CN106874901A - A kind of driving license recognition methods and device - Google Patents
A kind of driving license recognition methods and device Download PDFInfo
- Publication number
- CN106874901A CN106874901A CN201710030824.7A CN201710030824A CN106874901A CN 106874901 A CN106874901 A CN 106874901A CN 201710030824 A CN201710030824 A CN 201710030824A CN 106874901 A CN106874901 A CN 106874901A
- Authority
- CN
- China
- Prior art keywords
- driving license
- region
- dicode
- character
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
- G06V10/225—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition based on a marking or identifier characterising the area
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/63—Scene text, e.g. street names
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/24—Character recognition characterised by the processing or recognition method
- G06V30/242—Division of the character sequences into groups prior to recognition; Selection of dictionaries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
A kind of driving license recognition methods and device are the embodiment of the invention provides, wherein, methods described includes：Receive and pre-process driving license image；Determine the position in seal region on driving license image；According to the position relationship in seal region and driving license dicode region, and seal region position, coarse positioning driving license dicode region；By the contrast of the pixel gray value of Gray Projection and character and background, fine positioning by coarse positioning driving license dicode region；It is in shape and preset shape rule according to each single character, character row region is split, the single character picture after being split；By depth character recognition model, single character picture is identified, obtains the corresponding driving license dicode character string of driving license image.By present invention method and device, the recognition accuracy of driving license is improved.
Description
Technical field
The present invention relates to technical field of computer vision, more particularly to a kind of driving license recognition methods and device.
Background technology
In vehicle insurance finance services, Second-hand Vehicle Transaction business, palm such as insure at the application, car owner's driving license information is directed to
Typing, because driving license is a kind of certificate without chip, can only manually typing, if manually go be input into vehicle identification
Code and engine information, speed is very slow, and poor user experience, and efficiency is low.In order to improve the speed in input driving license information
And accuracy, OCR (Optical Character Recognition, optical character identification) technology is gradually applied to traveling
In the identification of card information, to meet application demand, preferably experience is brought to user.
But, existing driving license recognition methods is a profit based on general OCR frameworks or offline OCR graders
The character picture generated with printable character image (such as regular script/Song typeface) or computer is trained, and obtains information on driving license
Corresponding template, subsequently obtains the corresponding figure of information on driving license by carrying out interception to driving license image according to template matches
Picture, and then the identification of information on driving license is carried out according to OCR identifications, so that the rough fixed of driving license information can only be realized
Position, identification so that recognition accuracy is low；Meanwhile, existing driving license recognition methods will to natural conditions such as image angle/illumination
Ask strict so that the recognition effect for natural environment downward driving card image is not good enough, and recognition accuracy is low.
The content of the invention
The purpose of the embodiment of the present invention is to provide a kind of driving license recognition methods and device, to improve the identification of driving license
Accuracy rate.Concrete technical scheme is as follows：
On the one hand, a kind of driving license recognition methods is the embodiment of the invention provides, including：
Receive and pre-process driving license image；
Determine the position in seal region on the driving license image；
According to the position relationship in seal region and driving license dicode region, and the seal region position, coarse positioning
Driving license dicode region, wherein, the driving license dicode region includes VIN and engine number；
By the pixel gray value and the contrast of background of Gray Projection and character, fine positioning is by coarse positioning
Driving license dicode region, wherein, fine positioning includes by the driving license dicode region of coarse positioning：Isolated from driving license image
Character row region where driving license dicode；
It is in shape and preset shape rule according to each single character, the character row region is split,
Single character picture after being split；
By depth character recognition model, the single character picture is identified, obtains the driving license image pair
The driving license dicode character string answered.
Optionally, the position for determining seal region on the driving license image, including：
The salient region on the driving license image is detected, wherein, the salient region includes preset shape and face
The salient region that color is protruded；
The support vector machine classifier completed using training, from salient region described at least one, filters out classification
The maximum region of confidence level is the seal region, wherein, the support vector machine classifier is according to the driving license image
Color histogram feature and histograms of oriented gradients feature, training obtains.
Optionally, the position relationship according to seal region and driving license dicode, and the seal region position,
Driving license dicode region described in coarse positioning, including：
In position relationship from the seal region for pre-building with driving license dicode region, the seal region and row are obtained
Sail the position relationship in card dicode region；
According to the position relationship and the position in the seal region, driving license dicode region described in coarse positioning.
Optionally, the contrast of the pixel gray value by Gray Projection and character and background, fine positioning warp
The driving license dicode region of coarse positioning is crossed, including：
Horizontal direction projection is carried out to the driving license dicode region, in character row and the corresponding position of character row interval region
Put, respectively obtain maximum pixel and select the corresponding crest of gray value and the corresponding trough of minimum image vegetarian refreshments gray value, to the ripple
Peak and the trough carry out vertical direction projection, and rough segmentation separates out the first character row region；
According to character and the contrast of background, and using the extreme value of pixel gray level value changes, determine the individual of the extreme value
Number is the second candidate characters row region more than the first character row region of predetermined number；
The character classifier completed using training, from the second candidate characters row region described at least one, is filtered out point
The maximum region of class confidence level is the second character row region；Wherein, the character classifier includes being instructed according to characteristics of image
Practice the character classifier for completing.
Optionally, described is in shape and preset shape rule according to each single character, by the character row area
Domain is split, the single character picture after being split, including：
According to rectangle rule, the second character row region is split, obtained digital or alphabetical corresponding list
Individual character picture；
According to square rule, the second character row region is split, obtained the corresponding single character figure of Chinese character
Picture.
