CN113591715A - LNG vehicle station-entering certificate verification full-process post-processing method based on deep learning - Google Patents
LNG vehicle station-entering certificate verification full-process post-processing method based on deep learning Download PDFInfo
- Publication number
- CN113591715A CN113591715A CN202110876381.XA CN202110876381A CN113591715A CN 113591715 A CN113591715 A CN 113591715A CN 202110876381 A CN202110876381 A CN 202110876381A CN 113591715 A CN113591715 A CN 113591715A
- Authority
- CN
- China
- Prior art keywords
- certificate
- information
- face
- republic
- people
- 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 62
- 238000012795 verification Methods 0.000 title claims abstract description 26
- 238000013135 deep learning Methods 0.000 title claims abstract description 24
- 238000012805 post-processing Methods 0.000 title claims abstract description 21
- 230000006870 function Effects 0.000 claims abstract description 9
- 238000004806 packaging method and process Methods 0.000 claims abstract description 6
- 238000001514 detection method Methods 0.000 claims description 20
- 239000013598 vector Substances 0.000 claims description 18
- 230000008569 process Effects 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 12
- 230000002596 correlated effect Effects 0.000 claims description 2
- 238000005538 encapsulation Methods 0.000 claims description 2
- 238000012986 modification Methods 0.000 abstract description 4
- 230000004048 modification Effects 0.000 abstract description 4
- 238000004886 process control Methods 0.000 abstract description 2
- 230000011218 segmentation Effects 0.000 description 4
- 239000012634 fragment Substances 0.000 description 3
- 102100032202 Cornulin Human genes 0.000 description 2
- 101000920981 Homo sapiens Cornulin Proteins 0.000 description 2
- 230000000875 corresponding effect Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Abstract
The invention provides a full-process post-processing method for verification of an LNG vehicle station entering certificate based on deep learning, and relates to the technical field of safety management and process control. The method comprises the following steps: extracting certificate information of a plurality of certificates, wherein the certificate information comprises certificate names, keywords and other information; comparing the certificate information of each certificate, and determining comparison fields and the associated certificates; then comparing the certificate information with the incoming vehicle information and the face information; the configuration file uses a yaml format, is indented and aligned, and can be converted into a dictionary in a key-value pair format after being read by a python program; the comparison comprises a face recognition model, an OCR model, API packaging and flash service; the API packages and instantiates an OCR model, packages functions of OCR objects needing to interact with the outside into interfaces, and carries out friendly prompt on some input exceptions; and writing a flash server and providing an http server interface. The method strips the identified and matched logic from the code to form a configuration file, and the modification is flexible.
Description
Technical Field
The invention relates to the technical field of safety management and process control, in particular to a LNG vehicle station-entering certificate verification full-process post-processing method based on deep learning.
Background
In the prior art, a mainstream algorithm generally comprises two stages of text detection and text recognition, and for a text-non-dense image, an end-to-end (that is, detection and recognition are performed simultaneously in one model) deep learning algorithm is also provided.
The output result of the general OCR pre-training model is a line-level text enclosure box and text content, and how to effectively extract structured effective information from the whole image OCR output result is a very important ring in the floor application process of the deep learning OCR model. In the field of OCR certificate recognition, the prior art mostly locates the required information by certificate format matching.
The chinese patent (CN112348022A) provides a technical solution, which requires a pre-defined template, wherein the template includes a plurality of rectangular fragment coordinate positions. Firstly, positioning the template by using the full-image OCR keyword, then cutting according to the preset rectangular fragments, and sending the cut fragments into the recognition model. The recognition model used by the method is RCNN, belongs to pure image recognition, and does not add sequence learning of natural language processing.
In chinese patent (CN112380957A), the template is determined by directly performing feature comparison between certificates. Requires a pre-defined template and is very dependent on the effect of the image pre-processing. If the certificate layout has a plurality of versions, the certificate layout cannot be dealt with. And the OCR models used by the OCR models are only recognized from the image perspective without considering the semantics, so that the OCR models are very limited and are easy to be recognized wrongly.
