CN108108734A - A kind of licence plate recognition method and device - Google Patents

A kind of licence plate recognition method and device Download PDF

Info

Publication number
CN108108734A
CN108108734A CN201611052182.2A CN201611052182A CN108108734A CN 108108734 A CN108108734 A CN 108108734A CN 201611052182 A CN201611052182 A CN 201611052182A CN 108108734 A CN108108734 A CN 108108734A
Authority
CN
China
Prior art keywords
character
point
license plate
image region
plate 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
Application number
CN201611052182.2A
Other languages
Chinese (zh)
Other versions
CN108108734B (en
Inventor
韦立庆
何海峰
钮毅
罗兵华
朱江
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.)
Hangzhou Hikvision Digital Technology Co Ltd
Original Assignee
Hangzhou Hikvision Digital Technology Co Ltd
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 Hangzhou Hikvision Digital Technology Co Ltd filed Critical Hangzhou Hikvision Digital Technology Co Ltd
Priority to CN201611052182.2A priority Critical patent/CN108108734B/en
Publication of CN108108734A publication Critical patent/CN108108734A/en
Application granted granted Critical
Publication of CN108108734B publication Critical patent/CN108108734B/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
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • 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
    • 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
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Data Mining & Analysis (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Character Input (AREA)
  • Traffic Control Systems (AREA)

Abstract

The embodiment of the present application provides a kind of licence plate recognition method and device, is related to technical field of intelligent traffic.The described method includes：Obtain the license plate image region of the number-plate number to be identified；Determine the Character segmentation point in the license plate image region；According to identified Character segmentation point, the corresponding alternative character zone partitioning scheme in the license plate image region is obtained；For each alternative character zone partitioning scheme, character recognition is carried out to the license plate image region, obtains character identification result；According to the character identification result obtained, the corresponding number-plate number in the license plate image region is obtained.Car license recognition is carried out using the technical solution in the embodiment of the present application, the compatibility of Car license recognition can be improved.