Optionally, it is described the single character picture is identified by depth character recognition model, finally give institute
The corresponding driving license dicode character string of driving license image is stated, including：
Respectively to single character picture each described, it is identified by the depth character recognition model, returns to each
The corresponding individual probability of recognition result of the recognition result of the single character picture and each single character picture；
According to the individual probability, the overall probability of the whole driving license image is identified；
Determine corresponding character string during the overall maximum probability, be the corresponding driving license dicode word of the driving license image
Symbol string.
Optionally, it is identified by the depth character recognition model described, returns to each described single character figure
The recognition result of picture, afterwards, methods described also includes：
Check the recognition result of the single character picture, if meet the dicode rule of driving license dicode；
When the recognition result of the single character picture does not meet the dicode rule, by the single character picture
Recognition result is adjusted to meet the result of the dicode rule；
The corresponding character string in the determination overall maximum probability, is the corresponding driving license of the driving license image
Dicode character string, afterwards, methods described also includes：
Check whether the result of the driving license dicode character string meets the dicode rule, in the driving license dicode word
When the result for according with string does not meet the dicode rule, the driving license dicode character string is re-recognized.
Optionally, the position relationship according to seal region and driving license dicode region, and the seal region
Position, driving license dicode region described in coarse positioning, including：
The result of the straight-line detection according to the seal region, determines the putting position of the driving license image；
Based on the putting position, layout position analysis is carried out to the driving license image, obtain the seal region with
The position relationship in driving license dicode region；
According to the position relationship and the position in the seal region, driving license dicode region described in coarse positioning.
On the other hand, the embodiment of the present invention additionally provides a kind of driving license identifying device, including：
Receiving processing module, for receiving and pre-processes driving license image；
Determining module, the position for determining seal region on the driving license image；
Coarse positioning module, for the position relationship according to seal region and driving license dicode region, and the seal area
The position in domain, coarse positioning driving license dicode region, wherein, the driving license dicode region includes VIN and engine
Number；
Fine positioning module, it is thin fixed for the pixel gray value by Gray Projection and character and the contrast of background
Position by coarse positioning driving license dicode region, wherein, fine positioning includes by the driving license dicode region of coarse positioning：From traveling
Character row region where isolating driving license dicode in card image；
Segmentation module, for being in shape and preset shape rule according to each single character, by the character row
Region is split, the single character picture after being split；
Identification module, for by depth character recognition model, being identified to single character picture, obtains driving license figure
As corresponding driving license dicode character string.
Optionally, the determining module includes：
Detection sub-module, for detecting the salient region on driving license image, wherein, salient region includes default shape
The salient region that shape and color are protruded；
Screening submodule, for the support vector machine classifier completed using training, from least one salient region,
It is seal region to filter out the maximum region of classification confidence level, wherein, support vector machine classifier is according to driving license image
Color histogram feature and histograms of oriented gradients feature, what training was obtained.
A kind of driving license recognition methods provided in an embodiment of the present invention and device, can according to the color of driving license image and
Shape, the seal region on positioning driving license image；According to the position in seal region, coarse positioning driving license dicode region passes through
Gray Projection, fine positioning driving license dicode region；According to the regular shape of character, character row region is split, divided
Single character picture after cutting；And then by depth character recognition model, each single character picture is identified, final
To the corresponding driving license dicode character string of driving license image.Improve the recognition accuracy of natural environment downward driving card.Certainly, implement
Any product of the invention or method must be not necessarily required to while reaching all the above advantage.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
The accompanying drawing to be used needed for having technology description is briefly described, it should be apparent that, drawings in the following description are only this
Some embodiments of invention, for those of ordinary skill in the art, on the premise of not paying creative work, can be with
Other accompanying drawings are obtained according to these accompanying drawings.
Fig. 1 is embodiment of the present invention driving license recognition methods flow chart；
Fig. 2 is the structural representation of embodiment of the present invention driving license identifying device.
Specific embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation is described, it is clear that described embodiment is only a part of embodiment of the invention, rather than whole embodiments.It is based on
Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under the premise of creative work is not made
Embodiment, belongs to the scope of protection of the invention.
Fig. 1 is embodiment of the present invention driving license recognition methods flow chart, and 1 couple of present invention of reference picture implements driving license identification side
Method is described in detail, including：
Step 101：Receive and pre-process driving license image.
When needing to be identified driving license, directly can be identified using actual driving license picture；Meanwhile,
Driving license image can be obtained by acquisition technology, be identified by driving license image, completed to actual traveling
The identification of card.Wherein, collection driving license image can be taken pictures completion by camera to actual driving license picture, certainly, this
It is not the unique implementation for gathering driving license image, the embodiment of the present invention is not limited the mode of IMAQ, any
The mode of driving license image can be collected all in embodiment of the present invention protection domain.
It should be noted that the driving license picture of the driving license image that will be gathered in the embodiment of the present invention and reality, is referred to as
It is driving license image.After driving license image is received, the driving license image is pre-processed first, wherein, to driving license
Image carries out pretreatment to be included：Drop is except noise, unitary of illumination and inclines rule just etc..
Step 102, determines the position in seal region on driving license image.
Include the number-plate number, vehicle, vehicle owner and VIN and engine mumber on driving license image
Etc. information, meanwhile, also including seal.By observing actual driving license it can be found that seal and the number-plate number, vehicle, all
The information such as people and VIN and engine mumber compare, very prominent.
So the embodiment of the present invention considers, the other information on driving license image is searched by seal.Seal region has
Significant feature, such as color, shape, Texture features etc..Specifically can be by these features of seal, from driving license image
On determine the position in seal region, and then according to the position in seal region, determine the information on the driving license image of needs
Position, be that accurately identifying for driving license provides good condition below.
Step 103, according to the position relationship in seal region and driving license dicode region, and seal region position, slightly
Positioning driving license dicode region, wherein, driving license dicode region includes VIN and engine number.
