CN106886778A - A kind of car plate segmentation of the characters and their identification method under monitoring scene - Google Patents

A kind of car plate segmentation of the characters and their identification method under monitoring scene Download PDF

Info

Publication number
CN106886778A
CN106886778A CN201710278593.1A CN201710278593A CN106886778A CN 106886778 A CN106886778 A CN 106886778A CN 201710278593 A CN201710278593 A CN 201710278593A CN 106886778 A CN106886778 A CN 106886778A
Authority
CN
China
Prior art keywords
character
characters
license plate
image
training
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
Application number
CN201710278593.1A
Other languages
Chinese (zh)
Other versions
CN106886778B (en
Inventor
柯逍
张毓峰
郭文忠
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fuzhou University
Original Assignee
Fuzhou University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Fuzhou University filed Critical Fuzhou University
Priority to CN201710278593.1A priority Critical patent/CN106886778B/en
Publication of CN106886778A publication Critical patent/CN106886778A/en
Application granted granted Critical
Publication of CN106886778B publication Critical patent/CN106886778B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • 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
    • G06V30/14—Image acquisition
    • G06V30/148—Segmentation of character regions
    • G06V30/158—Segmentation of character regions using character size, text spacings or pitch estimation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • 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
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
    • G06V10/507—Summing image-intensity values; Histogram projection analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/56—Extraction of image or video features relating to colour
    • 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
    • G06V20/625—License plates
    • 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)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Biomedical Technology (AREA)
  • Evolutionary Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Character Discrimination (AREA)
  • Image Analysis (AREA)

Abstract

The present invention relates to a kind of car plate segmentation of the characters and their identification method under monitoring scene, comprise the following steps：Step S1：Characters on license plate image is classified, is that the training of explicit features grader and implicit features grader is ready；Step S2：Input license plate image；Step S3：Character segmentation is carried out with the character segmentation method that domain method is combined is connected using projection localization method to license plate image；Step S4：The method blended using explicit features grader and implicit features grader is identified to characters on license plate.The present invention improves the effect of License Plate Character Segmentation using projection localization method with method that domain method is combined is connected, consider single problem for Recognition of License Plate Characters feature, propose the method based on explicit features grader Yu implicit features Multiple Classifier Fusion, both advantages in feature extraction can be combined, so as to improve the effect of character recognition.