Description

A kind of licence plate recognition method and device
Technical field
This application involves technical field of intelligent traffic, more particularly to a kind of licence plate recognition method and device.
Background technology
Car plate is vehicle " identity card ", is an important information for being different from other motor vehicles.License plate recognition technology It has been widely used in the scenes such as bayonet, parking lot and electronic police, to obtain the number plate information of vehicle in scene, in public security Numerous aspects such as management play the power of " intelligent transportation algorithm ".
In the prior art, in the number-plate number in identifying license plate image to be identified, generally directed to license plate image to be identified, It is matched to the multiple plate templates pre-saved one by one, and then identifies the number-plate number.Detailed process is：Position vehicle to be identified License plate area in board image according to selected plate template, carries out license plate image to be identified Character segmentation, and to segmentation after Each character zone carry out character recognition.If character recognition success, then it is assumed that above-mentioned plate template successful match, by word Symbol recognition result is determined as the number-plate number of license plate image to be identified.If character recognition is unsuccessful, another car plate is selected Template repeats the above process.
For different types of car plate, it includes character feature it is different.In order to can recognize that the vehicle of these types Board, it usually needs build plate template for each type of car plate.
Under normal conditions, when carrying out Car license recognition using the above method, it can recognize that and meet in the car plate of the above-mentioned type The number-plate number.But if the type of the car plate in license plate image to be identified not in the range of the above-mentioned type, can not basis Above-mentioned plate template correctly splits license plate area, also just can not carry out Car license recognition to it using the above method, therefore The compatibility of licence plate recognition method is not high in the prior art.
The content of the invention
The embodiment of the present application has been designed to provide a kind of licence plate recognition method and device, to improve the simultaneous of Car license recognition Capacitive.Specific technical solution is as follows.
In order to achieve the above object, this application discloses a kind of licence plate recognition method, the described method includes：
Obtain the license plate image region of the number-plate number to be identified；
Determine the Character segmentation point in the license plate image region；
According to identified Character segmentation point, the corresponding alternative character zone segmentation side in the license plate image region is obtained Formula；
For each alternative character zone partitioning scheme, character recognition is carried out to the license plate image region, obtains word Accord with recognition result；
According to the character identification result obtained, the corresponding number-plate number in the license plate image region is obtained.
Optionally, it is described determine the license plate image region in Character segmentation point the step of, including：
According to vertical projection method, the upright projection value of each pixel column in the license plate image region is determined；
According to identified upright projection value, Character segmentation point is determined.
Optionally, upright projection value determined by the basis, the step of determining Character segmentation point, including：
Determine the width of sliding window；
According to the width and identified each upright projection value, corresponding first threshold of each upright projection value is calculated Value；
According to each upright projection value and corresponding first threshold, Character segmentation point is determined；
Wherein, it is described according to the width and identified each upright projection value, it calculates each upright projection value and corresponds to First threshold the step of, including：
The corresponding first threshold of each upright projection value is calculated in the following way：
According to the width, according to putting in order for upright projection value, selection includes the company including target vertical projection value Continuous first quantity upright projection value, wherein, the upright projection value puts in order and pixel in the license plate image region Putting in order for row is consistent, and the target vertical projection value is：One in identified upright projection value；
The average value of selected upright projection value is calculated, and the average value is determined as the target vertical projection value Corresponding first threshold.
Optionally, the step of width of the definite sliding window, including：
The height in the license plate image region is obtained, according to the height, determines the width of sliding window.
Optionally, the type of Character segmentation point includes left segmentation vertex type and right segmentation vertex type；The upright projection value It is to be obtained according to the binaryzation license plate image region of wrongly written or mispronounced character black matrix；
It is described according to each upright projection value and corresponding first threshold, the step of determining Character segmentation point, including：
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as left point The Character segmentation point of cutpoint type：
Proj (i) ＜ proj_th (i) and proj (i+1) >=proj_th (i+1)
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as right point The Character segmentation point of cutpoint type：
Proj (i) >=proj_th (i) and proj (i+1) ＜ proj_th (i+1)
Wherein, the proj (i) is i-th of upright projection value, and the proj_th (i) is and i-th of upright projection value pair The first threshold answered.
Optionally, described according to each upright projection value and corresponding first threshold, the step of determining Character segmentation point it Afterwards, the method further includes：
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as described The segmentation candidates point in license plate image region：
Proj (i)-proj_th (i) ＜ Th
Wherein, the Th is default second threshold；
In the following way, the type of each segmentation candidates point is determined：
The distance between target candidate cut-point and each Character segmentation point value are calculated, by the corresponding character point of lowest distance value The type of cutpoint is determined as the type of the target candidate cut-point, wherein, the target candidate cut-point is identified time Select one in cut-point；
Character segmentation point determined by the basis obtains the corresponding alternative character zone point in the license plate image region The step of cutting mode, including：
According to identified Character segmentation point and segmentation candidates point, it is corresponding alternative to obtain the license plate image region Character zone partitioning scheme.
Optionally, described according to each upright projection value and corresponding first threshold, the step of determining Character segmentation point it Afterwards, the method further includes：
According to most stable extremal region algorithm, the stability region in the license plate image region is obtained；
It is non-character cut-point to set the Character segmentation point inside each stability region；
Character segmentation point determined by the basis obtains the corresponding alternative character zone point in the license plate image region The step of cutting mode, including：
According to remaining Character segmentation point, the corresponding alternative character zone segmentation side in the license plate image region is obtained Formula.
Optionally, Character segmentation point determined by the basis obtains the corresponding alternative word in the license plate image region The step of according with region segmentation mode, including：
According to identified Character segmentation point, character zone to be selected is obtained；
The character zone to be selected that width is in the range of [w1, w2] is determined as target character region, wherein, the w1 is Default first width threshold value, the w2 are default second width threshold value, and the w2 is not less than the w1；
According to identified target character region, the corresponding alternative character zone segmentation in the license plate image region is obtained Mode.
In order to achieve the above object, this application discloses a kind of license plate recognition device, described device includes：
Image-region obtains module, for obtaining the license plate image region of the number-plate number to be identified；
Cut-point determining module, for determining the Character segmentation point in the license plate image region；
Partitioning scheme obtains module, for according to identified Character segmentation point, obtaining the license plate image region and corresponding to Alternative character zone partitioning scheme；
Character recognition module, for being directed to each alternative character zone partitioning scheme, to the license plate image region into Line character identifies, obtains character identification result；
The number-plate number obtains module, for according to the character identification result obtained, obtaining the license plate image region pair The number-plate number answered.
Optionally, the cut-point determining module, including：
Projection value determination sub-module, for according to vertical projection method, determining each pixel column in the license plate image region Upright projection value；
Cut-point determination sub-module, for according to identified upright projection value, determining Character segmentation point.
Optionally, the cut-point determination sub-module, including：
Width determination unit, for determining the width of sliding window；
Threshold computation unit, for according to the width and identified each upright projection value, calculating each vertical throwing The corresponding first threshold of shadow value；
Cut-point determination unit, for according to each upright projection value and corresponding first threshold, determining Character segmentation point；
Wherein, the threshold computation unit, is specifically used for：
The corresponding first threshold of each upright projection value is calculated in the following way：
According to the width, according to putting in order for upright projection value, selection includes the company including target vertical projection value Continuous first quantity upright projection value, wherein, the upright projection value puts in order and pixel in the license plate image region Putting in order for row is consistent, and the target vertical projection value is：One in identified upright projection value；
The average value of selected upright projection value is calculated, and the average value is determined as the target vertical projection value Corresponding first threshold.
Optionally, the width determination unit, is specifically used for：
The height in the license plate image region is obtained, according to the height, determines the width of sliding window.
Optionally, the type of Character segmentation point includes left segmentation vertex type and right segmentation vertex type；The upright projection value It is to be obtained according to the binaryzation license plate image region of wrongly written or mispronounced character black matrix；
The cut-point determination unit, is specifically used for：
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as left point The Character segmentation point of cutpoint type：
Proj (i) ＜ proj_th (i) and proj (i+1) >=proj_th (i+1)
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as right point The Character segmentation point of cutpoint type：
Proj (i) >=proj_th (i) and proj (i+1) ＜ proj_th (i+1)
Wherein, the proj (i) is i-th of upright projection value, and the proj_th (i) is and i-th of upright projection value pair The first threshold answered.