The position in seal region on driving license image can be determined according to above step, the position in seal region determines
Afterwards, according to the position relationship of other information on seal region and driving license, it is possible to determine the position of corresponding other information.
In actual application process, VIN and engine number are critically important information, are known by vehicle
Other code name and engine number can uniquely navigate to a specific car, and then to be handled insurance in real life, being repaired
And the provides convenients such as payment of breaking rules and regulations.Meanwhile, it can be found that VIN and engine number from actual driving license
Position and seal region position it is also more intimate, so embodiment of the present invention emphasis is directed to VIN and engine
Number is positioned, recognized.
It should be noted that on the position positioning driving license image for passing through seal region, in addition to driving license dicode region
Other information position method and driving license dicode regional location on driving license image is positioned by the position in seal region
Method be similar to, just repeat no more here.When needing to be identified the other information on driving license image, with reference to by seal
The location method in driving license dicode region, determines the position of other information on the position positioning driving license image in region, and then right
Other information is identified.
It should be noted that in natural environment, the putting position of driving license is different, to seal region on driving license image
, there is very big influence the positioning of position and the position according to seal region, the process in positioning driving license dicode region.Institute
In a kind of optional implementation method of the present invention, to be determined by the putting position first to driving license, according to driving license
The difference of putting position, coarse positioning driving license dicode region, specifically process, including：
First, the result of the straight-line detection according to seal region, determines the putting position of driving license image.
Then, the putting position based on driving license image, layout position analysis is carried out to driving license image, obtains seal area
Domain and the position relationship in driving license dicode region.
Finally, according to the position relationship in seal region and driving license dicode region, and seal region position, coarse positioning
Driving license dicode region.
For example, the seal region of the normal driving license put, the Chinese character in its seal region is from left to right arranged line by line, because
This is readily detected horizontal direction straight line.Driving license placing direction can so be judged, so according to the result of straight-line detection
Seal region and the relative position relation of whole driving license image are combined afterwards, judge the placing direction of driving license (such as with level
The angle of direction straight line is：0 degree, 90 degree, 180 degree or 270 degree etc.).
After determining the putting position of driving license image, then according to the putting position of actual driving license image, print is found
Chapter region and the relativeness in driving license dicode region, and then coarse positioning driving license dicode region.So that, the present invention is implemented
Example driving license recognition methods, is not only restricted to the influence that natural environment downward driving demonstrate,proves putting position, improves natural environment downward driving card
The accuracy rate of recognition result.
Step 104, by the pixel gray value and the contrast of background of Gray Projection and character, fine positioning is passed through
The driving license dicode region of coarse positioning, wherein, fine positioning includes by the driving license dicode region of coarse positioning：From driving license image
In isolate driving license dicode where character row region.
According to the position relationship in seal region and driving license dicode region, and seal region position, coarse positioning traveling
Card dicode region, is only the coarse positioning driving license dicode region from position.In order to more accurately position driving license dicode region
Position, by the feature fine positioning driving license dicode region of more detail, wherein, the feature of detail can be driving license
On image, the contrast of the background of the gray value of pixel and whole driving license image in driving license dicode region.
In addition, after navigating to the position in driving license dicode region, driving license dicode is isolated from driving license image, because
For driving license dicode is present in the form of character on driving license image, so the fine positioning driving license dicode for finally obtaining
Region as a result, character row region where isolating driving license dicode.
Step 105, is in shape and preset shape rule according to each single character, and character row region is carried out
Segmentation, the single character picture after being split.
It is above-mentioned obtain driving license dicode where character row region after, be exactly next character to all characters together
Row region is processed, and obtains single each character.Because each character is displayed on driving license image in certain
Shape, for example, can regard the square shape of Chinese character, digital rectangular in shape as.So being split by character row region
During, the shape that can be according to each character is split character row region, and then it is single to obtain each
Character picture.
Specifically, in a kind of optional embodiment of the embodiment of the present invention, according to each single character be in shape,
And preset shape rule, character row region is split, the single character picture after being split, including：
On the one hand, according to rectangle rule, the second character row region is split, obtains numeral or letter is corresponding
Single character picture.
On the other hand, according to square rule, the second character row region is split, is obtained the corresponding single word of Chinese character
Symbol image.
Specifically, can be distinguished by the different piece of the preset shape of preset shape rule definition and character row region
Matched, when result is matched, split, extract corresponding region and be single character, so repeated, until segmentation is obtained
All of single character.Or can be interval first with several pixel values, judge the shape of character row region different piece, then
Then compare whether the shape result for obtaining meets preset shape rule, if met, illustrate that the part is correct single
Character zone, and then, so repeat, until segmentation obtains all of single character. by the single character separation out.
Step 106, by depth character recognition model, is identified to single character picture, obtains driving license image pair
The driving license dicode character string answered.
After obtaining each single character picture, single character picture is input to the depth character recognition mould for training
Application OCR technique in type, you can return to recognition result.Wherein, depth character recognition model passes through CNN (Convolutional
Neural Network, convolutional neural networks) it is trained and obtains, depth character recognition mould is specifically obtained by CNN training
Type is prior art, is just repeated no more here；Meanwhile, OCR technique is the very ripe technology in driving license identification field, here also not
Repeat again.
Single character picture is identified, the corresponding driving license dicode character string of driving license image, i.e. driving license is obtained
Character row region on image where driving license dicode, complete identification, the result for obtaining be exactly separate car// know/not/generation/
Number/, hair/dynamic/machine/number/yard and the number of specific numeral and letter composition below.