Description

A kind of car plate segmentation of the characters and their identification method under monitoring scene
Technical field
The present invention relates to pattern-recognition and computer vision field, under particularly a kind of monitoring scene License Plate Character Segmentation with Recognition methods.
Background technology
As the quickening of safety China Reconstructs paces, monitoring camera distribution are increasingly wider, the resolution ratio of camera is more next It is higher, directly spread all over the monitors environment in streets and lanes compared with the vehicle image harvester using standard bayonet socket using these, Bring some new challenges.Traditional Vehicle License Plate Recognition System is at aspects such as highway toll, parking lots with extensive but most of The support that some special hardware equipment are required for scene, such as ground induction coil can only once be identified to a car, know Other speed is slow, and image transmitting uses analog signal mostly, and image level sense is not strong, and contrast is poor, in order to ensure resolution ratio Tend not to collect vehicle panoramic, so as to cause to meet the related service requirement of criminal investigation, public security.Under monitors environment scene Car plate detection based on dynamic video stream need not install other hardware devices with identifying system, can be simultaneously to the multiple in image Car plate is detected, not restricted by hardware and environment, efficiency high, multi-frame video image can be identified respectively, selected Confidence level highest is used as final result, so as to reduce influence of the single-frame images to recognition result, recognition accuracy is higher.But base Many challenges are also faced with the car plate detecting system under monitors environment scene, such as car plate angle change is big, long-term use It is image blurring that the dust of surface of camera head attachment can make to collect, and noise increases.These all bring to Recognition of License Plate Characters Huge challenge.In recent years, with the continuous hair of the technologies such as Computer Image Processing, artificial intelligence, pattern-recognition, transmission of video Exhibition, the car plate algorithm based on dynamic video stream is obtained increasingly in Criminal Investigation, the traffic accident quickly social life such as treatment It is widely applied.
Although domestic and foreign scholars also expand research to this one after another, it is proposed that some high levels and the strong car plate word of usability Symbol segmentation and recognizer, but it is still present that Character segmentation effect is undesirable, and Recognition of License Plate Characters feature considers single asking Topic.Upright projection split plot design algorithm is simple, and speed is fast, but more sensitive to noise, to the disconnected situation effect of characters on license plate It is poor；Connection domain method is insensitive to character deformation, illumination, weather environment, but to the disconnected situation effect ratio of character less It is preferable.Therefore we improve the effect of License Plate Character Segmentation using projection localization method with method that domain method is combined is connected.Pin Single, recognition accuracy situation not high is considered to character recognition feature, we use explicit features grader and implicit features The method that grader is blended.The artificial explicit features extracted based on priori have stronger specific aim, but extract The feature of the feature for arriving generally shallow-layer, and feature extraction too relies on the experience and subjective consciousness of people, the feature extracted Difference is very big on classification performance influence, or even the order of the feature extracted can also influence last classification performance, and is based on convolution The feature of the Machine self-learnings such as neutral net (CNN) can automatically learn the profound feature of image, it is to avoid artificial to explicit Feature is selected, automatically from training data learning feature.Implicit features are extracted and can reduce feature selecting to grader Influence, but the interpretation of feature extraction is poor, and feature selecting places one's entire reliance upon the selection of model.Methods herein can be combined Both advantages in feature extraction, so as to improve the effect of character recognition.
The content of the invention
In view of this, it is an object of the invention to provide a kind of car plate segmentation of the characters and their identification method under monitoring scene, with gram Take defect present in prior art.
The present invention is realized using following scheme：A kind of car plate segmentation of the characters and their identification method under monitoring scene, including it is following Step：
Step S1：Characters on license plate image is classified, is the training of explicit features grader and implicit features grader It is ready；
Step S2：Input license plate image；
Step S3：Line character is entered with the character segmentation method that domain method is combined is connected using projection localization method to license plate image Segmentation；
Step S4：The method blended using explicit features grader and implicit features grader is known to characters on license plate Not.
Further, in the step S3, characters on license plate is split as follows：
Step S31：Projection localization method first uses formula with the character segmentation method that domain method is combined is connected：F (i, j)=α R (i, j)+β G (i, j)+χ B (i, j)), wherein α=0.30 is made, β=0.59, χ=0.11 carries out at gray processing license plate image Reason；
Step S32：One width license plate image is carried out into mesh segmentation, the histogram of image, and normalizing is calculated each grid Change, calculate cumulative mean value mu, and global gray average, then calculate the probability q for being assigned to class AA, and the probability for being assigned to class B qB；Using formula：Sigma=qA*qB*(muA-muB)*(muA-muB) inter-class variance is calculated, it is maximum that inter-class variance is found in circulation It is worth, and writes down threshold value now, as optimal threshold, thresholding is finally carried out to grid using this optimal threshold, repeats this Individual process is finished until whole license plate image all binaryzations；