Optionally, after the cut-point determination unit, described device further includes candidate point determination unit；The candidate Point determination unit, is used for：
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as described The segmentation candidates point in license plate image region：
Proj (i)-proj_th (i) ＜ Th
Wherein, the Th is default second threshold；
In the following way, the type of each segmentation candidates point is determined：
The distance between target candidate cut-point and each Character segmentation point value are calculated, by the corresponding character point of lowest distance value The type of cutpoint is determined as the type of the target candidate cut-point, wherein, the target candidate cut-point is identified time Select one in cut-point；
The partitioning scheme obtains module, is specifically used for：
According to identified Character segmentation point and segmentation candidates point, it is corresponding alternative to obtain the license plate image region Character zone partitioning scheme.
Optionally, after the cut-point determination unit, described device further includes：
Stability region obtaining unit, for according to most stable extremal region algorithm, obtaining the steady of the license plate image region Determine region；
Cut-point setting unit is non-character cut-point for setting the Character segmentation point inside each stability region；
The partitioning scheme obtains module, is specifically used for：
According to remaining Character segmentation point, the corresponding alternative character zone segmentation side in the license plate image region is obtained Formula.
Optionally, the partitioning scheme obtains module, including：
Treat that favored area obtains submodule, for according to identified Character segmentation point, obtaining character zone to be selected；
Target area determination sub-module, the character zone to be selected for width to be in the range of [w1, w2] are determined as mesh Character zone is marked, wherein, the w1 is default first width threshold value, and the w2 is default second width threshold value, and the w2 is not small In the w1；
Partitioning scheme obtains subelement, for according to identified target character region, obtaining the license plate image region Corresponding alternative character zone partitioning scheme.
As seen from the above technical solution, the embodiment of the present application is for the license plate image area of the number-plate number to be identified obtained Domain first determines the Character segmentation point in the license plate image region, then according to identified Character segmentation point, obtains car plate figure As the corresponding alternative character zone partitioning scheme in region, and for each alternative character zone partitioning scheme, to car plate figure As region carries out character recognition, acquisition character identification result, according to the character identification result obtained, acquisition license plate image region The corresponding number-plate number.
That is, the embodiment of the present application is directed to the license plate image region of the number-plate number to be identified, it is corresponding every to obtain its A alternative character zone partitioning scheme carries out character recognition, without basis according to each partitioning scheme to license plate image region Default plate template carries out character recognition to license plate image region, that is, does not need to distinguish the corresponding car plate class in license plate image region Type.Therefore, Car license recognition is carried out using the technical solution of the embodiment of the present application, the compatibility of Car license recognition can be improved.
Description of the drawings
It in order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, below will be to embodiment or existing There is attached drawing needed in technology description to be briefly described.It should be evident that the accompanying drawings in the following description is only this Some embodiments of application, for those of ordinary skill in the art, without creative efforts, can be with Other attached drawings are obtained according to these attached drawings.
Fig. 1 is a kind of flow diagram of licence plate recognition method provided by the embodiments of the present application；
Fig. 2 a are a kind of exemplary plot of license plate image region and corresponding Character segmentation point；
Fig. 2 b are that a kind of exemplary plot after Character segmentation result is obtained according to alternative character zone partitioning scheme；
Fig. 3 is part car plate exemplary plot；
Fig. 4 is another flow diagram of licence plate recognition method provided by the embodiments of the present application；
Fig. 5 is a kind of flow diagram of step S102B in Fig. 4；
Fig. 6 a are a kind of exemplary plot of the corresponding upright projection curve in license plate image region and dynamic threshold curve；
Fig. 6 b are a kind of exemplary plot obtained for license plate image region after Character segmentation point；
Fig. 7 is a kind of structure diagram of license plate recognition device provided by the embodiments of the present application；
Fig. 8 is another structure diagram of license plate recognition device provided by the embodiments of the present application.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, the technical solution in the embodiment of the present application is carried out clear, complete Whole description.Obviously, described embodiment is only the part of the embodiment of the application, instead of all the embodiments.Base Embodiment in the application, those of ordinary skill in the art are obtained all on the premise of creative work is not made Other embodiment shall fall in the protection scope of this application.
The embodiment of the present application provides a kind of licence plate recognition method and device, and applied to electronic equipment, which can To be terminal device or server etc., wherein, terminal device can include computer, tablet computer, smart mobile phone, driving recording The equipment such as instrument.Car license recognition is carried out using the technical solution in the embodiment of the present application, the compatibility of Car license recognition can be improved.
Below by specific embodiment, the application is described in detail.
Fig. 1 is a kind of flow diagram of licence plate recognition method provided by the embodiments of the present application, applied to electronic equipment.It should Method includes the following steps：
Step S101：Obtain the license plate image region of the number-plate number to be identified.
Wherein, the license plate image region of the number-plate number to be identified can be understood as：Need the car plate figure of progress Car license recognition As region.License plate image region is the image-region that car plate is included in license plate image.License plate image refers to the car plate for including vehicle Partial image.As a kind of preferable embodiment, license plate image region can be that the outermost side frame of characters on license plate is formed Image-region.Certainly, license plate image region can also be the region for including other image sections outside characters on license plate.It is logical Often, license plate image region can be arranged to rectangular area.
For example, the image-region that dotted line frame is surrounded in Fig. 2 a is exactly a license plate image region, it can from the figure Go out in dotted line frame comprising car plate " A4170 ".
Specifically, the license plate image region of the number-plate number to be identified, directly obtains or uses following What mode obtained：The license plate image of the number-plate number to be identified is obtained, License Plate is carried out to the license plate image, obtains license plate image Region.Wherein, the license plate image of the number-plate number to be identified can be understood as：Need the license plate image of progress Car license recognition.
The license plate image of the above-mentioned number-plate number to be identified can be the image comprising vehicle captured on road or Image comprising vehicle of shooting etc. in parking lot.Certainly, the license plate image of the above-mentioned number-plate number to be identified, which can also be, passes through What other modes obtained, the application is not defined the acquisition pattern of the license plate image of the number-plate number to be identified.
Electronic equipment internal as executive agent can include image capture device, can not also be set comprising Image Acquisition It is standby.
When the electronic equipment internal as executive agent includes image capture device, electronic equipment is obtaining vehicle to be identified During the license plate image of trade mark code, it can include：Receive the license plate image of the number-plate number to be identified of image capture device acquisition.
When the electronic equipment internal as executive agent does not include image capture device, which can be with outside Image capture device be connected, electronic equipment can include when obtaining the license plate image of the number-plate number to be identified：Obtain image The license plate image of the number-plate number to be identified of collecting device acquisition.
The license plate image of the number-plate number to be identified obtained can be that image capture device gathers in real time, may not be It gathers in real time, but image capture device is collected what is stored afterwards in advance.
Step S102：Determine the Character segmentation point in the license plate image region.
Wherein, the mode of the coordinate of definite Character segmentation point may be employed to determine Character segmentation point in the present embodiment.Specifically , Character segmentation point can be understood as the pixel in longitudinal pixel column in license plate image region, it is understood that for longitudinal picture The pixel of bottom in element row.
Specifically, the methods of being directed to license plate image region, upright projection, connected domain, stroke width may be employed determines vehicle Character segmentation point in board image-region.The present embodiment is to determining that the specific method of Character segmentation point does not limit.
Step S103：According to identified Character segmentation point, the corresponding alternative character in the license plate image region is obtained Region segmentation mode.
Wherein, " alternative " also refer to " possible ".Character zone partitioning scheme is for entire license plate image region Combination between Character segmentation point.It is understood that two different Character segmentation points can determine a character area Domain, the two Character segmentation points form a Character segmentation point group.Each character zone partitioning scheme include it is at least one so Character segmentation point group.
In general, according to identified Character segmentation point, at least one character corresponding with license plate image region can be obtained Region segmentation mode.Character zone partitioning scheme is obtained, also just can word be carried out to license plate image region according to the partitioning scheme Symbol segmentation, obtains corresponding Character segmentation result.
In general, comprising a certain number of " spaces " in car plate, i.e., there are white space between character, in these white spaces There is no character.Therefore, can be the character of white space by the character zone formed in definite character zone partitioning scheme Cut-point group is arranged to non-character cut-point group, can so be eliminated as much as disturbing factor, improves the efficiency of character recognition.
Specifically, which Character segmentation point group institute can be determined according to the methods of upright projection, connected domain, stroke width Definite character zone is white space.