Embodiment of the present invention driving license recognition methods, by determining the position in seal region on driving license image, according to print
Chapter region and the position relationship in driving license dicode region, coarse positioning go out driving license dicode region；Then by Gray Projection and
The pixel gray value of character and the contrast of background, fine positioning by coarse positioning driving license dicode region；Obtained in positioning
It is in that shape and preset shape are advised according to each single character after character row region where driving license dicode region
Then, character row region is split, the single character picture after being split is right finally by depth character recognition model
Single character picture is identified, and obtains the corresponding driving license dicode area characters string of driving license image, completes driving license identification.
By embodiment of the present invention driving license recognition methods, the recognition accuracy of driving license is improved.
In a kind of optional embodiment of the invention, the position in seal region on driving license image is determined, including：
First, the salient region on detection driving license image, wherein, salient region includes preset shape and color is prominent
The salient region for going out.
The salient region that preset shape and color are protruded on detection driving license image, wherein, preset shape can be just
It is square, because by the actual driving license of statistical observation it can be found that seal region is usually square, certainly, actual
In application process, preset shape can also be other shapes, as long as meet the regular shape in seal region, all in the present invention
In the protection domain of embodiment.Meanwhile, the seal on driving license in actual application is usually that to be different from other regions black
The red of color, so in a kind of optional implementation method of the embodiment of the present invention, salient region is to criticize square and red
Region.
Then, the support vector machine classifier for being completed using training, from least one salient region, filters out classification
The maximum region of confidence level is seal region, wherein, support vector machine classifier is the color histogram according to driving license image
Feature and histograms of oriented gradients feature, what training was obtained.
By above-mentioned detection salient region process, multiple doubtful seal regions can be obtained.From multiple doubtful seals
In region, the seal region of determination is further found, in addition it is also necessary to further treatment.
In the embodiment of the present invention, final positioning finds the seal region of determination from multiple doubtful seal regions, under
The step of mask body, completes, including：
The first step, extracts the color histogram feature and histograms of oriented gradients feature of multiple dimensioned driving license image,
The color histogram feature and histograms of oriented gradients feature are connected as characteristics of image.
Second step, by the characteristics of image, training svm (Support Vector Machine, SVMs) classification
Device.
3rd step, the stage is accurately positioned in seal, using the svm graders for training, to all of doubtful seal region
Screened, the maximum region of classification confidence level is chosen, as accurately seal regional location.
It is determined that after seal region, according to seal region and the position relationship in driving license dicode region, and seal area
The position in domain, coarse positioning driving license dicode region, including：
In position relationship from the seal region for pre-building with driving license dicode region, seal region and driving license are obtained
The position relationship in dicode region.
According to position relationship and the position in seal region, coarse positioning driving license dicode region.
Wherein, the position relationship in seal region and driving license dicode region, i.e. seal region and driving license dicode region
Relativeness, generally, seal region is to determine with the relativeness in driving license dicode region, for example, seal region
A how many pixel are driving license dicode regions afterwards, what this was all to determine, so after the position in seal region determines, according to seal
Region and the relativeness in driving license dicode region, you can obtain the position in driving license dicode region.
It is not difficult to find out from the description above, by the position in seal region, the position in positioning driving license dicode region obtains
Result be coarse positioning result.In order to the character row region comprising driving license dicode region is more precisely located, also need
Will processing procedure in further detail, specific processing procedure is as follows.
By the pixel gray value and the contrast of background of Gray Projection and character, fine positioning is by coarse positioning
Driving license dicode region, including：
First, horizontal direction projection is carried out to driving license dicode region, it is corresponding in character row and character row interval region
On position, respectively obtain maximum pixel and select the corresponding crest of gray value and the corresponding trough of minimum image vegetarian refreshments gray value, to crest
Vertical direction projection is carried out with trough, rough segmentation separates out the first character row region.
Character zone is present on driving license image, often with obvious texture or marginal information, to coarse positioning
Image comprising driving license dicode region, carries out horizontal direction projection, in character row and the corresponding position of character row interval region
On, maximum pixel can be respectively obtained and select the corresponding crest of gray value and the corresponding trough of minimum image vegetarian refreshments gray value, subsequently
Vertical direction projection is carried out to crest and trough, you can isolate character row region roughly.
Then, according to character and the contrast of background, and using the extreme value of pixel gray level value changes, the individual of extreme value is determined
Number is the second candidate characters row region more than the region of predetermined number.
In addition, character zone is present on driving license image, not only with obvious texture or marginal information, and
And have very big contrast on driving license image, so on the basis of above-mentioned, pixel gray level value changes can also be utilized
Extreme value, the number for determining extreme value is character row region more than the region of predetermined number.Obtained by Gray Projection to above-mentioned
Character row area results carry out further accurate.
Finally, the character classifier for being completed using training, from least one second candidate characters row regions, is filtered out point
The maximum region of class confidence level is the second character row region；Wherein, character classifier includes being trained according to characteristics of image
Into character classifier.
Handmarking is carried out by substantial amounts of driving license, character classifier is obtained using character feature training, and then will
Above by Gray Projection and the contrast of character and background, the character row region for obtaining carries out final positioning.
The purpose of embodiment of the present invention driving license recognition methods be exactly in order to be more accurately identified to driving license, and
It is significant process therein to orient driving license dicode picture, and the embodiment of the present invention is by combining Gray Projection, character and background
Contrast and training complete character classifier, driving license dicode region is positioned progressively so that traveling is precisely located
Card dicode region, and then for finally accurately driving license identification provides important basis.
Fine positioning driving license dicode region, i.e., the character row area where isolating driving license dicode from driving license image
Domain, is next accomplished by splitting whole character row region, single character picture is obtained, specifically, according to each
Single character is in shape and preset shape rule, character row region is split, the single character figure after being split
Picture, is introduced the content in detail above, is just repeated no more here.