Step S33：Extract character outline；
Step S34：Do boundary rectangle operation；
Step S35：If meeting the number of the boundary rectangle of size less than 7, illustrate to there may be Characters Stuck or The situation that chinese character is lost in car plate；Part when Characters Stuck to adhesion carries out projection localization, the position of segmentation It is set to the trough for being closer to image midpoint；
Step S36：When chinese character is lost using by the anti-method for pushing away Chinese character of spcial character；Make word Character block of the Fu Kuai centers in the interval of car plate 1/7~2/7 is spcial character, and the left side of spcial character is Chinese character；
Step S37：If projection character block number is more than 7, illustrate after License Plate Segmentation that some characters may be divided into many Individual character block, thus further to projection character block merge treatment.
Further, in the step S4, characters on license plate is identified as follows：
Step S41：The convolutional neural networks that input training image is extracted to implicit features；
Step S42：It is input into training image to explicit features sorter network；
Step S43：The convolutional neural networks that training implicit features are extracted, the training of convolutional neural networks mainly includes two Stage：
First stage is the propagated forward stage：Propagated forward stage main process is one sample of taking-up from training sample This X (xp,yp) using training sample X as network input, by formula Ox=fn(...(f2(f1(XpW(1))W(2))...)W(n)) meter Calculate the reality output of training sample X；
Second stage is back-propagation phase：Back-propagation phase will calculate the reality output O of training sample XxWith reason Want to export YpDifference, the method according to minimization error is adjusted to the parameter of model parameter；
Step S44：One explicit features sorter network of training；Training explicit features sorter network includes three phases：
First stage is the extraction character feature stage：The stage is first rectified using Gamma orthosises to characters on license plate Just, regulation picture contrast, then the gradient of each pixel is calculated, profile information is obtained, character picture is then divided into n × n Junior unit, wherein n=6；Then the histogram of gradients of each junior unit is calculated, the Feature Descriptor of junior unit is obtained, then Again by junior unit according to 3 × 3 model split into m × n groups, every group of Feature Descriptor is together in series can be arrived the feature of group and retouch Son is stated, the Feature Descriptor of mn group is together in series and is obtained character feature Fexture1；By the n of character × n center matrix mark Non-character pixel, wherein n=8 are designated as, and are become low-resolution image, as character feature Fexture2, will Fexture1 and Fexture2 are together in series and obtain character feature；
Second stage is the training data preparatory stage：To every image zooming-out character feature of training data, and stamp All training images are organized into a matrix by class label；
Three phases are the training stage：Training matrix is input in SVMs, using RBF kernel functions to explicit Feature classifiers are trained；
Step S45：Image to be sorted is separately input in explicit features sorter network and implicit features sorter network, Obtain classification results；
Step S46：For each width test image I, the confidence level vector E=that implicit features extract network will be all obtained {e1,e2,...,eNAnd implicit features network confidence level vector E '={ e '1,e′2,...,e′N, wherein N is characters on license plate Class number；
Step S47：Obtain in step S45 corresponding | | the E | | of two vectors∞With | | E ' | |∞, the corresponding classifications of image IFor：Wherein i is | | E | |∞Corresponding classification, j is | | E ' | |∞Corresponding classification.
Compared to prior art, the invention has the advantages that：Projection localization method and connected domain that the present invention builds The character segmentation method that method is combined can improve the accuracy rate of Character segmentation.Set forth herein based on explicit features grader with The method of implicit features Multiple Classifier Fusion can either give full play to the artificial explicit features extracted based on priori to be had Stronger targetedly feature and based on convolutional neural networks (CNN), the machine such as sparse autocoder (AutoEncoder) from The feature of study can automatically learn the profound feature of image, it is to avoid manually explicit features are selected, automatically from instruction The characteristics of practicing data learning feature, the feature of the feature generally shallow-layer that explicit features extract, and feature can be avoided again The experience and subjective consciousness for too relying on people are extracted, the different of the feature extracted influence very big on classification performance, or even extract Feature order can also influence last classification performance shortcoming and implicit features extract interpretation it is poor, feature selecting Place one's entire reliance upon model selection shortcoming, improve the accuracy of character classification.
Brief description of the drawings
Fig. 1 is the flow chart of car plate segmentation of the characters and their identification method under monitoring scene of the present invention.
Specific embodiment
Below in conjunction with the accompanying drawings and embodiment the present invention will be further described.
The present embodiment provides a kind of car plate segmentation of the characters and their identification method under monitoring scene, as shown in figure 1, being directed to car plate word Symbol identification feature considers single problem, it is proposed that the method based on explicit features grader Yu implicit features Multiple Classifier Fusion. The artificial explicit features extracted based on priori have a stronger specific aim, but the feature extracted generally shallow-layer Feature, and feature extraction too relies on the experience and subjective consciousness of people, the different of the feature extracted influence very on classification performance Greatly, or even the order of feature extracted can also influence last classification performance, convolutional neural networks (CNN) are based on, it is sparse from The feature of the Machine self-learnings such as dynamic encoder (AutoEncoder) can automatically learn the profound feature of image, it is to avoid people Work is selected explicit features, automatically from training data learning feature.Implicit features are extracted and can reduce feature selecting pair The influence of grader, but the interpretation of feature extraction is poor, feature selecting places one's entire reliance upon the selection of model, methods herein Both advantages in feature extraction can be combined, so as to improve the effect of character recognition, with comprising the following steps：
