S17 to S18 that flow performing is following.
Wherein, step S16, decision module 105 generates Claims Resolution and determines, to compensate described Claims Resolution application.
Wherein, step S17, alarm modules 102 reminds Claims Resolution application corresponding to the photo of acquisition to there is fraud, and in step S18, decision module 105 generates claims rejected and determines, to carry out claims rejected to described Claims Resolution application.
Step S19, communication module 100 notifies to determine described in user, i.e. the decision of claims rejected or Claims Resolution.
Consult described in Fig. 4, be the detailed implementing procedure figure of step S13 in Fig. 3, namely how identify the shooting time in the photo of acquisition.
Step S130, described time identification module 103 positions the segment of predeterminated position in the picture obtained according to the segment locating rule preset, and with in the picture of photo, finds and determines the particular location at the date-time information place in photo.Described predeterminated position is the picture lower right corner.
In present pre-ferred embodiments, described default segment locating rule is: application Gaussian Blur method comparison film predeterminated position carries out pre-service, (to the level of detail of reduction photo predeterminated position the object of pre-treatment step be that subsequent calculations removes noise, this pre-treatment step of preferred employing, in other embodiments of the invention, also this pre-treatment step can not be comprised); By pretreated predeterminated position image gray processing; The computing of Sobel rim detection is carried out to the image after gray processing, thus obtains the single order horizontal direction derivative (now shooting time information also will obviously be distinguished) of image; The image of gray processing is converted to bianry image (after this step, shooting time region will be connected to a rectangular-shaped region); Determine all profile blocks in bianry image, and filter out profile block to be analyzed according to the screening rule preset, and the profile block generation minimum enclosed rectangle frame for filtering out; The profile block filtering of being surrounded by rectangle frame at angle of inclination will be there is.In the present embodiment, described default screening rule is the minimum enclosed rectangle verifying every block profile according to the display size estimated in advance, filters out finally alternative rectangle frame.
Step S131, described time identification module 103 identifies in oriented segment whether comprise temporal information according to the classification forecast model generated in advance.The described classification forecast model generated in advance is for holding vector machine (SupportVectorMachine, SVM) model.
In present pre-ferred embodiments, the generative process of described supporting vector machine model comprises: the photo sample obtaining predetermined number, according to described default segment locating rule, the segment of predeterminated position in each the picture sample obtained is positioned, determine the particular location at the date-time information place in photo; Pre-service is carried out to each segment oriented, to filter out the segment not meeting pre-conditioned (such as, angle of inclination is greater than predetermined angle extreme value, area is greater than preset area extreme value); The segment comprising temporal information and the segment that do not comprise temporal information are distributed in two different files; From two files, each extraction the first preset ratio (such as, 70%) segment is as training data, to carry out support vector machine (SupportVectorMachine, referred to as SVM) training of model, under two files, each the second remaining preset ratio (such as, 30%) segment as test data, in order to the Classification and Identification effect of assessment models; Utilize the tile data of the first preset ratio extracted to carry out SVM model training to generate corresponding SVM model, utilize the tile data of the second remaining preset ratio to carry out Accuracy Verification to the SVM model generated; If train the SVM Model Identification accuracy rate obtained to be less than default accuracy rate (such as, 99%), then increase the figure number of blocks of training dataset, repeat the generative process of above-mentioned SVM model, until the SVM model accuracy rate generated is more than or equal to default accuracy rate (such as, 99%).
Step S132, described time identification module 103 carries out character figure block comminute according to the Character segmentation rule preset to the segment identified.
In present pre-ferred embodiments, described default Character segmentation rule is: will comprise the segment gray processing of time; Otsu threshold method is adopted to do binary conversion treatment to gray processing segment; Adopt findContours function to binaryzation segment contouring, and obtain the minimum enclosed rectangle frame of all character segments; Separated one by one for the segment in the minimum enclosed rectangle frame obtained, to be divided into each monocase segment.
Step S133, the character data that each character segment that described time identification module 103 goes out to split according to the character recognition Model Identification generated in advance is corresponding.In the present embodiment, the described character recognition model generated in advance is the MLP model that artificial neural network (ArtificialNeuralNetworks is abbreviated as ANN) training generates.