Character row region is split, the single character picture after being split, next need completion is exactly most
Identification process afterwards, specifically, in a kind of optional embodiment of the embodiment of the present invention, by depth character recognition model,
Single character picture is identified, the corresponding driving license dicode character string of driving license image is finally given, including：
The first step, respectively to each single character picture, is identified by depth character recognition model, returns to each list
The corresponding individual probability of recognition result of the recognition result of individual character picture and each single character picture.
Second step, according to individual probability, is identified the overall probability of whole driving license image.
3rd step, it is determined that corresponding character string during overall maximum probability, is the corresponding driving license dicode word of driving license image
Symbol string.
To being split by character row region, all single character picture for obtaining is identified, and can obtain each
The recognition result of single character picture；Meanwhile, in identification process, different probability is pressed to a specific single character, can
To be identified as different characters, so, while returning again to recognition result, it is tool to return the image recognition of the single character
The probability of body character.Same treatment is carried out to each single character, the identification of all single characters can be finally obtained
As a result and it is identified as the probability of corresponding character.
Then, by each single character recognition into all probability of corresponding character, it is identified entirely by computing
The overall probability of driving license image, specific calculating process can simply be added all of individual probability, or calculate
The weighted sum of all individual probabilitys, specific weight can as needed be chosen during actual use.Certainly, calculate
Recognize that the computing of the overall probability of whole driving license image is not limited to both modes mentioned here, it is any to calculate knowledge
The mode of the overall probability of not whole driving license image is all allowed.
It is different specific characters for each single character recognition, calculates the corresponding whole driving license image of identification
Overall probability, it is final to determine corresponding character string during overall maximum probability, be the corresponding driving license dicode word of driving license image
Symbol string, completes the identification process of whole driving license dicode.
It should be noted that in actual identification process, the character for inevitably identifying is driving license
The character that dicode can not possibly occur in practice, in order that recognition result is more accurate, the embodiment of the present invention will also be tied to identification
Fruit optimizes.Specifically optimization process can after to each single character recognition, that is, check whether identification mistake；Or
Person can also whole identification process after the completion of, whole recognition result is optimized.Specifically, by depth word
Symbol identification model is identified, and returns to the recognition result of each single character picture, afterwards, including：
Check the recognition result of single character picture, if meet the dicode rule of driving license dicode；
When the recognition result of single character picture does not meet dicode rule, the recognition result of single character picture is adjusted
To meet the result of dicode rule.
In such as identification of the vehicle, first position represents production country or area code, is only
One of 123469wtjsklvryz, if recognition result is not character therein, illustrates that recognition result is wrong, then again
Identification, tuning is carried out to result.
Similarly, it is determined that corresponding character string during overall maximum probability, is the corresponding driving license dicode of driving license image
Character string, afterwards, including：
Check driving license dicode character string result whether meet dicode rule, driving license dicode character string result not
When meeting dicode rule, driving license dicode character string is re-recognized.For example, checking identification of the vehicle, the corresponding character for obtaining
The recognition result of string, if meeting has 17 characters etc., and specific driving license dicode rule is ripe those skilled in the art
The technology known, just repeats no more here.
In a kind of optional embodiment of the present invention, embodiment of the present invention driving license recognition methods can be applied to by client
In the system of end and back-end server composition.Client is based on ios system developments, and user selects specific work by client
Pattern, specific mode of operation can include：Standard driving license identification, rotation driving license identification, driving license picture collection with
And recognition result display isotype.Back-end server runs embodiment of the present invention recognition methods.Such back-end server, for row
Card dicode identification scene is sailed, algorithm customization is carried out, accuracy of identification is high, and recognition speed is fast；Client need not perform specific meter
Calculation process, it is only necessary to pay task to the back-end so that client is light, and complicated calculations give server process, low-power consumption.Together
When, embodiment of the present invention driving license recognition methods is easy to extension, is easy to expand to other entries to be identified, or card to be identified
Part.
In addition, the embodiment of the present invention additionally provides a kind of driving license identifying device, Fig. 2 knows for embodiment of the present invention driving license
The structural representation of other device, reference picture 2 is described in detail to embodiment of the present invention driving license identifying device.Including：
Receiving processing module 201, for receiving and pre-processes driving license image.
Determining module 202, the position for determining seal region on driving license image.
Coarse positioning module 203, for the position relationship according to seal region and driving license dicode region, and seal region
Position, coarse positioning driving license dicode region, wherein, driving license dicode region include VIN and engine number.
Fine positioning module 204, for the pixel gray value by Gray Projection and character and the contrast of background,
Fine positioning by coarse positioning driving license dicode region, wherein, fine positioning includes by the driving license dicode region of coarse positioning：From
Character row region where driving license dicode is isolated in driving license image.
Segmentation module 205, for being in shape and preset shape rule according to each single character, by character row
Region is split, the single character picture after being split.
Identification module 206, for by depth character recognition model, being identified to single character picture, is travelled
The corresponding driving license dicode character string of card image.
Embodiment of the present invention driving license identifying device, by receiving processing module 201, determining module 202, coarse positioning module
203rd, fine positioning module 204, segmentation module 205 and identification module 206, determine the position in seal region on driving license image,
According to seal region and the position relationship in driving license dicode region, coarse positioning goes out driving license dicode region；Then thrown by gray scale
The pixel gray value of shadow and character and the contrast of background, fine positioning by coarse positioning driving license dicode region；Fixed
Position obtain driving license dicode where character row region after, be in shape and preset shape according to each single character
Rule, character row region is split, the single character picture after being split, finally by depth character recognition model,
Single character picture is identified, the corresponding driving license dicode character string of driving license image is obtained, driving license identification is completed, led to
Embodiment of the present invention driving license identifying device is crossed, the recognition accuracy of driving license is improved.