For example, 10 Character segmentation points in license plate image region, triangle are marked with triangle symbol in fig. 2 a The number of Character segmentation point has been marked among symbol, these number be respectively 1,2 ..., 10.According to this 10 Character segmentation points, Following two character zone partitioning schemes can be obtained, the first is, (1,2) (3,4) (5,6) (7,8) (9,10), second It is (1,2) (3,6) (7,8) (9,10).Wherein, the combination of (1,2) etc is exactly a Character segmentation point group.Also, according to vertical The methods of delivering directly shadow, connected domain, stroke width, it may be determined that right between (2,3) (4,5) (6,7) (8,9) in license plate image region The character zone answered is white space, therefore gives up these combinations.
As a kind of specific embodiment, in order to more accurately split car plate image-region, according to identified character Cut-point obtains the corresponding alternative character zone partitioning scheme in the license plate image region, can include step 1~step 3：
Step 1：According to identified Character segmentation point, character zone to be selected is obtained.
Wherein, character zone to be selected can be the region using the coordinate representation of two Character segmentation points.
Step 2：The character zone to be selected that width is in the range of [w1, w2] is determined as target character region, wherein, institute It is default first width threshold value to state w1, and the w2 is default second width threshold value, and the w2 is not less than the w1.In general, w1 and The value of w2 can be obtained rule of thumb.
It is understood that width is in character zone to be selected in the range of [w1, w2] it is more likely that real character area Domain.To be selected character zone of the width more than w2 or less than w1 may be wide or narrow, belongs to the possibility of true character zone It is smaller, therefore given up, to improve the accuracy of segmentation.
Step 3：According to identified target character region, the corresponding alternative character area in the license plate image region is obtained Regional partition mode.
Step S104：For each alternative character zone partitioning scheme, the license plate image region is known into line character Not, character identification result is obtained.
It, can be first according to standby when carrying out character recognition to the license plate image region as a kind of specific embodiment The character zone partitioning scheme of choosing, is split license plate image region, obtains and each alternative character zone partitioning scheme Then corresponding character zone segmentation result carries out character recognition to the character picture in each character zone segmentation result again, Obtain character identification result.
Specifically, when carrying out character recognition to the character picture in each character zone segmentation result, it can be by each word The characteristic value input Character recognizer of image is accorded with, Character recognizer exports the corresponding character of each character picture and confidence level, That is each character identification result can include character and confidence level.
The example in step S103 is continued to use, it, can be to license plate image area for two kinds of alternative character zone partitioning schemes Domain is split, and two kinds of character zone segmentation results shown in Fig. 2 b is obtained, by the character in each character zone segmentation result Image inputs character classifier, can obtain two kinds of character identification results shown in table 1.
Table 1
When carrying out character recognition to character picture, Hog (histograms of oriented gradients)+SVM (supporting vectors may be employed Machine) mode of grader carries out character recognition.Specifically, the Hog characteristic values of each character picture can be extracted, by this feature value Input SVM classifier, grader output recognition result.
Wherein, SVM classifier includes 37 output units, be respectively 10 numbers, 26 English alphabets and one " not Know " character output.The codomain of confidence level is [0,1000].
It should be noted that the embodiment of the present application can also use other Character recognizers to carry out character recognition, this Shen Please embodiment character recognition mode is not specifically limited.
Step S105：According to the character identification result obtained, the corresponding number-plate number in the license plate image region is obtained.
As a kind of specific embodiment, according to the character identification result obtained, the license plate image region is obtained During the corresponding number-plate number, the confidence level and value of each character identification result can be first calculated, by confidence level and value maximum pair The character identification result answered is determined as the corresponding number-plate number in license plate image region.
The example in step S104 is continued to use, for two kinds of character identification results in table 1, each character recognition can be obtained As a result confidence level and value are respectively：The confidence level and value of the first character identification result are 904+935+956+923+960 =4678, the confidence level and value of second character identification result are：904+300+923+960=3087.As it can be seen that the first character The confidence level and value of recognition result are larger, therefore the first character identification result is determined as the corresponding car plate in license plate image region Number.
It is pointed out that if it is determined that the recognition result of grader is " unknown " this character, i.e. expression fails from word Symbol identifies character in region.And the recognition result of grader why is determined as " unknown ", it is usually because in grader bag In each output unit contained, the confidence level of " unknown " output unit is higher.Thus have a problem that.Know obtaining character During other result, the confidence level of each character is higher in the character identification result, and corresponding confidence level and value are higher.However, When character identification result includes " unknown " this character, the confidence level and value of the character identification result should be relatively low, and The confidence level of " unknown " this character is higher, and the confidence level and value for but causing the character identification result also accordingly become higher, so Just do not square with the fact.
For example, to shown in Fig. 2 b second of segmentation result carry out character recognition in, in recognition result each character and Corresponding confidence level is respectively：A-904, " unknown "-700,7-923；0—960.As it can be seen that exist " not in the recognition result Know ".Correspondingly, the confidence level and value of the recognition result can recognize that the situation phase of non-" unknown " character with second character Than should be relatively low.But if directly the confidence level of the above results is taken and, obtained confidence level and value is very high.
During in order to ensure to ask confidence level and value, it can add up to the confidence level of each character in recognition result, work as knowledge When there is " unknown " character in other result, confidence level maximum is subtracted into its confidence level, obtains putting after " unknown " character change Reliability just solves the above problem using the amended confidence level.
For example, the confidence level of " unknown " character is 700, confidence level extreme value scope is [0,1000], then can be by confidence level Maximum 1000-700=300, the confidence level being determined as after " unknown " character change.
As shown in the above, the present embodiment is first determined for the license plate image region of the number-plate number to be identified obtained Character segmentation point in the license plate image region then according to identified Character segmentation point, obtains license plate image region pair The alternative character zone partitioning scheme answered, and for each alternative character zone partitioning scheme, to license plate image region into Line character identifies, obtains character identification result, according to the character identification result obtained, obtains the corresponding vehicle in license plate image region Trade mark code.
That is, the present embodiment is directed to the license plate image region of the number-plate number to be identified, it is corresponding each standby to obtain its The character zone partitioning scheme of choosing carries out character recognition, without according to default according to each partitioning scheme to license plate image region Plate template to license plate image region carry out character recognition, that is, do not need to distinguish the corresponding car plate type in license plate image region.Cause This, carries out Car license recognition using the technical solution of the present embodiment, can improve the compatibility of Car license recognition.
For example, in existing licence plate recognition method, can plate template be established according to the type of Chinese car plate, to know Other CHINESE REGION car plate.But for the car plate of these other countries of Fig. 3, then the plate template of CHINESE REGION can not be used to carry out Car license recognition.It can also be seen that not having fixed form between these car plates from Fig. 3, variation is more, using in the present embodiment Licence plate recognition method without matching plate template, does not need to distinguish what car plate type these car plates belong to, but according to car plate figure As Character segmentation point, acquisition possible character zone partitioning scheme corresponding with license plate image region, according to these are found in region Mode obtains the possible character identification result in license plate image region, and license plate number is determined from these possible character identification results Code.As it can be seen that the compatibility of the licence plate recognition method provided in the present embodiment is higher.
Fig. 4 is another flow diagram of licence plate recognition method provided by the embodiments of the present application, and this method is applied to electronics Equipment.This method embodiment is the improvement to embodiment of the method shown in Fig. 1.Non- improvements and embodiment illustrated in fig. 1 content phase Together, non-improvements are no longer discussed in detail in the present embodiment, and related description can refer to embodiment illustrated in fig. 1.
In embodiment illustrated in fig. 1, step S102 the step of determining the Character segmentation point in the license plate image region, can To include：
Step S102A：According to vertical projection method, the upright projection of each pixel column in the license plate image region is determined Value.
Wherein it is determined that in the license plate image region during upright projection value of each pixel column, it can be understood as：By described in The pixel value summation of each pixel column in license plate image region, using the upright projection value corresponding as the pixel column with value.Picture Element row refer to a row pixel longitudinal in image.
Specifically, in order to reduce the complexity of processing, each pixel column is vertical in the license plate image region is determined During projection value, can license plate image region be first converted into gray level image, binary conversion treatment then is carried out to gray level image again, is obtained Binary image is obtained, determines the upright projection value of each pixel column in binary image.
Wherein it is possible to binary conversion treatment is carried out to gray level image using big rule algorithm (OSTU).The binary image of acquisition Can be the image of surplus white background or the image of wrongly written or mispronounced character black matrix.
Step S102B：According to identified upright projection value, Character segmentation point is determined.