In present pre-ferred embodiments, the generative process of the described character recognition model generated in advance comprises: the segment sample comprising temporal information obtaining predetermined number, carries out character figure block comminute according to above-mentioned default Character segmentation rule to each segment sample; All character segments of segmentation are classified according to character types, so that the character segment of same character types is divided into a class, the character segment of kinds of characters type is divided into inhomogeneity (such as to comprise numeral, ten classes are divided into by 0-9, comprise symbol, be divided into other a few class by "/", "-", ": "); From each class, each extraction the first preset ratio (such as, 70%) character segment is as training data, to carry out the multilayer perceptron (multi-layerperceptrons in artificial neural network, referred to as MLP) training of model, under each class, each the second remaining preset ratio (such as, 30%) character segment as test data, in order to the Classification and Identification effect of assessment models; Utilize the character tile data of the first preset ratio extracted to carry out artificial neural network training to generate corresponding MLP model, utilize the character tile data of the second remaining preset ratio to carry out Accuracy Verification to the MLP model generated; If the MLP model accuracy rate generated is less than default accuracy rate (such as, 99%), then increase the figure number of blocks of training dataset, repeat the generative process of above-mentioned MLP model, until the MLP model accuracy rate generated is more than or equal to default accuracy rate (such as, 99%).
Step S134, generates the shooting time in photo according to identified character data.
It should be noted last that, above embodiment is only in order to illustrate technical scheme of the present invention and unrestricted, although with reference to preferred embodiment to invention has been detailed description, those of ordinary skill in the art is to be understood that, can modify to technical scheme of the present invention or equivalent replacement, and not depart from the spirit and scope of technical solution of the present invention.
Claims (10)
1. realize the anti-method of swindling of settlement of insurance claim based on the study of Claims Resolution picture depth, it is characterized in that, the method comprises:
When receiving the Claims Resolution photo that user submits to, based on two-dimension discrete cosine transform function, frequency domain conversion is carried out to the photo obtained, according to preset analysis rule and based on frequency domain conversion after the color value of photo on each Color Channel, authenticity verification is carried out to described photo;
If the picture obtained is false, then generates prompting message and there is fraud to remind Claims Resolution application corresponding to the picture of acquisition;
If the picture obtained is real, then according to predetermined time recognition rule, identify the shooting time in the photo of acquisition;
Extract the Claims Resolution Time To Event filled in Claims Resolution application corresponding to the photo of acquisition; And
When the Claims Resolution Time To Event extracted does not mate with the shooting time of identification, generate prompting message and there is fraud to remind Claims Resolution application corresponding to the picture of acquisition.
2. the method for claim 1, it is characterized in that, described default analysis rule is: extract the color value amplitude of variation of photo on each RGB Color Channel after frequency domain conversion, if there is the color value amplitude of variation on Color Channel to be greater than predetermined threshold value, then judge that the photo obtained is false photo; If do not have the color value amplitude of variation on color channel to be greater than described predetermined threshold value, then judge that the photo obtained is real photo.
3. the method for claim 1, is characterized in that, described predetermined time recognition rule is: position the segment of predeterminated position in the picture obtained according to the segment locating rule preset; Identify in oriented segment whether comprise temporal information according to the supporting vector machine model generated in advance; According to the Character segmentation rule preset, character figure block comminute is carried out to the segment identified; The character data that each character segment going out to split according to the character recognition Model Identification generated in advance is corresponding.
4. method as claimed in claim 3, it is characterized in that, described default segment locating rule is: application Gaussian Blur method comparison film predeterminated position carries out pre-service, reduces the level of detail of photo predeterminated position; By pretreated predeterminated position image gray processing; The computing of Sobel rim detection is carried out to the image after gray processing, thus obtains the single order horizontal direction derivative of image; Convert the image of gray processing to bianry image; Determine all profile blocks in bianry image, and filter out profile block to be analyzed according to the screening rule preset, and the profile block generation minimum enclosed rectangle frame for filtering out; The profile block filtering of being surrounded by rectangle frame at angle of inclination will be there is.
5. method as claimed in claim 3, it is characterized in that, described default Character segmentation rule is: will comprise the segment gray processing of time; Otsu threshold method is adopted to do binary conversion treatment to gray processing segment; Adopt findContours function to binaryzation segment contouring, and obtain the minimum enclosed rectangle frame of all character segments; Separated one by one for the segment in the minimum enclosed rectangle frame obtained, to be divided into each monocase segment.