Step S1：Characters on license plate image is classified, is the training of explicit features grader and implicit features grader It is ready；
Step S2：Input license plate image；
Step S3：Line character is entered with the character segmentation method that domain method is combined is connected using projection localization method to license plate image Segmentation；
Step S4：The method blended using explicit features grader and implicit features grader is known to characters on license plate Not.
In the present embodiment, in the step S3, characters on license plate is split as follows：
Step S31：Projection localization method first uses formula with the character segmentation method that domain method is combined is connected：F (i, j)=α R (i, j)+β G (i, j)+χ B (i, j)), wherein α=0.30 is made, β=0.59, χ=0.11 carries out at gray processing license plate image Reason；
Step S32：One width license plate image is carried out into mesh segmentation, the histogram of image, and normalizing is calculated each grid Change, calculate cumulative mean value mu, and global gray average, then calculate the probability q for being assigned to class AA, and the probability for being assigned to class B qB；Using formula：Sigma=qA*qB*(muA-muB)*(muA-muB) inter-class variance is calculated, it is maximum that inter-class variance is found in circulation It is worth, and writes down threshold value now, as optimal threshold, thresholding is finally carried out to grid using this optimal threshold, repeats this Individual process is finished until whole license plate image all binaryzations；
Step S33：Extract character outline；
Step S34：Do boundary rectangle operation；
Step S35：If meeting the number of the boundary rectangle of size less than 7, illustrate to there may be Characters Stuck or The situation that chinese character is lost in car plate；Part when Characters Stuck to adhesion carries out projection localization, the position of segmentation It is set to the trough for being closer to image midpoint；
Step S36：When chinese character is lost using by the anti-method for pushing away Chinese character of spcial character；Make word Character block of the Fu Kuai centers in the interval of car plate 1/7~2/7 is spcial character, and the left side of spcial character is Chinese character；
Step S37：If projection character block number is more than 7, illustrate after License Plate Segmentation that some characters may be divided into many Individual character block, thus further to projection character block merge treatment.
In the present embodiment, in the step S4, characters on license plate is identified as follows：
Step S41：The convolutional neural networks that input training image is extracted to implicit features；
Step S42：It is input into training image to explicit features sorter network；
Step S43：The convolutional neural networks that training implicit features are extracted, the training process of convolutional neural networks be one by Initial " low layer " feature is gradually converted into the expression process of " high level " feature by the automatic study of parameter, convolutional neural networks Training mainly includes two stages：
First stage is the propagated forward stage：Propagated forward stage main process is one sample of taking-up from training sample This X (xp,yp) using training sample X as network input, by formula Ox=fn(...(f2(f1(XpW(1))W(2))...)W(n)) meter Calculate the reality output of training sample X；
Second stage is back-propagation phase：Back-propagation phase will calculate the reality output O of training sample XxWith reason Want to export YpDifference, the method according to minimization error is adjusted to the parameter of model parameter；
Step S44：One explicit features sorter network of training；Training explicit features sorter network includes three phases：
First stage is the extraction character feature stage：The stage is first rectified using Gamma orthosises to characters on license plate Just, regulation picture contrast, then the gradient of each pixel is calculated, profile information is obtained, character picture is then divided into n × n Junior unit, wherein n=6；Then the histogram of gradients of each junior unit is calculated, the Feature Descriptor of junior unit is obtained, then Again by junior unit according to 3 × 3 model split into m × n groups, every group of Feature Descriptor is together in series can be arrived the feature of group and retouch Son is stated, the Feature Descriptor of mn group is together in series and is obtained character feature Fexture1；By the n of character × n center matrix mark Non-character pixel, wherein n=8 are designated as, and are become low-resolution image, as character feature Fexture2, will Fexture1 and Fexture2 are together in series and obtain character feature；
Second stage is the training data preparatory stage：To every image zooming-out character feature of training data, and stamp All training images are organized into a matrix by class label；
Three phases are the training stage：Training matrix is input in SVMs, using RBF kernel functions to explicit Feature classifiers are trained；
Step S45：Image to be sorted is separately input in explicit features sorter network and implicit features sorter network, Obtain classification results；
Step S46：For each width test image I, the confidence level vector E=that implicit features extract network will be all obtained {e1,e2,...,eNAnd implicit features network confidence level vector E '={ e '1,e′2,...,e′N, wherein N is characters on license plate Class number；
Step S47：Obtain in step S45 corresponding | | the E | | of two vectors∞With | | E ' | |∞, the corresponding classifications of image IFor：Wherein i is | | E | |∞Corresponding classification, j is | | E ' | |∞Corresponding classification.
The foregoing is only presently preferred embodiments of the present invention, all impartial changes done according to scope of the present invention patent with Modification, should all belong to covering scope of the invention.