Specifically, step S102B can include numerous embodiments, can by identified upright projection value respectively in advance If threshold value is compared, the pixel that will be greater than in the corresponding license plate image region of upright projection value of predetermined threshold value is determined as word Accord with cut-point.In this embodiment, predetermined threshold value is a predetermined fixed value, and value can rule of thumb really It is fixed.This method is properly termed as fixed threshold method.
Can also be, by identified upright projection value compared with the threshold value of dynamic change.The threshold of the dynamic change Value can be determined according to license plate image region.That is, each upright projection value is compared with corresponding threshold value Compared with the corresponding threshold value of each upright projection value is different.The threshold value of the dynamic change can be determined according to upright projection value.This Kind method is properly termed as dynamic thresholding method.
As it can be seen that in the present embodiment, according to vertical projection method, and utilize character portion and background portion in license plate image region There are the Character segmentation point for the characteristics of notable difference, determining license plate image region between the pixel value of point pixel, institute can be improved The accuracy of definite Character segmentation point, so as to improve the accuracy of character recognition process.
Based on shown in Fig. 4 in a kind of specific embodiment of embodiment, in order to more accurately true according to upright projection value Determine Character segmentation point, step S102B, the step of determining Character segmentation point, can be according to figure according to identified upright projection value Flow diagram shown in 5 carries out, which specifically includes following steps：
Step S102B1：Determine the width of sliding window.
Wherein, in the present embodiment, when the width of license plate image region transverse direction with pixel quantity come when weighing, sliding window Mouthful width also represented with pixel quantity.When license plate image region it is longitudinally wide with normalized numerical value come when weighing, it is sliding The width of dynamic window is also represented with normalizing numerical value.In short, the width of sliding window, for representing quantitative value.The width can With the value for being preset value or being determined according to license plate image region.
In a kind of specific embodiment, in order to more accurately determine the first threshold in the description below, sliding window is determined The step of width of mouth, it can include：The height in the license plate image region is obtained, according to the height, determines sliding window Width.
It further, can be using the product of preset value and the height in license plate image region as the width of sliding window.Its In, preset value can be taken as 0.6 etc numerical value.For example, license plate image region is the image of wide 100 pixel * high, 20 pixels, it is sliding The width of dynamic window can using value as：The pixel of 0.6*20 pixels=12.
It can also be using the height and value in preset value and license plate image region as the width of sliding window.Certainly, true When determining the width of sliding window, other embodiments can also be included.
It should be noted that due to license plate image region longitudinally wide, there are certain associations between transverse height Property, therefore, the width of sliding window is determined according to the height in license plate image region, accuracy can be improved.
Step S102B2：According to the width and identified each upright projection value, each upright projection value pair is calculated The first threshold answered.
Wherein, for step S102B2, each vertical throwing can be calculated in the way of shown in following steps 1 and step 2 The corresponding first threshold of shadow value：
Step 1：According to the width, according to putting in order for upright projection value, selection exists comprising target vertical projection value Interior continuous first quantity upright projection value.
Wherein, putting in order for the upright projection value puts in order one with pixel column in the license plate image region It causes, the target vertical projection value is：One in identified upright projection value.
Step 2：The average value of selected upright projection value is calculated, and the average value is determined as the target vertical The corresponding first threshold of projection value.
Step S102B3：According to each upright projection value and corresponding first threshold, Character segmentation point is determined.
It should be noted that left segmentation vertex type and right minute can be usually divided into according to the type of Character segmentation point Cutpoint type, both types are respectively used to represent the cut-point of left and right two of character zone.
It is once mentioned in the above, license plate image region can be converted to binary image, be obtained according to binary image To the upright projection value of each pixel column, wherein, binary image can be surplus white background or wrongly written or mispronounced character black matrix. That is upright projection value can be obtained according to the binary image of surplus white background or according to wrongly written or mispronounced character black matrix Binary image.Below to obtain the situation of upright projection value, introduction step S102B3 according to the binary image of wrongly written or mispronounced character black matrix A kind of specific embodiment.
It is the binaryzation license plate image region according to wrongly written or mispronounced character black matrix in upright projection value as a kind of specific embodiment In the case of obtaining, step S102B3 according to each upright projection value and corresponding first threshold, determines the step of Character segmentation point Suddenly, 1 and step 2 be may comprise steps of：
Step 1：By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, determine For the Character segmentation point of left segmentation vertex type：
Proj (i) ＜ proj_th (i) and proj (i+1) >=proj_th (i+1)
Step 2：By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, determine For the Character segmentation point of right segmentation vertex type：
Proj (i) >=proj_th (i) and proj (i+1) ＜ proj_th (i+1)
Wherein, the proj (i) is i-th of upright projection value, and the proj_th (i) is and i-th of upright projection value pair The first threshold answered.
It is understood that for the Character segmentation point of left segmentation vertex type, corresponding upright projection value, which is less than, to be corresponded to First threshold, the corresponding upright projection value of character late cut-point be not less than corresponding first threshold.For right cut-point The Character segmentation point of type, corresponding upright projection value are not less than corresponding first threshold, and character late cut-point corresponds to Upright projection value be less than corresponding first threshold.
It, will be with using the coordinate of each horizontal pixel point in license plate image region as transverse axis as a specific embodiment The corresponding each upright projection value of each pixel column can obtain the corresponding upright projection curve in license plate image region as the longitudinal axis.It will First threshold corresponding with each upright projection value is used as and indulges as transverse axis by the coordinate of each horizontal pixel point in license plate image region Axis can obtain the corresponding dynamic threshold curve in license plate image region.Upright projection curve include wave crest and trough, rise and fall compared with Greatly, dynamic threshold curve is relatively smooth, rises and falls little.
Compare upright projection curve and the point in dynamic threshold curve, it is found that in the point of two curve intersections, be in The point of upright projection curve rising stage is the Character segmentation point (referred to as left cut-point) of left segmentation vertex type, is thrown in vertical The point that shadow curve declines the phase is the Character segmentation point (referred to as right cut-point) of right segmentation vertex type.
For example, in Fig. 6 a, the binaryzation license plate image overlying regions of wrongly written or mispronounced character black matrix depict corresponding upright projection curve 1 With dynamic threshold curve 2, it is seen that upright projection curve includes many wave crests and trough, and there are a small amount of wave crests for dynamic threshold curve And trough, it is smoother, it rises and falls little.Many left cut-points and right cut-point can be obtained by comparing this two curves.In Fig. 6 a In, the triangle symbol with " l " is represented into left cut-point, the triangle symbol with " r " is represented into right cut-point.
The above illustrates, in upright projection value is obtained according to the binaryzation license plate image region of wrongly written or mispronounced character black matrix In the case of, a kind of specific embodiment that step S102B3 is included.It is according to surplus in upright projection value based on same thinking In the case that the binaryzation license plate image region of white background obtains, the detailed description below that step S102B3 is included can be obtained, Specifically include following steps 1 and step 2：
Step 1：By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, determine For the Character segmentation point of left segmentation vertex type：
Proj (i) >=proj_th (i) and proj (i+1) ＜ proj_th (i+1)
Step 2：By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, determine For the Character segmentation point of right segmentation vertex type：
Proj (i) ＜ proj_th (i) and proj (i+1) >=proj_th (i+1)
Wherein, the proj (i) is i-th of upright projection value, and the proj_th (i) is and i-th of upright projection value pair The first threshold answered.
Present embodiment illustrates the embodiment that may refer to above-mentioned " wrongly written or mispronounced character black matrix " situation.
In conclusion in this embodiment, Character segmentation point, dynamic first threshold are determined using dynamic thresholding method It is determined according to upright projection value, can more accurately reflect the changing rule of upright projection value, so as to more accurately Determine Character segmentation point.
The quantity for the Character segmentation point determined by the above process may be very big, wherein containing many non-character segmentations Point.As shown in FIG. 6 a, a character may correspond to that there are two left cut-points and two right cut-points.In order to improve character point The accuracy cut, reduces the quantity of cut-point, and the present embodiment can also include implementation below.
As a kind of specific embodiment, after step S102B3, i.e., according to each upright projection value and corresponding One threshold value, after the step of determining Character segmentation point, the method can also include：According to most stable extremal region algorithm, obtain The stability region in the license plate image region is obtained, it is non-character cut-point to set the Character segmentation point inside each stability region. Character segmentation point inside each stability region does not include the Character segmentation point of each stability region both sides of the edge.
Wherein, most stable extremal region algorithm (Maximally Stable Extremal Regions, MSER) is a kind of Ask for the image segmentation algorithm of most stable extremal region.According to most stable extremal region algorithm, the license plate image area is obtained It, can be using the license plate image region as the input of most stable extremal region algorithm, output and car plate during the stability region in domain The corresponding each stability region of image-region.
That is, the Character segmentation point inside stability region is arranged to non-character cut-point, i.e., from definite The Character segmentation point inside stability region is deleted in Character segmentation point.
It is understood that stability region is considered as a complete character zone, it should not be partitioned from, and think Its internal Character segmentation point may be inaccurate.