6. method as claimed in claim 3, it is characterized in that, the generative process of described supporting vector machine model comprises: the photo sample obtaining predetermined number, positions the segment of predeterminated position in each the picture sample obtained according to the segment locating rule preset; Pre-service is carried out to each segment oriented, does not meet pre-conditioned segment to filter out; The segment comprising temporal information and the segment that do not comprise temporal information are distributed in two different files; From two files, the segment of each extraction the first preset ratio is as training data, and to carry out the training of supporting vector machine model, under two files, the segment of each the second remaining preset ratio is as test data, in order to the Classification and Identification effect of assessment models; Utilize the tile data of the first preset ratio extracted to carry out supporting vector machine model training to generate corresponding supporting vector machine model, utilize the tile data of the second remaining preset ratio to carry out Accuracy Verification to the supporting vector machine model generated; If train the supporting vector machine model recognition accuracy obtained to be less than default accuracy rate, then increase the figure number of blocks of training dataset, repeat the generative process of above-mentioned supporting vector machine model, until the supporting vector machine model accuracy rate generated is more than or equal to default accuracy rate.
7. method as claimed in claim 3, it is characterized in that, the generative process of described character recognition model comprises: the segment sample comprising temporal information obtaining predetermined number, carries out character figure block comminute according to above-mentioned default Character segmentation rule to each segment sample; Classify according to character types to all character segments of segmentation, so that the character segment of same character types is divided into a class, the character segment of kinds of characters type is divided into inhomogeneity; From each class, the character segment of each extraction the first preset ratio is as training data, and to carry out the training of multilayer perceptron model, under each class, the character segment of each the second remaining preset ratio is as test data, in order to the Classification and Identification effect of assessment models; Utilize the character tile data of the first preset ratio extracted to carry out artificial neural network training to generate corresponding multilayer perceptron model, utilize the character tile data of the second remaining preset ratio to carry out Accuracy Verification to the multilayer perceptron model generated; If the multilayer perceptron model accuracy rate generated is less than default accuracy rate, then increases the figure number of blocks of training dataset, repeat the generative process of above-mentioned multilayer perceptron model, until the multilayer perceptron model accuracy rate generated is more than or equal to default accuracy rate.
8. be applicable to a server for method described in any one of claim 1 to 7, it is characterized in that, this server comprises memory device and processor, wherein:
Described storage unit, stores one and realizes the anti-system of swindling of settlement of insurance claim based on the study of Claims Resolution picture depth;
Described processor, described realizes the anti-system of swindling of settlement of insurance claim, to perform following steps based on the study of Claims Resolution picture depth for calling and performing:
When receiving the Claims Resolution photo that user submits to, based on two-dimension discrete cosine transform function, frequency domain conversion is carried out to the photo obtained, according to preset analysis rule and based on frequency domain conversion after the color value of photo on each Color Channel, authenticity verification is carried out to described photo;
If the picture obtained is false, then generates prompting message and there is fraud to remind Claims Resolution application corresponding to the picture of acquisition;
If the picture obtained is real, then according to predetermined time recognition rule, identify the shooting time in the photo of acquisition;
Extract the Claims Resolution Time To Event filled in Claims Resolution application corresponding to the photo of acquisition; And
When the Claims Resolution Time To Event extracted does not mate with the shooting time of identification, generate prompting message and there is fraud to remind Claims Resolution application corresponding to the picture of acquisition.
9. server as claimed in claim 8, it is characterized in that, described default analysis rule is: extract the color value amplitude of variation of photo on each RGB Color Channel after frequency domain conversion, if there is the color value amplitude of variation on Color Channel to be greater than predetermined threshold value, then judge that the photo obtained is false photo; If do not have the color value amplitude of variation on color channel to be greater than described predetermined threshold value, then judge that the photo obtained is real photo.