Claims (3)

1. a kind of car plate segmentation of the characters and their identification method under monitoring scene, it is characterised in that：Comprise the following steps：
Step S1：Characters on license plate image is classified, is that the training of explicit features grader and implicit features grader is carried out Prepare；
Step S2：Input license plate image；
Step S3：Line character point is entered with the character segmentation method that domain method is combined is connected using projection localization method to license plate image Cut；
Step S4：The method blended using explicit features grader and implicit features grader is identified to characters on license plate.
2. car plate segmentation of the characters and their identification method under a kind of monitoring scene according to claim 1, it is characterised in that：Institute State in step S3, characters on license plate is split as follows：
Step S31：Projection localization method first uses formula with the character segmentation method that domain method is combined is connected：F (i, j)=α R (i, j)+ β G (i, j)+χ B (i, j)), wherein α=0.30 is made, β=0.59, license plate image is carried out gray processing treatment by χ=0.11；
Step S32：One width license plate image is carried out into mesh segmentation, the histogram of image is calculated each grid, and normalized, counted Cumulative mean value mu, and global gray average are calculated, then calculates the probability q for being assigned to class AA, and the probability q for being assigned to class BB；Using Formula：Sigma=qA*qB*(muA-muB)*(muA-muB) inter-class variance is calculated, inter-class variance maximum is found in circulation, and writes down Threshold value now, as optimal threshold, finally carry out thresholding to grid using this optimal threshold, repeat this process until Whole license plate image all binaryzations are finished；
Step S33：Extract character outline；
Step S34：Do boundary rectangle operation；
Step S35：If meeting the number of the boundary rectangle of size less than 7, illustrate to there may be Characters Stuck or car plate The situation that middle chinese character is lost；Part when Characters Stuck to adhesion carries out projection localization, and the position of segmentation is It is closer to the trough at image midpoint；
Step S36：When chinese character is lost using by the anti-method for pushing away Chinese character of spcial character；Make character block Character block of the center in the interval of car plate 1/7~2/7 is spcial character, and the left side of spcial character is Chinese character；
Step S37：If projection character block number is more than 7, some characters may be divided into multiple words after illustrating License Plate Segmentation Symbol block, therefore further to projection character block merge treatment.
3. car plate segmentation of the characters and their identification method under a kind of monitoring scene according to claim 1, it is characterised in that：Institute State in step S4, characters on license plate is identified as follows：
Step S41：The convolutional neural networks that input training image is extracted to implicit features；
Step S42：It is input into training image to explicit features sorter network；
Step S43：The convolutional neural networks that training implicit features are extracted, the training of convolutional neural networks mainly includes two ranks Section：
First stage is the propagated forward stage：Propagated forward stage main process is one sample X of taking-up from training sample (xp,yp) using training sample X as network input, by formula Ox=fn(...(f2(f1(XpW(1))W(2))...)W(n)) calculate The reality output of training sample X；
Second stage is back-propagation phase：Back-propagation phase will calculate the reality output O of training sample XxExported with ideal YpDifference, the method according to minimization error is adjusted to the parameter of model parameter；
Step S44：One explicit features sorter network of training；Training explicit features sorter network includes three phases：
First stage is the extraction character feature stage：The stage is first corrected using Gamma orthosises to characters on license plate, is adjusted Section picture contrast, then the gradient of each pixel is calculated, profile information is obtained, then character picture is divided into the small list of n × n Unit, wherein n=6；Then the histogram of gradients of each junior unit is calculated, the Feature Descriptor of junior unit is obtained, then again will be small According to 3 × 3 model split into m × n groups, every group of Feature Descriptor is together in series can arrive the Feature Descriptor of group to unit, The Feature Descriptor of mn group is together in series and obtains character feature Fexture1；By the n of character × n center matrix labeled as non- Character pixels, wherein n=8, and become low-resolution image, as character feature Fexture2, by Fexture1 and Fexture2 is together in series and obtains character feature；
Second stage is the training data preparatory stage：To every image zooming-out character feature of training data, and stamp classification All training images are organized into a matrix by label；
Three phases are the training stage：Training matrix is input in SVMs, using RBF kernel functions to explicit features Grader is trained；
Step S45：Image to be sorted is separately input in explicit features sorter network and implicit features sorter network, is obtained Classification results；
Step S46：For each width test image I, the confidence level vector E={ e that implicit features extract network will be all obtained1, e2,...,eNAnd implicit features network confidence level vector E '={ e '1,e′2,...,e′N, wherein N is the classification of characters on license plate Number；
Step S47：Obtain in step S45 corresponding | | the E | | of two vectors∞With | | E ' | |∞, the corresponding classifications of image IFor：Wherein i is | | E | |∞Corresponding classification, j is | | E ' | |∞Corresponding classification.
CN201710278593.1A 2017-04-25 2017-04-25 License plate character segmentation and recognition method in monitoring scene Active CN106886778B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710278593.1A CN106886778B (en) 2017-04-25 2017-04-25 License plate character segmentation and recognition method in monitoring scene