Corresponding, according to identified Character segmentation point, it is corresponding alternative to obtain the license plate image region by step S103 Character zone partitioning scheme the step of, can include：
According to remaining Character segmentation point, the corresponding alternative character zone segmentation side in the license plate image region is obtained Formula.
To sum up, in the present embodiment, some inaccurate Character segmentation points are removed according to stability region, standard can be improved Exactness, while the quantity of cut-point can be also reduced, reduce the complexity of processing.
In general, according to the comparison of upright projection value and first threshold, it can be by left segmentation vertex type in license plate image region Character segmentation point and the Character segmentation point of right segmentation vertex type determine, but the Character segmentation point of omission may also be had Do not identify.
For example, the Character segmentation point determined for license plate image region is listed in Fig. 6 b, it can be seen that wherein " 41 " Character segmentation point among part fails effectively to determine.
Further, in order to identify the Character segmentation of omission point, improve license plate identification accuracy, step S102B3 it Afterwards, i.e., left segmentation vertex type and the right character for splitting vertex type are determined according to each upright projection value and corresponding first threshold After cut-point, the method can also include：
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as described The segmentation candidates point in license plate image region：
Proj (i)-proj_th (i) ＜ Th
Wherein, the Th is default second threshold, i.e. second threshold is pre-set fixed value, is become with above-mentioned dynamic The first threshold of change is different.
In the following way, the type of each segmentation candidates point is determined：
The distance between target candidate cut-point and each Character segmentation point value are calculated, by the corresponding character point of lowest distance value The type of cutpoint is determined as the type of the target candidate cut-point, wherein, the target candidate cut-point is identified time Select one in cut-point.
Corresponding, according to identified Character segmentation point, it is corresponding alternative to obtain the license plate image region by step S103 Character zone partitioning scheme the step of, including：
According to identified Character segmentation point and segmentation candidates point, it is corresponding alternative to obtain the license plate image region Character zone partitioning scheme.
In this embodiment, the segmentation candidates point in license plate image region is determined using fixed threshold method, can be incited somebody to action The Character segmentation point identifying processing omitted in license plate image region, so as to improve the accuracy of Character segmentation process.
Certainly, after segmentation candidates point, can also be set each according to the stability region in the license plate image region of acquisition Segmentation candidates point inside stability region is non-candidate cut-point, to improve the accuracy of Character segmentation, while reduces cut-point Quantity, to reduce processing complexity.
Fig. 7 is a kind of structure diagram of license plate recognition device provided by the embodiments of the present application, which is applied to electronics Equipment, the embodiment are corresponding with embodiment of the method shown in Fig. 1.Specifically, the device includes：
Image-region obtains module 701, for obtaining the license plate image region of the number-plate number to be identified；
Cut-point determining module 702, for determining the Character segmentation point in the license plate image region；
Partitioning scheme obtains module 703, for according to identified Character segmentation point, obtaining the license plate image region pair The alternative character zone partitioning scheme answered；
Character recognition module 704, for being directed to each alternative character zone partitioning scheme, to the license plate image region Character recognition is carried out, obtains character identification result；
The number-plate number obtains module 705, for according to the character identification result obtained, obtaining the license plate image region The corresponding number-plate number.
In a kind of embodiment based on embodiment illustrated in fig. 7, the partitioning scheme obtains module 703, can specifically wrap It includes：
Treat that favored area obtains submodule (not shown), for according to identified Character segmentation point, word selection to be treated in acquisition Accord with region；
Target area determination sub-module (not shown), for width to be in the character to be selected in the range of [w1, w2] Region is determined as target character region, wherein, the w1 is default first width threshold value, and the w2 is default second width threshold Value, the w2 are not less than the w1；
Partitioning scheme obtains subelement (not shown), for according to identified target character region, described in acquisition The corresponding alternative character zone partitioning scheme in license plate image region.
Fig. 8 is another structure diagram of license plate recognition device provided by the embodiments of the present application, which is Improvement project based on embodiment illustrated in fig. 7, non-improvements are identical with embodiment illustrated in fig. 7, and particular content is referred to Fig. 7 Illustrated embodiment.The embodiment is corresponding with embodiment of the method shown in Fig. 4.
Wherein, the cut-point determining module 702, specifically includes：
Projection value determination sub-module 702A, for according to vertical projection method, determining each picture in the license plate image region The upright projection value of element row；
Cut-point determination sub-module 702B, for according to identified upright projection value, determining Character segmentation point.
In a kind of embodiment based on embodiment illustrated in fig. 8, the cut-point determination sub-module 702B can specifically be wrapped It includes：
Width determination unit (not shown), for determining the width of sliding window；
Threshold computation unit (not shown), for according to the width and identified each upright projection value, meter Calculate the corresponding first threshold of each upright projection value；
Cut-point determination unit (not shown), for according to each upright projection value and corresponding first threshold, determining Character segmentation point；
Wherein, the threshold computation unit, is specifically used for：
The corresponding first threshold of each upright projection value is calculated in the following way：
According to the width, according to putting in order for upright projection value, selection includes the company including target vertical projection value Continuous first quantity upright projection value, wherein, the upright projection value puts in order and pixel in the license plate image region Putting in order for row is consistent, and the target vertical projection value is：One in identified upright projection value；
The average value of selected upright projection value is calculated, and the average value is determined as the target vertical projection value Corresponding first threshold.
In a kind of embodiment based on embodiment illustrated in fig. 8, the width determination unit is specifically used for：
The height in the license plate image region is obtained, according to the height, determines the width of sliding window.
In a kind of embodiment based on embodiment illustrated in fig. 8, the type of Character segmentation point include left segmentation vertex type and Right segmentation vertex type；The upright projection value is obtained according to the binaryzation license plate image region of wrongly written or mispronounced character black matrix；
The cut-point determination unit, is specifically used for：
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as left point The Character segmentation point of cutpoint type：
Proj (i) ＜ proj_th (i) and proj (i+1) >=proj_th (i+1)
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as right point The Character segmentation point of cutpoint type：
Proj (i) >=proj_th (i) and proj (i+1) ＜ proj_th (i+1)
Wherein, the proj (i) is i-th of upright projection value, and the proj_th (i) is and i-th of upright projection value pair The first threshold answered.
In a kind of embodiment based on embodiment illustrated in fig. 8, after the cut-point determination unit, described device is also It can include candidate point determination unit (not shown)；The candidate point determination unit, is used for：
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as described The segmentation candidates point in license plate image region：
Proj (i)-proj_th (i) ＜ Th
Wherein, the Th is default second threshold；
In the following way, the type of each segmentation candidates point is determined：
The distance between target candidate cut-point and each Character segmentation point value are calculated, by the corresponding character point of lowest distance value The type of cutpoint is determined as the type of the target candidate cut-point, wherein, the target candidate cut-point is identified time Select one in cut-point；
The partitioning scheme obtains module 703, specifically can be used for：
According to identified Character segmentation point and segmentation candidates point, it is corresponding alternative to obtain the license plate image region Character zone partitioning scheme.
In a kind of embodiment based on embodiment illustrated in fig. 8, after the cut-point determination unit, described device is also It can include：
Stability region obtaining unit (not shown), for according to most stable extremal region algorithm, obtaining the car plate The stability region of image-region；
Cut-point setting unit (not shown) is non-word for setting the Character segmentation point inside each stability region Accord with cut-point；
The partitioning scheme obtains module 703, specifically can be used for：According to remaining Character segmentation point, the vehicle is obtained The corresponding alternative character zone partitioning scheme of board image-region.
Since above device embodiment is obtained based on embodiment of the method, there is identical technique effect with this method, Therefore details are not described herein for the technique effect of device embodiment.For device embodiment, since it is substantially similar to method Embodiment, so describing fairly simple, the relevent part can refer to the partial explaination of embodiments of method.
It should be noted that herein, relational terms such as first and second and the like are used merely to a reality Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation In any this actual relation or order.Moreover, term " comprising ", "comprising" or any other variant be intended to it is non- It is exclusive to include, so that process, method, article or equipment including a series of elements not only include those elements, But also it including other elements that are not explicitly listed or further includes solid by this process, method, article or equipment Some elements.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including Also there are other identical elements in the process of the element, method, article or equipment.
It will appreciated by the skilled person that all or part of step in the above embodiment is can to pass through journey Sequence instructs relevant hardware, and come what is completed, the program can be stored in computer read/write memory medium.It is designated herein Storage medium refers to ROM/RAM, magnetic disc, CD etc..
The foregoing is merely the preferred embodiment of the application, the protection domain of the application is not intended to limit.It is all Any modification, equivalent substitution, improvement and etc. done within spirit herein and principle are all contained in the protection domain of the application It is interior.