In the scene that the LNG transport vehicle enters the natural gas station for unloading, the problems that the format is not fixed, the requirement is changed frequently and the like due to various certificates and different versions are faced. Meanwhile, the problems are likely to occur in other occasions requiring certificate identification.
Disclosure of Invention
In order to realize a complete process from text information identification to structural information extraction and corresponding information comparison in a certificate identification task; when the certificate version is updated or the comparison requirement is changed, the effective text information can be extracted from the output result of the OCR model and the comparison result can be returned only by changing the general configuration file designed by the invention without modifying codes; the invention provides a full-process post-processing method for verification of an LNG vehicle station entering certificate based on deep learning, and the specific technical scheme is as follows.
An LNG vehicle station entering certificate verification full-process post-processing method based on deep learning comprises the following steps:
s1, extracting certificate information of each certificate, wherein the certificate information comprises a certificate name and information keywords;
s2, comparing the certificate information of each certificate, including determining comparison fields and correlated certificate names;
s3, comparing certificate information with incoming vehicles to obtain information and face information;
the configuration file is in a yaml format, is indented and aligned, and is converted into a dictionary in a key-value pair format after being read by a python program; the implementation logic of the code comprises a face recognition model, an OCR model, API packaging and flash service; the API packages and instantiates an OCR model, packages functions of OCR objects needing to interact with the outside into interfaces, and prompts input abnormity; and writing a flash server and providing an http server interface.
Preferably, the LNG vehicle arrival certificate includes: the characteristic equipment uses registration certificate, the card of road transportation of the people's republic of China-semitrailer, the card of road transportation of people's republic of China-tractor, the card of driver of road transportation of dangerous goods, the card of escort personnel of road transportation of dangerous goods, the card of driving motor vehicles of people's republic of China-semitrailer, the card of driving motor vehicles of people's republic of China-tractor.
Preferably, the certificate information extracted by the feature device using the registration certificate includes a license plate and a device variety; certificate information extracted by the road transport certificate-semitrailer of the people's republic of China comprises a validity period, an operation range and a license plate; certificate information extracted by the road transport certificate-tractor of the people's republic of China comprises a validity period, an operation range and a license plate; certificate information extracted by the road dangerous goods transport driver certificate comprises a validity period, a name and a certificate number; certificate information extracted by the road dangerous goods transportation escort certificate comprises an expiration date, a name and a certificate number; certificate information extracted from the motor vehicle driving license of the people's republic of China comprises an expiration date, a name and a certificate number; certificate information extracted by a semitrailer which is a motor vehicle driving certificate of the people's republic of China comprises an expiration date and a license plate; the certificate information extracted by the motor vehicle driving certificate-tractor of the people's republic of China comprises the validity period and the license plate.
Preferably, the associated document specifically comprises: the license plate numbers of the road transportation certificate-semitrailer of the people's republic of China and the motor vehicle driving certificate-semitrailer of the people's republic of China are related; the license plate numbers of the road transportation certificate-tractor of the people's republic of China and the motor vehicle driving certificate-tractor of the people's republic of China are related; the road dangerous goods transport driver's license is related to the name, face and license number of the people's republic of China's motor vehicle license.
Preferably, the comparing of the certificate information with the information and face information of the vehicle entering the vehicle specifically comprises: the license plate number of the incoming vehicle is consistent with the license plate number of the certificate; the scene shooting of the vehicle related personnel is matched with the certificate photo.
Preferably, the key is "certificate" and the value is a dictionary object; the certificate name is identified, the keywords matched with the certificate, the information keywords needing to be extracted from the certificate, the direction of the information relative to the keywords and the information regular expression are obtained.
Preferably, in the comparison process of the certificate information, the first layer key represents the configuration information of the first part for the certificate; the second layer of keys are 'special equipment use registration certificates' which represent certificate names to be identified; the third layer of keys are specific structured information matching rules.