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710278593.1A CN106886778B (en) 2017-04-25 2017-04-25 License plate character segmentation and recognition method in monitoring scene

Publications (2)

Publication Number Publication Date
CN106886778A true CN106886778A (en) 2017-06-23
CN106886778B CN106886778B (en) 2020-02-07

Family

ID=59183579

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710278593.1A Active CN106886778B (en) 2017-04-25 2017-04-25 License plate character segmentation and recognition method in monitoring scene

Country Status (1)

Country Link
CN (1) CN106886778B (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107688784A (en) * 2017-08-23 2018-02-13 福建六壬网安股份有限公司 A kind of character identifying method and storage medium based on further feature and shallow-layer Fusion Features
CN107958219A (en) * 2017-12-06 2018-04-24 电子科技大学 Image scene classification method based on multi-model and Analysis On Multi-scale Features
CN107992897A (en) * 2017-12-14 2018-05-04 重庆邮电大学 Commodity image sorting technique based on convolution Laplce's sparse coding
CN108509912A (en) * 2018-04-03 2018-09-07 深圳市智绘科技有限公司 Multipath network video stream licence plate recognition method and system
CN109086767A (en) * 2018-07-11 2018-12-25 于洋 The dividing method of fuzzy license plate in a kind of monitoring scene
WO2020088338A1 (en) * 2018-10-30 2020-05-07 杭州海康威视数字技术股份有限公司 Method and apparatus for building recognition model
CN113392838A (en) * 2021-08-16 2021-09-14 智道网联科技(北京)有限公司 Character segmentation method and device and character recognition method and device
CN115171092A (en) * 2022-09-08 2022-10-11 松立控股集团股份有限公司 End-to-end license plate detection method based on semantic enhancement

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101398894A (en) * 2008-06-17 2009-04-01 浙江师范大学 Automobile license plate automatic recognition method and implementing device thereof
CN104298976A (en) * 2014-10-16 2015-01-21 电子科技大学 License plate detection method based on convolutional neural network
CN105354574A (en) * 2015-12-04 2016-02-24 山东博昂信息科技有限公司 Vehicle number recognition method and device
CN105354572A (en) * 2015-12-10 2016-02-24 苏州大学 Automatic identification system of number plate on the basis of simplified convolutional neural network
CN106096602A (en) * 2016-06-21 2016-11-09 苏州大学 Chinese license plate recognition method based on convolutional neural network

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101398894A (en) * 2008-06-17 2009-04-01 浙江师范大学 Automobile license plate automatic recognition method and implementing device thereof
CN104298976A (en) * 2014-10-16 2015-01-21 电子科技大学 License plate detection method based on convolutional neural network
CN105354574A (en) * 2015-12-04 2016-02-24 山东博昂信息科技有限公司 Vehicle number recognition method and device
CN105354572A (en) * 2015-12-10 2016-02-24 苏州大学 Automatic identification system of number plate on the basis of simplified convolutional neural network
CN106096602A (en) * 2016-06-21 2016-11-09 苏州大学 Chinese license plate recognition method based on convolutional neural network