In a kind of optional embodiment of the embodiment of the present invention, determining module 202 in driving license identifying device, including：
Detection sub-module, for detecting the salient region on driving license image, wherein, salient region includes default shape
The salient region that shape and color are protruded.
Screening submodule, for the support vector machine classifier completed using training, from least one salient region,
It is seal region to filter out the maximum region of classification confidence level, wherein, support vector machine classifier is according to driving license image
Color histogram feature and histograms of oriented gradients feature, what training was obtained.
In a kind of optional embodiment of the embodiment of the present invention, coarse positioning module 203 in driving license identifying device, including：
First position relation obtains submodule, for the position from the seal region that pre-builds and driving license dicode region
In relation, the position relationship in seal region and driving license dicode region is obtained.
First coarse positioning submodule, for the position according to position relationship and seal region, coarse positioning driving license dicode
Region.
In a kind of optional embodiment of the embodiment of the present invention, fine positioning module 204 in driving license identifying device, including：
Crude separation submodule, for carrying out horizontal direction projection to driving license dicode region, in character row and character in the ranks
On the corresponding position in septal area domain, respectively obtain that maximum pixel selects the corresponding crest of gray value and minimum image vegetarian refreshments gray value is corresponding
Trough, vertical direction projection is carried out to crest and trough, and rough segmentation separates out the first character row region.
Character row submodule is determined, for the contrast according to character and background, and using pixel gray level value changes
Extreme value, the number for determining extreme value is the second candidate characters row region more than the first character row region of predetermined number.
Screening character row submodule, for the character classifier completed using training, from least one second candidate characters
In row region, it is the second character row region to filter out the maximum region of classification confidence level；Wherein, character classifier is included according to figure
The character classifier of completion is trained as feature.
In a kind of optional embodiment of the embodiment of the present invention, module 205 is split in driving license identifying device, including：
First segmentation submodule, for according to rectangle rule, the second character row region is split, obtain numeral or
The corresponding single character picture of person's letter.
Second segmentation submodule, for according to square rule, the second character row region being split, obtains Chinese character pair
The single character picture answered.
In a kind of optional embodiment of the embodiment of the present invention, identification module 206 in driving license identifying device, including：
Single result returns to submodule, for each single character picture, being entered by depth character recognition model respectively
Row identification, return each single character picture recognition result and each single character picture recognition result it is corresponding single
Probability.
Overall probability calculation submodule, for according to individual probability, being identified the overall probability of whole driving license image.
As a result determination sub-module, is that driving license image is corresponding for determining corresponding character string during overall maximum probability
Driving license dicode character string.
In a kind of optional embodiment of the embodiment of the present invention, driving license identifying device also includes：
Check module, the recognition result for checking single character picture, if meet the dicode rule of driving license dicode.
First adjusting module, for when the recognition result of single character picture does not meet dicode rule, by single character
The recognition result of image is adjusted to meet the result of dicode rule.
Whether the second adjusting module, the result for checking driving license dicode character string meets dicode rule, in driving license
When the result of dicode character string does not meet dicode rule, driving license dicode character string is re-recognized.
In a kind of optional embodiment of the embodiment of the present invention, coarse positioning module 203, also wraps in driving license identifying device
Include：
Putting position determination sub-module, for the result of the straight-line detection according to seal region, determines driving license image
Putting position.
Second place relation obtains submodule, for based on putting position, layout position analysis being carried out to driving license image,
Obtain the position relationship in seal region and driving license dicode region.
Second coarse positioning submodule, for the position according to position relationship and seal region, coarse positioning driving license dicode
Region.
It should be noted that the device of the embodiment of the present invention is using the device of above-mentioned driving license recognition methods, then it is above-mentioned
All embodiments of driving license recognition methods are applied to the device, and can reach same or analogous beneficial effect.
It should be noted that herein, such as first and second or the like relational terms are used merely to a reality
Body or operation make a distinction with another entity or operation, and not necessarily require or imply these entities or deposited between operating
In any this actual relation or order.And, term " including ", "comprising" or its any other variant be intended to
Nonexcludability is included, so that process, method, article or equipment including a series of key elements not only will including those
Element, but also other key elements including being not expressly set out, or also include being this process, method, article or equipment
Intrinsic key element.In the absence of more restrictions, the key element limited by sentence "including a ...", it is not excluded that
Also there is other identical element in process, method, article or equipment including the key element.
Each embodiment in this specification is described by the way of correlation, identical similar portion between each embodiment
Divide mutually referring to what each embodiment was stressed is the difference with other embodiment.Especially for system reality
Apply for example, because it is substantially similar to embodiment of the method, so description is fairly simple, related part is referring to embodiment of the method
Part explanation.
Presently preferred embodiments of the present invention is the foregoing is only, is not intended to limit the scope of the present invention.It is all
Any modification, equivalent substitution and improvements made within the spirit and principles in the present invention etc., are all contained in protection scope of the present invention
It is interior.
Claims (10)
1. a kind of driving license recognition methods, it is characterised in that including：
Receive and pre-process driving license image；
Determine the position in seal region on the driving license image；
According to the position relationship in seal region and driving license dicode region, and the seal region position, coarse positioning traveling
Card dicode region, wherein, the driving license dicode region includes VIN and engine number；
By the contrast of the pixel gray value of Gray Projection and character and background, fine positioning by coarse positioning traveling
Card dicode region, wherein, fine positioning includes by the driving license dicode region of coarse positioning：Traveling is isolated from driving license image
Character row region where card dicode；
It is in shape and preset shape rule according to each single character, the character row region is split, obtains
Single character picture after segmentation；
By depth character recognition model, the single character picture is identified, obtains the driving license image corresponding
Driving license dicode character string.