Preferably, in the comparison process of the certificate information and the incoming vehicle information and the face information, the key of the first layer is "matching" to indicate a configuration file, the key of the second layer is "license plate" and the key of the third layer in the comparison process of the certificate information of the certificate, the value of the key of the third layer is a list, the elements in the list are certificate names to indicate that the certificates need to be compared with the field.
Further preferably, the face recognition module encapsulates the whole face recognition module, and provides an interface for the OCR model to use; the face registration is used for detecting a face in the certificate, calculating a face feature vector and storing the face feature vector in an OCR result dictionary; the face comparison interface inputs the image and an OCR recognition result dictionary, detects the face in the image, and when the face feature vectors temporarily stored in the OCR recognition result are matched, the comparison is successful; the OCR model comprises text detection and recognition, information extraction, information comparison and calling of a face recognition interface.
Preferably, the certificate matching interface is called after the recognition is finished, the API encapsulation instantiates an OCR model, encapsulates functions of the OCR object which need to interact with the outside into the interface, and prompts input abnormity.
The LNG vehicle station-entering verification full-process post-processing method based on deep learning has the advantages that: an OCR deep learning algorithm is adopted, so that the robustness is high; the adopted mode is used for extracting the structured information, so that the flexibility is high, and the method is not limited by the format; the identified and matched logic is stripped from the code to form a configuration file, and the modification is flexible. The method adds the verification links of certificate face and field photo in the whole process, uses three cascade models of face detection, face key point detection and 128-dimensional feature vector extraction, and completes two tasks of detection and identification.
Drawings
In order to more clearly illustrate the embodiments or technical solutions of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained based on these drawings without creative efforts.
FIG. 1 is a flow chart of a full-process LNG vehicle station-entering verification post-processing method based on deep learning.
Detailed Description
With reference to fig. 1, a detailed description is given of an LNG vehicle arrival verification full-process post-processing method based on deep learning according to the present invention.
An LNG vehicle station entering certificate verification full-process post-processing method based on deep learning comprises the following steps:
s1, certificate information of each certificate is extracted, wherein the certificate information comprises certificate names and information keywords.
The LNG car certificate of arrival includes: the characteristic equipment uses registration certificate, the card of road transportation of the people's republic of China-semitrailer, the card of road transportation of people's republic of China-tractor, the card of driver of road transportation of dangerous goods, the card of escort personnel of road transportation of dangerous goods, the card of driving motor vehicles of people's republic of China-semitrailer, the card of driving motor vehicles of people's republic of China-tractor.
And S2, comparing the certificate information of each certificate, including determining comparison fields and the associated certificate names.
The certificate information extracted by the characteristic equipment using the registration certificate comprises a license plate and equipment varieties; certificate information extracted by the road transport certificate-semitrailer of the people's republic of China comprises a validity period, an operation range and a license plate; certificate information extracted by the road transport certificate-tractor of the people's republic of China comprises a validity period, an operation range and a license plate; certificate information extracted by the road dangerous goods transport driver certificate comprises a validity period, a name and a certificate number; certificate information extracted by the road dangerous goods transportation escort certificate comprises an expiration date, a name and a certificate number; certificate information extracted from the motor vehicle driving license of the people's republic of China comprises an expiration date, a name and a certificate number; certificate information extracted by a semitrailer which is a motor vehicle driving certificate of the people's republic of China comprises an expiration date and a license plate; the certificate information extracted by the motor vehicle driving certificate-tractor of the people's republic of China comprises the validity period and the license plate.
The associated documents include: the license plate numbers of the road transportation certificate-semitrailer of the people's republic of China and the motor vehicle driving certificate-semitrailer of the people's republic of China are related; the license plate numbers of the road transportation certificate-tractor of the people's republic of China and the motor vehicle driving certificate-tractor of the people's republic of China are related; the road dangerous goods transport driver's license is related to the name, face and license number of the people's republic of China's motor vehicle license.
And S3, comparing the certificate information with the incoming vehicle information and the face information.