10. server as claimed in claim 8, it is characterized in that, described predetermined time recognition rule is: position the segment of predeterminated position in the picture obtained according to the segment locating rule preset; Identify in oriented segment whether comprise temporal information according to the supporting vector machine model generated in advance; According to the Character segmentation rule preset, character figure block comminute is carried out to the segment identified; The character data that each character segment going out to split according to the character recognition Model Identification generated in advance is corresponding.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510925404.6A CN105405054A (en) | 2015-12-11 | 2015-12-11 | Insurance claim antifraud implementation method based on claim photo deep learning and server |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510925404.6A CN105405054A (en) | 2015-12-11 | 2015-12-11 | Insurance claim antifraud implementation method based on claim photo deep learning and server |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105405054A true CN105405054A (en) | 2016-03-16 |
Family
ID=55470524
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510925404.6A Pending CN105405054A (en) | 2015-12-11 | 2015-12-11 | Insurance claim antifraud implementation method based on claim photo deep learning and server |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105405054A (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107085814A (en) * | 2017-04-11 | 2017-08-22 | 武汉华创欣网科技有限公司 | A kind of analysis method and system of vehicle insurance Claims Resolution photo |
| CN107087025A (en) * | 2017-04-11 | 2017-08-22 | 武汉华创欣网科技有限公司 | A kind of car damage identification Claims Resolution method and device based on mobile network communication |
| CN107680046A (en) * | 2017-07-25 | 2018-02-09 | 平安科技（深圳）有限公司 | Image rectification method, device, storage medium and computer equipment |
| WO2018036276A1 (en) * | 2016-08-22 | 2018-03-01 | 平安科技（深圳）有限公司 | Image quality detection method, device, server and storage medium |
| WO2018055340A1 (en) * | 2016-09-21 | 2018-03-29 | Emergent Network Intelligence Ltd | Automatic image based object damage assessment |
| CN108090747A (en) * | 2018-01-25 | 2018-05-29 | 中国平安财产保险股份有限公司 | Self-service method, apparatus of validating the car, equipment and the computer storage media that vehicle insurance core is protected |
| WO2018120723A1 (en) * | 2016-12-30 | 2018-07-05 | 平安科技（深圳）有限公司 | Video compressive sensing reconstruction method and system, and electronic apparatus and storage medium |
| CN108323209A (en) * | 2018-01-29 | 2018-07-24 | 深圳前海达闼云端智能科技有限公司 | Information processing method, system, cloud processing device and computer program product |
| CN108647712A (en) * | 2018-05-08 | 2018-10-12 | 阿里巴巴集团控股有限公司 | Processing method, processing equipment, client and the server of vehicle damage identification |
| CN108682010A (en) * | 2018-05-08 | 2018-10-19 | 阿里巴巴集团控股有限公司 | Processing method, processing equipment, client and the server of vehicle damage identification |
| CN109214781A (en) * | 2018-09-12 | 2019-01-15 | 医倍思特（北京）医疗信息技术有限公司 | A kind of people hurts the check method of Claims Resolution setting loss expense |
| CN109214937A (en) * | 2018-09-27 | 2019-01-15 | 上海远眸软件有限公司 | The anti-fraud determination method of settlement of insurance claim intelligence and system |
| CN110136009A (en) * | 2019-04-17 | 2019-08-16 | 阿里巴巴集团控股有限公司 | Accident vehicle Claims Resolution method and apparatus, electronic equipment |
| CN110956149A (en) * | 2019-12-06 | 2020-04-03 | 中国平安财产保险股份有限公司 | Pet identity verification method, device and equipment and computer readable storage medium |
| CN112001785A (en) * | 2020-07-21 | 2020-11-27 | 小花网络科技(深圳)有限公司 | Network credit fraud identification method and system based on image identification |
| CN112053051A (en) * | 2020-08-27 | 2020-12-08 | 中信银行股份有限公司 | Due diligence application system and information processing method thereof |
| CN113011893A (en) * | 2021-03-22 | 2021-06-22 | 腾讯科技（深圳）有限公司 | Data processing method and device, computer equipment and storage medium |
| WO2023006974A1 (en) | 2021-07-30 | 2023-02-02 | Swiss Reinsurance Company Ltd. | Optical fraud detector for automated detection of fraud in digital imaginary-based automobile claims, automated damage recognition, and method thereof |