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107688784A (en) * 2017-08-23 2018-02-13 福建六壬网安股份有限公司 A kind of character identifying method and storage medium based on further feature and shallow-layer Fusion Features
CN107958219A (en) * 2017-12-06 2018-04-24 电子科技大学 Image scene classification method based on multi-model and Analysis On Multi-scale Features
CN107992897A (en) * 2017-12-14 2018-05-04 重庆邮电大学 Commodity image sorting technique based on convolution Laplce's sparse coding
CN108509912A (en) * 2018-04-03 2018-09-07 深圳市智绘科技有限公司 Multipath network video stream licence plate recognition method and system
CN108509912B (en) * 2018-04-03 2021-09-28 深圳市智绘科技有限公司 License plate recognition method and system for multi-channel network video stream
CN109086767A (en) * 2018-07-11 2018-12-25 于洋 The dividing method of fuzzy license plate in a kind of monitoring scene
CN109086767B (en) * 2018-07-11 2019-05-07 于洋 The dividing method of fuzzy license plate in a kind of monitoring scene
WO2020088338A1 (en) * 2018-10-30 2020-05-07 杭州海康威视数字技术股份有限公司 Method and apparatus for building recognition model
CN113392838A (en) * 2021-08-16 2021-09-14 智道网联科技(北京)有限公司 Character segmentation method and device and character recognition method and device
CN113392838B (en) * 2021-08-16 2021-11-19 智道网联科技(北京)有限公司 Character segmentation method and device and character recognition method and device
CN115171092A (en) * 2022-09-08 2022-10-11 松立控股集团股份有限公司 End-to-end license plate detection method based on semantic enhancement
CN115171092B (en) * 2022-09-08 2022-11-18 松立控股集团股份有限公司 End-to-end license plate detection method based on semantic enhancement

Also Published As

Publication number Publication date
CN106886778B (en) 2020-02-07

Similar Documents

Publication Publication Date Title
CN106886778A (en) A kind of car plate segmentation of the characters and their identification method under monitoring scene
CN110956094B (en) RGB-D multi-mode fusion personnel detection method based on asymmetric double-flow network
CN104778453B (en) A kind of night pedestrian detection method based on infrared pedestrian's brightness statistics feature
Yuan et al. Robust traffic sign recognition based on color global and local oriented edge magnitude patterns
CN107491762B (en) A kind of pedestrian detection method
CN107316035A (en) Object identifying method and device based on deep learning neutral net
CN107463892A (en) Pedestrian detection method in a kind of image of combination contextual information and multi-stage characteristics
CN107122776A (en) A kind of road traffic sign detection and recognition methods based on convolutional neural networks
CN111965636A (en) Night target detection method based on millimeter wave radar and vision fusion
CN108564673A (en) A kind of check class attendance method and system based on Global Face identification
CN110555465A (en) Weather image identification method based on CNN and multi-feature fusion
CN108764096B (en) Pedestrian re-identification system and method
CN111062381B (en) License plate position detection method based on deep learning
CN111242911A (en) Method and system for determining image definition based on deep learning algorithm
CN115601717B (en) Deep learning-based traffic offence behavior classification detection method and SoC chip
CN104217206A (en) Real-time attendance counting method based on high-definition videos
CN115115908A (en) Cross-domain target detection model training method, target detection method and storage medium
CN114187664B (en) Rope skipping counting system based on artificial intelligence
CN115019340A (en) Night pedestrian detection algorithm based on deep learning
CN112347967B (en) Pedestrian detection method fusing motion information in complex scene
CN117994573A (en) Infrared dim target detection method based on superpixel and deformable convolution
Jin et al. Fusing Canny operator with vibe algorithm for target detection
CN108520208A (en) Localize face recognition method
CN114694133B (en) Text recognition method based on combination of image processing and deep learning
CN109145744A (en) A kind of LSTM network pedestrian recognition methods again based on adaptive prediction mode

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