Claims (16)

1. a kind of licence plate recognition method, which is characterized in that the described method includes：
Obtain the license plate image region of the number-plate number to be identified；
Determine the Character segmentation point in the license plate image region；
According to identified Character segmentation point, the corresponding alternative character zone partitioning scheme in the license plate image region is obtained；
For each alternative character zone partitioning scheme, character recognition is carried out to the license plate image region, character is obtained and knows Other result；
According to the character identification result obtained, the corresponding number-plate number in the license plate image region is obtained.
2. the according to the method described in claim 1, it is characterized in that, Character segmentation determined in the license plate image region The step of point, including：
According to vertical projection method, the upright projection value of each pixel column in the license plate image region is determined；
According to identified upright projection value, Character segmentation point is determined.
3. according to the method described in claim 2, it is characterized in that, upright projection value determined by the basis, determines character The step of cut-point, including：
Determine the width of sliding window；
According to the width and identified each upright projection value, the corresponding first threshold of each upright projection value is calculated；
According to each upright projection value and corresponding first threshold, Character segmentation point is determined；
Wherein, it is described according to the width and identified each upright projection value, calculate each upright projection value corresponding The step of one threshold value, including：
The corresponding first threshold of each upright projection value is calculated in the following way：
According to the width, according to putting in order for upright projection value, selection includes continuous the including target vertical projection value One quantity upright projection value, wherein, the upright projection value puts in order and pixel column in the license plate image region It puts in order consistent, the target vertical projection value is：One in identified upright projection value；
The average value of selected upright projection value is calculated, and the average value is determined as the target vertical projection value and is corresponded to First threshold.
4. according to the method described in claim 3, it is characterized in that, the step of the width of the definite sliding window, including：
The height in the license plate image region is obtained, according to the height, determines the width of sliding window.
5. according to the method described in claim 3, it is characterized in that, the type of Character segmentation point includes left segmentation vertex type and the right side Split vertex type；The upright projection value is obtained according to the binaryzation license plate image region of wrongly written or mispronounced character black matrix；
It is described according to each upright projection value and corresponding first threshold, the step of determining Character segmentation point, including：
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as left cut-point The Character segmentation point of type：
Proj (i) ＜ proj_th (i) and proj (i+1) >=proj_th (i+1)
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as right cut-point The Character segmentation point of type：
Proj (i) >=proj_th (i) and proj (i+1) ＜ proj_th (i+1)
Wherein, the proj (i) is i-th of upright projection value, and the proj_th (i) is corresponding with i-th of upright projection value First threshold.
6. according to the method described in claim 5, it is characterized in that, described according to each upright projection value and corresponding first threshold After the step of being worth, determining Character segmentation point, the method further includes：
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as the car plate The segmentation candidates point of image-region：
Proj (i)-proj_th (i) ＜ Th
Wherein, the Th is default second threshold；
In the following way, the type of each segmentation candidates point is determined：
The distance between target candidate cut-point and each Character segmentation point value are calculated, by the corresponding Character segmentation point of lowest distance value Type be determined as the type of the target candidate cut-point, wherein, the target candidate cut-point is identified candidate point One in cutpoint；
Character segmentation point determined by the basis obtains the corresponding alternative character zone segmentation side in the license plate image region The step of formula, including：
According to identified Character segmentation point and segmentation candidates point, the corresponding alternative word in the license plate image region is obtained Accord with region segmentation mode.
7. according to the method described in claim 5, it is characterized in that, described according to each upright projection value and corresponding first threshold After the step of being worth, determining Character segmentation point, the method further includes：
According to most stable extremal region algorithm, the stability region in the license plate image region is obtained；
It is non-character cut-point to set the Character segmentation point inside each stability region；
Character segmentation point determined by the basis obtains the corresponding alternative character zone segmentation side in the license plate image region The step of formula, including：
According to remaining Character segmentation point, the corresponding alternative character zone partitioning scheme in the license plate image region is obtained.
8. according to the method described in claim 1, it is characterized in that, Character segmentation point determined by the basis, described in acquisition The step of corresponding alternative character zone partitioning scheme in license plate image region, including：
According to identified Character segmentation point, character zone to be selected is obtained；
The character zone to be selected that width is in the range of [w1, w2] is determined as target character region, wherein, the w1 is default First width threshold value, the w2 are default second width threshold value, and the w2 is not less than the w1；
According to identified target character region, the corresponding alternative character zone segmentation side in the license plate image region is obtained Formula.
9. a kind of license plate recognition device, which is characterized in that described device includes：
Image-region obtains module, for obtaining the license plate image region of the number-plate number to be identified；
Cut-point determining module, for determining the Character segmentation point in the license plate image region；
Partitioning scheme obtains module, for according to identified Character segmentation point, it is corresponding standby to obtain the license plate image region The character zone partitioning scheme of choosing；
For being directed to each alternative character zone partitioning scheme, word is carried out to the license plate image region for character recognition module Symbol identification, obtains character identification result；
The number-plate number obtains module, for according to the character identification result obtained, it is corresponding to obtain the license plate image region The number-plate number.
10. device according to claim 9, which is characterized in that the cut-point determining module, including：
Projection value determination sub-module, for according to vertical projection method, determining hanging down for each pixel column in the license plate image region Straight projection value；
Cut-point determination sub-module, for according to identified upright projection value, determining Character segmentation point.
11. device according to claim 10, which is characterized in that the cut-point determination sub-module, including：
Width determination unit, for determining the width of sliding window；
Threshold computation unit, for according to the width and identified each upright projection value, calculating each upright projection value Corresponding first threshold；
Cut-point determination unit, for according to each upright projection value and corresponding first threshold, determining Character segmentation point；
Wherein, the threshold computation unit, is specifically used for：
The corresponding first threshold of each upright projection value is calculated in the following way：
According to the width, according to putting in order for upright projection value, selection includes continuous the including target vertical projection value One quantity upright projection value, wherein, the upright projection value puts in order and pixel column in the license plate image region It puts in order consistent, the target vertical projection value is：One in identified upright projection value；
The average value of selected upright projection value is calculated, and the average value is determined as the target vertical projection value and is corresponded to First threshold.
12. according to the devices described in claim 11, which is characterized in that the width determination unit is specifically used for：
The height in the license plate image region is obtained, according to the height, determines the width of sliding window.
13. according to the devices described in claim 11, which is characterized in that the type of Character segmentation point include it is left segmentation vertex type and Right segmentation vertex type；The upright projection value is obtained according to the binaryzation license plate image region of wrongly written or mispronounced character black matrix；
The cut-point determination unit, is specifically used for：
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as left cut-point The Character segmentation point of type：
Proj (i) ＜ proj_th (i) and proj (i+1) >=proj_th (i+1)
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as right cut-point The Character segmentation point of type：
Proj (i) >=proj_th (i) and proj (i+1) ＜ proj_th (i+1)
Wherein, the proj (i) is i-th of upright projection value, and the proj_th (i) is corresponding with i-th of upright projection value First threshold.
14. device according to claim 13, which is characterized in that after the cut-point determination unit, described device Further include candidate point determination unit；The candidate point determination unit, is used for：
By the pixel in the corresponding license plate image region of the upright projection value for meeting following conditions, it is determined as the car plate The segmentation candidates point of image-region：
Proj (i)-proj_th (i) ＜ Th
Wherein, the Th is default second threshold；
In the following way, the type of each segmentation candidates point is determined：
The distance between target candidate cut-point and each Character segmentation point value are calculated, by the corresponding Character segmentation point of lowest distance value Type be determined as the type of the target candidate cut-point, wherein, the target candidate cut-point is identified candidate point One in cutpoint；
The partitioning scheme obtains module, is specifically used for：
According to identified Character segmentation point and segmentation candidates point, the corresponding alternative word in the license plate image region is obtained Accord with region segmentation mode.
15. device according to claim 13, which is characterized in that after the cut-point determination unit, described device It further includes：
Stability region obtaining unit, for according to most stable extremal region algorithm, obtaining the stable region in the license plate image region Domain；
Cut-point setting unit is non-character cut-point for setting the Character segmentation point inside each stability region；
The partitioning scheme obtains module, is specifically used for：
According to remaining Character segmentation point, the corresponding alternative character zone partitioning scheme in the license plate image region is obtained.
16. device according to claim 9, which is characterized in that the partitioning scheme obtains module, including：
Treat that favored area obtains submodule, for according to identified Character segmentation point, obtaining character zone to be selected；
Target area determination sub-module, the character zone to be selected for width to be in the range of [w1, w2] are determined as target word Region is accorded with, wherein, the w1 is default first width threshold value, and the w2 is default second width threshold value, and the w2 is not less than institute State w1；
Partitioning scheme obtains subelement, for according to identified target character region, obtaining the license plate image region and corresponding to Alternative character zone partitioning scheme.
CN201611052182.2A 2016-11-24 2016-11-24 License plate recognition method and device Active CN108108734B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201611052182.2A CN108108734B (en) 2016-11-24 2016-11-24 License plate recognition method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201611052182.2A CN108108734B (en) 2016-11-24 2016-11-24 License plate recognition method and device

Publications (2)

Publication Number Publication Date
CN108108734A true CN108108734A (en) 2018-06-01
CN108108734B CN108108734B (en) 2021-09-24

Family

ID=62204484

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201611052182.2A Active CN108108734B (en) 2016-11-24 2016-11-24 License plate recognition method and device

Country Status (1)

Country Link
CN (1) CN108108734B (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108960244A (en) * 2018-07-10 2018-12-07 北京智芯原动科技有限公司 A kind of character projection localization method and device of license plate
CN109034145A (en) * 2018-07-06 2018-12-18 西安电子科技大学 Bank's card number recognition methods based on OpenCV
CN109063702A (en) * 2018-08-08 2018-12-21 北京旷视科技有限公司 Licence plate recognition method, device, equipment and storage medium
CN109977959A (en) * 2019-03-29 2019-07-05 国家电网有限公司 A kind of train ticket character zone dividing method and device
CN110728687A (en) * 2019-10-15 2020-01-24 卓尔智联(武汉)研究院有限公司 File image segmentation method and device, computer equipment and storage medium
CN110728276A (en) * 2018-07-16 2020-01-24 杭州海康威视数字技术股份有限公司 License plate recognition method and device
CN111696364A (en) * 2020-06-22 2020-09-22 沈志强 Roadside parking management method
CN112232354A (en) * 2020-11-23 2021-01-15 中国第一汽车股份有限公司 Character recognition method, device, equipment and storage medium
CN112997190A (en) * 2020-12-29 2021-06-18 深圳市锐明技术股份有限公司 License plate recognition method and device and electronic equipment

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6037082A (en) * 1983-08-05 1985-02-26 Ricoh Co Ltd Character cutting system
JPS61206088A (en) * 1985-03-11 1986-09-12 Mitsubishi Electric Corp Character reading device
CN101770576A (en) * 2008-12-31 2010-07-07 北京新岸线网络技术有限公司 Method and device for extracting characters
CN101877050A (en) * 2009-11-10 2010-11-03 青岛海信网络科技股份有限公司 Automatic extracting method for characters on license plate
CN102129560A (en) * 2010-01-18 2011-07-20 富士通株式会社 Method and device for identifying characters
CN102254157A (en) * 2011-07-07 2011-11-23 北京文通图像识别技术研究中心有限公司 Evaluating method for searching character segmenting position between two adjacent characters
CN102402686A (en) * 2011-12-07 2012-04-04 北京云星宇交通工程有限公司 Method for dividing license plate characters based on connected domain analysis
CN102693431A (en) * 2012-05-31 2012-09-26 信帧电子技术（北京）有限公司 Method and device for identifying type of white number plate
CN103207998A (en) * 2012-12-24 2013-07-17 电子科技大学 License plate character segmentation method based on support vector machine
CN103530600A (en) * 2013-06-06 2014-01-22 东软集团股份有限公司 License plate recognition method and system under complicated illumination

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6037082A (en) * 1983-08-05 1985-02-26 Ricoh Co Ltd Character cutting system
JPS61206088A (en) * 1985-03-11 1986-09-12 Mitsubishi Electric Corp Character reading device
CN101770576A (en) * 2008-12-31 2010-07-07 北京新岸线网络技术有限公司 Method and device for extracting characters
CN101877050A (en) * 2009-11-10 2010-11-03 青岛海信网络科技股份有限公司 Automatic extracting method for characters on license plate
CN102129560A (en) * 2010-01-18 2011-07-20 富士通株式会社 Method and device for identifying characters
CN102254157A (en) * 2011-07-07 2011-11-23 北京文通图像识别技术研究中心有限公司 Evaluating method for searching character segmenting position between two adjacent characters
CN102402686A (en) * 2011-12-07 2012-04-04 北京云星宇交通工程有限公司 Method for dividing license plate characters based on connected domain analysis
CN102693431A (en) * 2012-05-31 2012-09-26 信帧电子技术（北京）有限公司 Method and device for identifying type of white number plate
CN103207998A (en) * 2012-12-24 2013-07-17 电子科技大学 License plate character segmentation method based on support vector machine
CN103530600A (en) * 2013-06-06 2014-01-22 东软集团股份有限公司 License plate recognition method and system under complicated illumination

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
张爱娟: "基于结构特征点的字符分割技术的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 *
李晨: "基于机器视觉的手写体区域检测和字符分割的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 *

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109034145A (en) * 2018-07-06 2018-12-18 西安电子科技大学 Bank's card number recognition methods based on OpenCV
CN109034145B (en) * 2018-07-06 2021-11-09 西安电子科技大学 Bank card number identification method based on OpenCV
CN108960244A (en) * 2018-07-10 2018-12-07 北京智芯原动科技有限公司 A kind of character projection localization method and device of license plate
CN110728276A (en) * 2018-07-16 2020-01-24 杭州海康威视数字技术股份有限公司 License plate recognition method and device
CN109063702A (en) * 2018-08-08 2018-12-21 北京旷视科技有限公司 Licence plate recognition method, device, equipment and storage medium
CN109977959A (en) * 2019-03-29 2019-07-05 国家电网有限公司 A kind of train ticket character zone dividing method and device
CN110728687A (en) * 2019-10-15 2020-01-24 卓尔智联(武汉)研究院有限公司 File image segmentation method and device, computer equipment and storage medium
CN110728687B (en) * 2019-10-15 2022-08-02 卓尔智联(武汉)研究院有限公司 File image segmentation method and device, computer equipment and storage medium
CN111696364A (en) * 2020-06-22 2020-09-22 沈志强 Roadside parking management method
CN112232354A (en) * 2020-11-23 2021-01-15 中国第一汽车股份有限公司 Character recognition method, device, equipment and storage medium
CN112997190A (en) * 2020-12-29 2021-06-18 深圳市锐明技术股份有限公司 License plate recognition method and device and electronic equipment
CN112997190B (en) * 2020-12-29 2024-01-12 深圳市锐明技术股份有限公司 License plate recognition method and device and electronic equipment

Also Published As

Publication number Publication date
CN108108734B (en) 2021-09-24

Similar Documents

Publication Publication Date Title
CN108108734A (en) A kind of licence plate recognition method and device
CN110991465B (en) Object identification method, device, computing equipment and storage medium
Khalid et al. Automatic measurement of the traffic sign with digital segmentation and recognition<? show [AQ ID= Q1]?>
CN104616009A (en) Character cutting and recognizing method
CN108229466B (en) License plate recognition method and device
CN108268867A (en) A kind of license plate locating method and device
CN108197644A (en) A kind of image-recognizing method and device
Azad et al. New method for optimization of license plate recognition system with use of edge detection and connected component
CN108073926A (en) A kind of licence plate recognition method and device
CN108090484B (en) License plate recognition method and device
CN107578011A (en) The decision method and device of key frame of video
Arafat et al. Connected component analysis integrated edge based technique for automatic vehicular license plate recognition framework<? show [AQ ID= Q1]?>
CN110969164A (en) Low-illumination imaging license plate recognition method and device based on deep learning end-to-end
CN108073925B (en) License plate recognition method and device
CN108205670A (en) A kind of licence plate recognition method and device
Chaturvedi et al. Automatic license plate recognition system using surf features and rbf neural network
CN116721396A (en) Lane line detection method, device and storage medium
Vidhyalakshmi et al. Text detection in natural images with hybrid stroke feature transform and high performance deep Convnet computing
CN113657378B (en) Vehicle tracking method, vehicle tracking system and computing device
CN110852353A (en) Intersection classification method and equipment
Deb et al. Automatic vehicle identification by plate recognition for intelligent transportation system applications
CN115984859A (en) Image character recognition method and device and storage medium
CN112446375A (en) License plate recognition method, device, equipment and storage medium
CN111325199A (en) Character inclination angle detection method and device
CN111639640B (en) License plate recognition method, device and equipment based on artificial intelligence

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