| EP4028977A4 (en) * | 2019-09-09 | 2023-09-27 | Neural Claim System, Inc. | Methods and systems for submitting and/or processing insurance claims for damaged motor vehicle glass |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101303761A (en) * | 2008-06-10 | 2008-11-12 | 裘炅 | Integrated system for claim settlement of vehicle insurance capable of shooting and uploading evidence-obtaining photograph or video by mobile phone |
| CN101546416A (en) * | 2009-04-27 | 2009-09-30 | 谢谦 | Method for remotely treating automobile insurance service and system thereof |
| CN104021567A (en) * | 2014-06-26 | 2014-09-03 | 福州大学 | Gaussian blur falsification detection method of image based on initial digital law |
-
2015
- 2015-12-11 CN CN201510925404.6A patent/CN105405054A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101303761A (en) * | 2008-06-10 | 2008-11-12 | 裘炅 | Integrated system for claim settlement of vehicle insurance capable of shooting and uploading evidence-obtaining photograph or video by mobile phone |
| CN101546416A (en) * | 2009-04-27 | 2009-09-30 | 谢谦 | Method for remotely treating automobile insurance service and system thereof |
| CN104021567A (en) * | 2014-06-26 | 2014-09-03 | 福州大学 | Gaussian blur falsification detection method of image based on initial digital law |
Non-Patent Citations (2)
| Title |
|---|
| 孙堡垒 等: "基于Benford定律的高斯模糊篡改取证", 《计算机研究与发展》 * |
| 朱曦: "彩色照片中日期文字的提取与识别", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018036276A1 (en) * | 2016-08-22 | 2018-03-01 | 平安科技（深圳）有限公司 | Image quality detection method, device, server and storage medium |
| WO2018055340A1 (en) * | 2016-09-21 | 2018-03-29 | Emergent Network Intelligence Ltd | Automatic image based object damage assessment |
| GB2554361A (en) * | 2016-09-21 | 2018-04-04 | Emergent Network Intelligence Ltd | Automatic image based object damage assessment |
| US11823365B2 (en) | 2016-09-21 | 2023-11-21 | Emergent Network Intelligence Ltd. | Automatic image based object damage assessment |
| GB2554361B (en) * | 2016-09-21 | 2022-05-11 | Emergent Network Intelligence Ltd | Automatic image based object damage assessment |
| WO2018120723A1 (en) * | 2016-12-30 | 2018-07-05 | 平安科技（深圳）有限公司 | Video compressive sensing reconstruction method and system, and electronic apparatus and storage medium |
| US10630995B2 (en) | 2016-12-30 | 2020-04-21 | Ping An Technology (Shenzhen) Co., Ltd. | Video compressed sensing reconstruction method, system, electronic device, and storage medium |
| CN107087025A (en) * | 2017-04-11 | 2017-08-22 | 武汉华创欣网科技有限公司 | A kind of car damage identification Claims Resolution method and device based on mobile network communication |
| CN107085814A (en) * | 2017-04-11 | 2017-08-22 | 武汉华创欣网科技有限公司 | A kind of analysis method and system of vehicle insurance Claims Resolution photo |
| WO2019019383A1 (en) * | 2017-07-25 | 2019-01-31 | 平安科技（深圳）有限公司 | Image correction method and apparatus, storage medium and computer device |
| CN107680046A (en) * | 2017-07-25 | 2018-02-09 | 平安科技（深圳）有限公司 | Image rectification method, device, storage medium and computer equipment |
| CN108090747A (en) * | 2018-01-25 | 2018-05-29 | 中国平安财产保险股份有限公司 | Self-service method, apparatus of validating the car, equipment and the computer storage media that vehicle insurance core is protected |
| WO2019144416A1 (en) * | 2018-01-29 | 2019-08-01 | 深圳前海达闼云端智能科技有限公司 | Information processing method and system, cloud processing device and computer program product |
| CN108323209A (en) * | 2018-01-29 | 2018-07-24 | 深圳前海达闼云端智能科技有限公司 | Information processing method, system, cloud processing device and computer program product |
| CN108323209B (en) * | 2018-01-29 | 2023-10-31 | 达闼机器人股份有限公司 | Information processing method, system, cloud processing device and computer storage medium |
| CN108647712A (en) * | 2018-05-08 | 2018-10-12 | 阿里巴巴集团控股有限公司 | Processing method, processing equipment, client and the server of vehicle damage identification |
| WO2019214321A1 (en) * | 2018-05-08 | 2019-11-14 | 阿里巴巴集团控股有限公司 | Vehicle damage identification processing method, processing device, client and server |
| CN108682010A (en) * | 2018-05-08 | 2018-10-19 | 阿里巴巴集团控股有限公司 | Processing method, processing equipment, client and the server of vehicle damage identification |
| TWI715932B (en) * | 2018-05-08 | 2021-01-11 | 開曼群島商創新先進技術有限公司 | Vehicle damage identification processing method and its processing device, data processing equipment for vehicle damage assessment, damage assessment processing system, client and server |
| CN109214781A (en) * | 2018-09-12 | 2019-01-15 | 医倍思特（北京）医疗信息技术有限公司 | A kind of people hurts the check method of Claims Resolution setting loss expense |