The information and face information comparison between the certificate information and the vehicles entering the parking lot specifically comprises the following steps: the license plate number of the incoming vehicle is consistent with the license plate number of the certificate; the scene shooting of the vehicle related personnel is matched with the certificate photo.
The method flexibly processes logic configuration after certificate identification, designs a complete processing flow aiming at the requirements of identification and verification of the certificate of the LNG vehicle entering the station, and the configuration file comprises the following steps: (1) the information extraction logic, in this section, describes the certificate name to be identified, the keywords that match the certificate, the information keywords to be extracted from the certificate, the orientation of the information relative to the keywords, and the information regular expression. (2) Comparing information among certificates; in this section, the fields that need to be aligned are described, as well as which certificates in which the fields need to be consistent. (3) Extracting human face features; the certificate names listed in the part indicate that the face needs to be detected and feature vectors are extracted from the certificates so as to be compared with the face on site for verification.
In the specific implementation, a detection and recognition two-stage algorithm is used, the text detection model is DB, and the text recognition model is CRNN. DB (differential localization) is a detection model based on segmentation, binarization is executed in a segmentation network, a probability graph generated by segmentation is converted into a text bounding box, post-processing logic is simplified, and the performance of the text detection model based on segmentation is improved. The CRNN text recognition algorithm is an image-based sequence recognition model, uses a convolutional neural network with a VGG structure to extract image characteristics, extracts sequence characteristics through a bidirectional LSTM network after serialization, and finally obtains a line-level text recognition result through a CTC translation layer. Meanwhile, the invention adds the verification links of certificate face and field photo in the whole process, uses three cascade models of face detection, face key point detection and 128-dimensional feature vector extraction, and completes two tasks of detection and identification.
The configuration file uses a yaml format, is indented and aligned, and is converted into a dictionary in a key-value pair format after being read by a python program. The key is 'certificate' and the value is a dictionary object; the certificate name is identified, the keywords matched with the certificate, the information keywords needing to be extracted from the certificate, the direction of the information relative to the keywords and the information regular expression are obtained.
In the extraction of information: the key is "certificate" and the value is a dictionary object. In this section, the name of the certificate to be recognized, the keyword matched to the certificate, the information keyword to be extracted from the certificate, the orientation of the information with respect to the keyword, and the information regular expression are described.
For example:
certificate:
special equipment use registration certificate:
certificate name-special equipment use registration certificate
License plate [ \ u4E00- \ u9FA5] [ A-Z ] [ A-Z0-9] {4} hanging' ]
Equipment variety [ equipment variety, RIGHT, 'semitrailer' ]
The outermost key "certificate" indicates that this is the configuration information of the first part, and the second key "special device use registry" indicates the name of the certificate to be recognized. Each key of the third layer corresponds to the information content to be extracted, wherein the value of "certificate name" is used to match to which certificate the image belongs, and may be a character string or a list. In the case of a list, all elements of the list must be matched simultaneously. The rest third-level keys are specific structural information matching rules, and values are expressed in a list form in the third-level key value pairs. The extracted information is also structured, i.e. in the form of key-value pairs, as far as possible.
The list thus contains three elements: the first element is a keyword of the structured information, such as "name", which may be a regular expression; the second element is the direction of the information relative to the keywords, and the total five values are respectively 'LEFT', 'RIGHT', 'UP', 'DOWN' and 'SAME', wherein 'SAME' indicates that the keywords and the corresponding information are positioned in the SAME text detection box; the third element is a regular matching expression of required information, and the matched information field is used as a value of a third-layer key value pair and written into an output dictionary.
In the information comparison between certificates: in the process of comparing certificate information with incoming vehicle information and face information, a key of a first layer is matched to represent a configuration file, a key of a second layer is the same as a key of a third layer in the process of comparing certificate information of certificates, the value of the key of the third layer is a list, elements in the list are certificate names, and the fact that the certificates need to be compared with the field is represented.
The fields that need to be aligned and which certificates need to have the fields in them consistent are described.