2. driving license recognition methods according to claim 1, it is characterised in that printed on the determination driving license image
The position in chapter region, including：
The salient region on the driving license image is detected, wherein, the salient region includes preset shape and color is prominent
The salient region for going out；
The support vector machine classifier completed using training, from salient region described at least one, filters out classification credible
It is the seal region to spend maximum region, wherein, the support vector machine classifier is the face according to the driving license image
Color Histogram feature and histograms of oriented gradients feature, what training was obtained.
3. driving license recognition methods according to claim 2, it is characterised in that described double according to seal region and driving license
Code position relationship, and the seal region position, driving license dicode region described in coarse positioning, including：
In position relationship from the seal region for pre-building with driving license dicode region, the seal region and driving license are obtained
The position relationship in dicode region；
According to the position relationship and the position in the seal region, driving license dicode region described in coarse positioning.
4. driving license recognition methods according to claim 3, it is characterised in that described by Gray Projection and character
Pixel gray value and background contrast, fine positioning by coarse positioning the driving license dicode region, including：
Horizontal direction projection is carried out to the driving license dicode region, in character row and the corresponding position of character row interval region
On, respectively obtain maximum pixel and select the corresponding crest of gray value and the corresponding trough of minimum image vegetarian refreshments gray value, to the crest
Vertical direction projection is carried out with the trough, rough segmentation separates out the first character row region；
According to character and the contrast of background, and using the extreme value of pixel gray level value changes, determine that the number of the extreme value surpasses
The the first character row region for crossing predetermined number is the second candidate characters row region；
The character classifier completed using training, from the second candidate characters row region described at least one, filtering out classification can
The maximum region of reliability is the second character row region；Wherein, the character classifier includes being trained according to characteristics of image
Into character classifier.
5. driving license recognition methods according to claim 4, it is characterised in that described to be according to each single character
Shape and preset shape rule, the character row region are split, the single character picture after being split, including：
According to rectangle rule, the second character row region is split, obtained numeral or the corresponding single word of letter
Symbol image；
According to square rule, the second character row region is split, obtained the corresponding single character picture of Chinese character.
6. driving license recognition methods according to claim 1, it is characterised in that described by depth character recognition model,
The single character picture is identified, the corresponding driving license dicode character string of the driving license image is finally given, including：
Respectively to single character picture each described, it is identified by the depth character recognition model, is returned described in each
The corresponding individual probability of recognition result of the recognition result of single character picture and each single character picture；
According to the individual probability, the overall probability of the whole driving license image is identified；
Determine corresponding character string during the overall maximum probability, be the corresponding driving license dicode character of the driving license image
String.
7. driving license recognition methods according to claim 6, it is characterised in that described by the depth character recognition
Model is identified, and returns to the recognition result of each single character picture, and afterwards, methods described also includes：
Check the recognition result of the single character picture, if meet the dicode rule of driving license dicode；
When the recognition result of the single character picture does not meet the dicode rule, by the identification of the single character picture
Result is adjusted to meet the result of the dicode rule；
The corresponding character string in the determination overall maximum probability, is the corresponding driving license dicode of the driving license image
Character string, afterwards, methods described also includes：
Check whether the result of the driving license dicode character string meets the dicode rule, in the driving license dicode character string
Result when not meeting dicode rule, re-recognize the driving license dicode character string.
8. driving license recognition methods according to claim 1, it is characterised in that described double according to seal region and driving license
Code region position relationship, and the seal region position, driving license dicode region described in coarse positioning, including：
The result of the straight-line detection according to the seal region, determines the putting position of the driving license image；
Based on the putting position, layout position analysis is carried out to the driving license image, obtain the seal region with traveling
The position relationship in card dicode region；
According to the position relationship and the position in the seal region, driving license dicode region described in coarse positioning.
9. a kind of driving license identifying device, it is characterised in that including：
Receiving processing module, for receiving and pre-processes driving license image；
Determining module, the position for determining seal region on the driving license image；
Coarse positioning module, for the position relationship according to seal region and driving license dicode region, and the seal region
Position, coarse positioning driving license dicode region, wherein, the driving license dicode region includes VIN and Motor Number
Code；
Fine positioning module, for the pixel gray value by Gray Projection and character and the contrast of background, fine positioning warp
The driving license dicode region of coarse positioning is crossed, wherein, fine positioning includes by the driving license dicode region of coarse positioning：From driving license figure
Character row region where driving license dicode is isolated as in；
Segmentation module, for being in shape and preset shape rule according to each single character, by the character row region
Split, the single character picture after being split；
Identification module, for by depth character recognition model, being identified to single character picture, obtains driving license image pair
The driving license dicode character string answered.