| CN109214937A (en) * | 2018-09-27 | 2019-01-15 | 上海远眸软件有限公司 | The anti-fraud determination method of settlement of insurance claim intelligence and system |
| CN110136009A (en) * | 2019-04-17 | 2019-08-16 | 阿里巴巴集团控股有限公司 | Accident vehicle Claims Resolution method and apparatus, electronic equipment |
| EP4028977A4 (en) * | 2019-09-09 | 2023-09-27 | Neural Claim System, Inc. | Methods and systems for submitting and/or processing insurance claims for damaged motor vehicle glass |
| CN110956149A (en) * | 2019-12-06 | 2020-04-03 | 中国平安财产保险股份有限公司 | Pet identity verification method, device and equipment and computer readable storage medium |
| CN110956149B (en) * | 2019-12-06 | 2025-01-14 | 中国平安财产保险股份有限公司 | Pet identity verification method, device, equipment and computer-readable storage medium |
| CN112001785A (en) * | 2020-07-21 | 2020-11-27 | 小花网络科技(深圳)有限公司 | Network credit fraud identification method and system based on image identification |
| CN112053051A (en) * | 2020-08-27 | 2020-12-08 | 中信银行股份有限公司 | Due diligence application system and information processing method thereof |
| CN113011893A (en) * | 2021-03-22 | 2021-06-22 | 腾讯科技（深圳）有限公司 | Data processing method and device, computer equipment and storage medium |
| CN113011893B (en) * | 2021-03-22 | 2024-04-26 | 腾讯科技（深圳）有限公司 | Data processing method, device, computer equipment and storage medium |
| WO2023006974A1 (en) | 2021-07-30 | 2023-02-02 | Swiss Reinsurance Company Ltd. | Optical fraud detector for automated detection of fraud in digital imaginary-based automobile claims, automated damage recognition, and method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105405054A (en) | Insurance claim antifraud implementation method based on claim photo deep learning and server | |
| JP6873237B2 (en) | Image-based vehicle damage assessment methods, equipment, and systems, as well as electronic devices | |
| Wang et al. | Asphalt pavement pothole detection and segmentation based on wavelet energy field | |
| EP3678049A1 (en) | Document verification and identity verification method and device | |
| CN112085952B (en) | Method and device for monitoring vehicle data, computer equipment and storage medium | |
| US11514526B1 (en) | Systems and methods for property damage restoration predictions based upon processed digital images | |
| WO2018166116A1 (en) | Car damage recognition method, electronic apparatus and computer-readable storage medium | |
| US8761498B1 (en) | Face and license plate detection in street level images with 3-D road width features estimated from laser data | |
| CN106203387A (en) | Face verification method and system | |
| CN111126098B (en) | Certificate image acquisition method, device, equipment and storage medium | |
| CN103810635A (en) | Method and device for realizing checking of loan business processing flow | |
| CN111932363A (en) | Identification and verification method, device, equipment and system for authorization book | |
| EP3265960B1 (en) | Methods for categorizing input images for use e.g. as a gateway to authentication systems | |
| Chauhan et al. | Intelligent toll collection system for moving vehicles in India | |
| CN103679209A (en) | Sparse theory based character recognition method | |
| CN111178347B (en) | Ambiguity detection method, ambiguity detection device, ambiguity detection equipment and ambiguity detection storage medium for certificate image | |
| CN114511866A (en) | Data auditing method, device, system, processor and machine-readable storage medium | |
| CN114386013A (en) | Automatic student status authentication method and device, computer equipment and storage medium | |
| CN106778765B (en) | License plate recognition method and device | |
| CN113077355B (en) | Insurance claim settlement method and device, electronic equipment and storage medium | |
| EP4105825A1 (en) | Generalised anomaly detection | |
| CN115131564A (en) | Vehicle component damage detection method based on artificial intelligence and related equipment | |
| Singh et al. | Evaluating the Performance of Ensembled YOLOv8 Variants in Smart Parking Applications for Vehicle Detection and License Plate Recognition under Varying Lighting Conditions | |
| Luna et al. | License Plate Recognition for Stolen Vehicles Using Optical Character Recognition | |
| Nguyen et al. | UIT-ANPR: toward an open framework for automatic number plate recognition on smartphones |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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: 20160316 |