For example:
matching:
license plate:
semitrailer [ road transport certificate of people's republic of China-semitrailer, motor vehicle driving certificate of people's republic of China-semitrailer ]
Tractor (road transport card of people's republic of China-tractor, motor vehicle driving card of people's republic of China-tractor)
The outermost key "matches" indicates that this is the second part of the configuration information. The key 'license plate' of the second layer needs to be correspondingly consistent with the key of the third layer of the first part, and represents a field needing to be matched. The third layer of keys is used for distinguishing when the same field belongs to a plurality of different objects, for example, two license plates in the scene respectively belong to a semi-trailer and a tractor, the third layer of keys is used for distinguishing, the value of the third layer of keys is a list, the elements in the list are certificate names, and the certificate names show that the fields need to be compared.
In the face feature extraction: the face recognition model packages the whole face recognition module and provides an interface for the OCR model to use; the face registration is used for detecting a face in the certificate, calculating a face feature vector and storing the face feature vector in an OCR result dictionary; the face comparison interface inputs the image and an OCR recognition result dictionary, detects the face in the image, and when the face feature vectors temporarily stored in the OCR recognition result are matched, the comparison is successful; the OCR model comprises text detection and recognition, information extraction, information comparison and calling of a face recognition interface.
The certificate names listed in the part indicate that the face needs to be detected and feature vectors are extracted from the certificates so as to be compared with the face on site for verification.
For example:
face:
[ driver's license for road transportation of dangerous goods, driver's license for motor vehicles in the people's republic of China, license for escort personnel for transportation of dangerous goods ]
The part has a very simple structure, and the outermost key 'face' represents that the part is the configuration information of the third part in terms of one layer of key value pairs. The value is a list, the list elements are certificate names, and the certificate names represent that the certificates need to be identified and face feature vectors need to be detected and extracted so as to be used for comparing the faces on site.
The implementation logic of the code comprises a face recognition model, an OCR model, API packaging and a flash service. And the API packages and instantiates an OCR model, packages functions of the OCR object needing to interact with the outside into an interface, and prompts input abnormity. And writing a flash server and providing an http server interface. And calling a certificate matching interface after the recognition is finished, instantiating an OCR model by API packaging, packaging functions of OCR objects needing to interact with the outside into an interface, and prompting input abnormity.
Specifically, the method comprises the following steps:
the code theme is divided into four parts: face recognition model, OCR model, API package, flash service. The logic of primary recognition, extraction, and alignment is embodied in the OCR model.
The face recognition model encapsulates the whole face recognition module and provides an interface for the OCR model to use. There are two main interfaces: face registration and face comparison. The face registration is used for detecting a face in the certificate, calculating a face feature vector and storing the face feature vector in an OCR result dictionary. And the face comparison interface inputs the image and the OCR recognition result dictionary, detects the face in the image, and if the face can be matched with the face feature vector temporarily stored in the OCR recognition result, the comparison is successful.
The OCR model comprises text detection and recognition, information extraction and information comparison and also comprises the call of a face recognition interface.
In the code implementation of the part, firstly, an initialization interface is provided, calling the interface reads the configuration file, and an initialization dictionary is generated according to the first part of outermost keys of the configuration file and is used as the initial value of the final return value of the OCR. The structure example is as follows: { "certificate name 1": None, "certificate name 2": None, "certificate name 3": None }.
Then an interface for identifying a single certificate is called, an image is input, OCR detects the identified deep learning model, and the surrounding frame coordinates of the text behavior unit and the text content character string are output. The bounding box coordinates contain eight values, starting from the top left vertex (x, y) and going clockwise through each vertex of the quadrilateral. Then, searching and matching are carried out according to the key 'certificate name' on the third layer of the configuration file, the certificate is found, and then a recognition result dictionary of the certificate is generated according to the rest keys on the third layer of the configuration file, and the structure example is as follows: { "certificate name 1 {" license plate: "None," validity period: "None }," certificate name 2 ": None," certificate name 3 ": None }. Structured information is then extracted based on the value of the third level key of the profile, i.e. the list representing the lookup logic. Firstly, the first element of the regular matching list is a keyword, then the orientation of the required information relative to the keyword is determined according to the second element of the list, a surrounding frame possibly existing in the information is found according to the relative position of the quadrilateral frame, and finally the information is searched regularly in the candidate surrounding frame. Some information may have multiple values, such as validity periods that are signed multiple times on a document, and thus the values of the structured key-value pairs are stored in a unified list. And finally, deleting the date earlier than the current time through a filtering function, and converting the rest unified output formats into a character string 'YYYY-MM-DD'.
And after the identification is finished, calling the certificate matching interface, and automatically giving a comparison result of the information among all certificates according to the second part of the configuration file.
In addition, a face comparison interface and a license plate comparison interface are provided, the input is a face or license plate picture shot on site, the OCR recognition result is compared with the field, and the comparison result is returned.
The API package can instantiate an OCR model, package functions of OCR objects needing to interact with the outside into interfaces, and perform friendly prompting on some input exceptions (for example, a face photo does not contain a face).
And finally, writing a flash server and providing an http server interface.
The method adopts an OCR deep learning algorithm, and has high robustness; the adopted mode is used for extracting the structured information, so that the flexibility is high, and the method is not limited by the format; the identified and matched logic is stripped from the code to form a configuration file, and the modification is flexible. The method adds the verification links of certificate face and field photo in the whole process, uses three cascade models of face detection, face key point detection and 128-dimensional feature vector extraction, and completes two tasks of detection and identification.
It is to be understood that the above description is not intended to limit the present invention, and the present invention is not limited to the above examples, and those skilled in the art may make modifications, alterations, additions or substitutions within the spirit and scope of the present invention.
Claims (10)
1. The LNG vehicle station-entering certificate verification full-process post-processing method based on deep learning is characterized by comprising the following steps of:
s1, extracting certificate information of each certificate, wherein the certificate information comprises a certificate name and information keywords;
s2, comparing the certificate information of each certificate, including determining comparison fields and correlated certificate names;
s3, comparing certificate information with incoming vehicles to obtain information and face information;
the configuration file is in a yaml format, is indented and aligned, and is converted into a dictionary in a key-value pair format after being read by a python program; the implementation logic of the code comprises a face recognition model, an OCR model, API packaging and flash service; the API packages and instantiates an OCR model, packages functions of OCR objects needing to interact with the outside into interfaces, and prompts input abnormity; and writing a flash server and providing an http server interface.
2. The deep learning-based LNG vehicle arrival evidence verification full-process post-processing method as claimed in claim 1, wherein the LNG vehicle arrival evidence comprises: the characteristic equipment uses registration certificate, the card of road transportation of the people's republic of China-semitrailer, the card of road transportation of people's republic of China-tractor, the card of driver of road transportation of dangerous goods, the card of escort personnel of road transportation of dangerous goods, the card of driving motor vehicles of people's republic of China-semitrailer, the card of driving motor vehicles of people's republic of China-tractor.
3. The LNG vehicle station-entering certificate verification full-process post-processing method based on deep learning of claim 2, wherein certificate information extracted by the feature device using a registration certificate comprises a license plate and a device variety; certificate information extracted by the road transport certificate-semitrailer of the people's republic of China comprises a validity period, an operation range and a license plate; certificate information extracted by the road transport certificate-tractor of the people's republic of China comprises a validity period, an operation range and a license plate; certificate information extracted by the road dangerous goods transport driver certificate comprises a validity period, a name and a certificate number; certificate information extracted by the road dangerous goods transportation escort certificate comprises an expiration date, a name and a certificate number; certificate information extracted from the motor vehicle driving license of the people's republic of China comprises an expiration date, a name and a certificate number; certificate information extracted by a semitrailer which is a motor vehicle driving certificate of the people's republic of China comprises an expiration date and a license plate; the certificate information extracted by the motor vehicle driving certificate-tractor of the people's republic of China comprises the validity period and the license plate.
4. The deep learning-based LNG vehicle inbound certificate verification full-process post-processing method as claimed in claim 3, wherein the associated certificate specifically comprises: the license plate numbers of the road transportation certificate-semitrailer of the people's republic of China and the motor vehicle driving certificate-semitrailer of the people's republic of China are related; the license plate numbers of the road transportation certificate-tractor of the people's republic of China and the motor vehicle driving certificate-tractor of the people's republic of China are related; the road dangerous goods transport driver's license is related to the name, face and license number of the people's republic of China's motor vehicle license.
5. The LNG vehicle station-entering certificate verification full-process post-processing method based on deep learning of claim 4, wherein the comparison of certificate information and incoming vehicle information and face information specifically comprises: the license plate number of the incoming vehicle is consistent with the license plate number of the certificate; the scene shooting of the vehicle related personnel is matched with the certificate photo.
6. The LNG vehicle station-entering certificate verification full-process post-processing method based on deep learning of claim 5, wherein the key is a 'certificate' and the value is a dictionary object; the certificate name is identified, the keywords matched with the certificate, the information keywords needing to be extracted from the certificate, the direction of the information relative to the keywords and the information regular expression are obtained.
7. The LNG vehicle station-entering certificate verification full-process post-processing method based on deep learning of claim 6, wherein in the comparison process of certificate information of certificates, a first layer key is 'certificate' and represents configuration information of a first part; the second layer of keys are 'special equipment use registration certificates' which represent certificate names to be identified; the third layer of keys are specific structured information matching rules.
8. The LNG vehicle station entering certificate verification full-process post-processing method based on deep learning of claim 6, wherein in the comparison process of certificate information and entering vehicle information and face information, a key of a first layer is "match" to indicate a configuration file, a key of a second layer, namely "license plate", is the same as a key of a third layer in the comparison process of certificate information of certificates, the value of the key of the third layer is a list, and elements in the list are certificate names to indicate that the certificates need to be compared with the field.
9. The LNG vehicle arrival evidence verification full-flow post-processing method based on deep learning of claim 6, wherein the face recognition model encapsulates an entire face recognition module and provides an interface for an OCR model to use; the face registration is used for detecting a face in the certificate, calculating a face feature vector and storing the face feature vector in an OCR result dictionary; the face comparison interface inputs the image and an OCR recognition result dictionary, detects the face in the image, and when the face feature vectors temporarily stored in the OCR recognition result are matched, the comparison is successful;
the OCR model comprises text detection and recognition, information extraction, information comparison and calling of a face recognition interface.
10. The LNG vehicle arrival certificate verification full-flow post-processing method based on deep learning of claim 6, wherein a certificate matching interface is called after recognition is finished, an API encapsulation instantiates an OCR model, functions of OCR objects needing to interact with the outside are encapsulated into an interface, and input abnormity is prompted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110876381.XA CN113591715A (en) | 2021-07-29 | 2021-07-29 | LNG vehicle station-entering certificate verification full-process post-processing method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110876381.XA CN113591715A (en) | 2021-07-29 | 2021-07-29 | LNG vehicle station-entering certificate verification full-process post-processing method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113591715A true CN113591715A (en) | 2021-11-02 |
Family
ID=78253342
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110876381.XA Pending CN113591715A (en) | 2021-07-29 | 2021-07-29 | LNG vehicle station-entering certificate verification full-process post-processing method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113591715A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022418A (en) * | 2016-05-17 | 2016-10-12 | 深圳泰首智能技术有限公司 | Ticket-certificate-person-combined comparison device |
| CN108730726A (en) * | 2018-08-08 | 2018-11-02 | 王清泉 | Driver's certificate compares all-in-one machine |
| CN110321804A (en) * | 2019-06-11 | 2019-10-11 | 福建亿安智能技术有限公司 | One kind is combined based on face and electronic Car license recognition and carries out identity authentication method |
| CN111461122A (en) * | 2020-05-18 | 2020-07-28 | 南京大学 | Certificate information detection and extraction method |
| CN111738239A (en) * | 2020-08-18 | 2020-10-02 | 江苏神彩科技股份有限公司 | Dangerous waste transfer people and vehicle information verification system |
| CN112669049A (en) * | 2020-12-31 | 2021-04-16 | 车主邦（北京）科技有限公司 | Method and device for verifying logistics vehicle certificate information |
| CN112699775A (en) * | 2020-12-28 | 2021-04-23 | 中国平安人寿保险股份有限公司 | Certificate identification method, device and equipment based on deep learning and storage medium |
-
2021
- 2021-07-29 CN CN202110876381.XA patent/CN113591715A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022418A (en) * | 2016-05-17 | 2016-10-12 | 深圳泰首智能技术有限公司 | Ticket-certificate-person-combined comparison device |
| CN108730726A (en) * | 2018-08-08 | 2018-11-02 | 王清泉 | Driver's certificate compares all-in-one machine |
| CN110321804A (en) * | 2019-06-11 | 2019-10-11 | 福建亿安智能技术有限公司 | One kind is combined based on face and electronic Car license recognition and carries out identity authentication method |
| CN111461122A (en) * | 2020-05-18 | 2020-07-28 | 南京大学 | Certificate information detection and extraction method |
| CN111738239A (en) * | 2020-08-18 | 2020-10-02 | 江苏神彩科技股份有限公司 | Dangerous waste transfer people and vehicle information verification system |
| CN112699775A (en) * | 2020-12-28 | 2021-04-23 | 中国平安人寿保险股份有限公司 | Certificate identification method, device and equipment based on deep learning and storage medium |
| CN112669049A (en) * | 2020-12-31 | 2021-04-16 | 车主邦（北京）科技有限公司 | Method and device for verifying logistics vehicle certificate information |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10546206B2 (en) | Methods for mobile image capture of vehicle identification numbers in a non-document | |
| Yang et al. | A framework for improved video text detection and recognition | |
| EP2671190B1 (en) | System for data extraction and processing | |
| CN111340020A (en) | Formula identification method, device, equipment and storage medium | |
| CN109635633A (en) | Electronic device, bank slip recognition method and storage medium | |
| CN112926700B (en) | Class identification method and device for target image | |
| CN113496212A (en) | Text recognition method and device for box-type structure and electronic equipment | |
| CN116246294B (en) | Image information identification method, device, storage medium and electronic equipment | |
| CN113591715A (en) | LNG vehicle station-entering certificate verification full-process post-processing method based on deep learning | |
| US8472719B2 (en) | Method of stricken-out character recognition in handwritten text | |
| Siddique et al. | Development of an automatic vehicle license plate detection and recognition system for Bangladesh | |
| CN115294593A (en) | Image information extraction method and device, computer equipment and storage medium | |
| CN114443834A (en) | Method and device for extracting license information and storage medium | |
| CN114296545A (en) | Unmanned aerial vehicle gesture control method based on vision | |
| CN113704474A (en) | Bank outlet equipment operation guide generation method, device, equipment and storage medium | |
| JPH11328315A (en) | Character recognizing device | |
| JP5657401B2 (en) | Document processing apparatus and document processing program | |
| CN111613225A (en) | Method and system for automatically reporting road violation based on voice and image processing | |
| Fan et al. | Joint segmentation and recognition of license plate characters | |
| CN116844182A (en) | Card character recognition method for automatically recognizing format | |
| CN110569836B (en) | Variable-length character string identification method and device | |
| KR20220097844A (en) | Apparatus for processing unstructured document capable of extracting units and unit value and method thereof | |
| CN117079301A (en) | Certificate text detection method and system | |
| CN117197798A (en) | Invoice information extraction method, invoice information extraction device, invoice information extraction equipment and invoice information storage medium | |
| JPH04289989A (en) | Roman letter recognizing device |
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 |