10. traveling card device according to claim 9, it is characterised in that the determining module includes：
Detection sub-module, for detecting the salient region on driving license image, wherein, salient region include preset shape and
The salient region that color is protruded；
Screening submodule, for the support vector machine classifier completed using training, from least one salient region, screening
The maximum region of the confidence level that goes out to classify is seal region, wherein, support vector machine classifier is the color according to driving license image
Histogram feature and histograms of oriented gradients feature, what training was obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710030824.7A CN106874901B (en) | 2017-01-17 | 2017-01-17 | Driving license identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710030824.7A CN106874901B (en) | 2017-01-17 | 2017-01-17 | Driving license identification method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106874901A true CN106874901A (en) | 2017-06-20 |
| CN106874901B CN106874901B (en) | 2020-07-03 |
Family
ID=59157678
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710030824.7A Active CN106874901B (en) | 2017-01-17 | 2017-01-17 | Driving license identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106874901B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109145723A (en) * | 2018-07-09 | 2019-01-04 | 长江大学 | A kind of seal recognition methods, system, terminal installation and storage medium |
| CN109190623A (en) * | 2018-09-15 | 2019-01-11 | 闽江学院 | A method of identification projector brand and model |
| CN109409421A (en) * | 2018-10-09 | 2019-03-01 | 杭州诚道科技股份有限公司 | Motor vehicle, driver's archival image recognition methods based on convolutional neural networks |
| CN109919160A (en) * | 2019-03-04 | 2019-06-21 | 深圳先进技术研究院 | Method for recognizing verification code, device, terminal and storage medium |
| CN110472623A (en) * | 2019-06-29 | 2019-11-19 | 华为技术有限公司 | Image detecting method, equipment and system |
| CN110569801A (en) * | 2019-09-10 | 2019-12-13 | 四川星盾科技股份有限公司 | method for identifying key content of driving license |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101059836A (en) * | 2007-06-01 | 2007-10-24 | 华南理工大学 | Human eye positioning and human eye state recognition method |
| US20110075145A1 (en) * | 2007-10-01 | 2011-03-31 | Maskless Lithography, Inc. | Alignment system for optical lithography |
| CN103870803A (en) * | 2013-10-21 | 2014-06-18 | 北京邮电大学 | Vehicle license plate recognition method and system based on coarse positioning and fine positioning fusion |
| CN106156768A (en) * | 2016-07-01 | 2016-11-23 | 屈桢深 | The vehicle registration certificate detection method of view-based access control model |
-
2017
- 2017-01-17 CN CN201710030824.7A patent/CN106874901B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101059836A (en) * | 2007-06-01 | 2007-10-24 | 华南理工大学 | Human eye positioning and human eye state recognition method |
| US20110075145A1 (en) * | 2007-10-01 | 2011-03-31 | Maskless Lithography, Inc. | Alignment system for optical lithography |
| CN103870803A (en) * | 2013-10-21 | 2014-06-18 | 北京邮电大学 | Vehicle license plate recognition method and system based on coarse positioning and fine positioning fusion |
| CN106156768A (en) * | 2016-07-01 | 2016-11-23 | 屈桢深 | The vehicle registration certificate detection method of view-based access control model |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109145723A (en) * | 2018-07-09 | 2019-01-04 | 长江大学 | A kind of seal recognition methods, system, terminal installation and storage medium |
| CN109190623A (en) * | 2018-09-15 | 2019-01-11 | 闽江学院 | A method of identification projector brand and model |
| CN109190623B (en) * | 2018-09-15 | 2020-12-29 | 闽江学院 | Method for identifying brand and model of projector |
| CN109409421A (en) * | 2018-10-09 | 2019-03-01 | 杭州诚道科技股份有限公司 | Motor vehicle, driver's archival image recognition methods based on convolutional neural networks |
| CN109409421B (en) * | 2018-10-09 | 2021-12-07 | 杭州诚道科技股份有限公司 | Motor vehicle and driver archive image identification method based on convolutional neural network |
| CN109919160A (en) * | 2019-03-04 | 2019-06-21 | 深圳先进技术研究院 | Method for recognizing verification code, device, terminal and storage medium |
| CN109919160B (en) * | 2019-03-04 | 2021-03-23 | 深圳先进技术研究院 | Verification code identification method, device, terminal and storage medium |
| CN110472623A (en) * | 2019-06-29 | 2019-11-19 | 华为技术有限公司 | Image detecting method, equipment and system |
| CN110472623B (en) * | 2019-06-29 | 2022-08-09 | 华为技术有限公司 | Image detection method, device and system |
| CN110569801A (en) * | 2019-09-10 | 2019-12-13 | 四川星盾科技股份有限公司 | method for identifying key content of driving license |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106874901B (en) | 2020-07-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106874901A (en) | A kind of driving license recognition methods and device | |
| CN104751187B (en) | Meter reading automatic distinguishing method for image | |
| CN108446678B (en) | Dangerous driving behavior identification method based on skeletal features | |
| CN101038686B (en) | Method for recognizing machine-readable travel certificate | |
| CN102043945B (en) | License plate character recognition method based on real-time vehicle tracking and binary index classification | |
| CN104778470B (en) | Text detection based on component tree and Hough forest and recognition methods | |
| CN103325122B (en) | Based on the pedestrian retrieval method of Bidirectional sort | |
| CN103870803A (en) | Vehicle license plate recognition method and system based on coarse positioning and fine positioning fusion | |
| CN103413145B (en) | Intra-articular irrigation method based on depth image | |
| CN111126224A (en) | Vehicle detection method and classification recognition model training method | |
| CN106023220A (en) | Vehicle exterior part image segmentation method based on deep learning | |
| CN106446952A (en) | Method and apparatus for recognizing score image | |
| CN103903018A (en) | Method and system for positioning license plate in complex scene | |
| CN104766046A (en) | Detection and recognition algorithm conducted by means of traffic sign color and shape features | |
| CN103065138A (en) | Recognition method of license plate number of motor vehicle | |
| CN103544484A (en) | Traffic sign identification method and system based on SURF | |
| CN104217217A (en) | Vehicle logo detection method and system based on two-layer classification | |
| CN108537143B (en) | A kind of face identification method and system based on key area aspect ratio pair | |
| CN107480585A (en) | Object detection method based on DPM algorithms | |
| CN110263215A (en) | A kind of video feeling localization method and system | |
| CN108960221B (en) | Bank card identification method and device based on image | |
| CN108537169A (en) | A kind of high-resolution remote sensing image method for extracting roads based on center line and detection algorithm of having a lot of social connections | |
| CN107169496A (en) | A kind of character recognition method and device | |
| CN113269161A (en) | Traffic signboard detection method based on deep learning | |
| CN109389167A (en) | Traffic sign recognition method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |