CN109034157A - Licence plate recognition method and device - Google Patents
Licence plate recognition method and device Download PDFInfo
- Publication number
- CN109034157A CN109034157A CN201710428442.XA CN201710428442A CN109034157A CN 109034157 A CN109034157 A CN 109034157A CN 201710428442 A CN201710428442 A CN 201710428442A CN 109034157 A CN109034157 A CN 109034157A
- Authority
- CN
- China
- Prior art keywords
- character
- picture
- license plate
- width
- mean breadth
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
-
- 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
- G06V10/273—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 removing elements interfering with the pattern to be recognised
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/63—Scene text, e.g. street names
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/158—Segmentation of character regions using character size, text spacings or pitch estimation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/625—License plates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Character Input (AREA)
Abstract
The present invention provides a kind of licence plate recognition method and devices, wherein this method comprises: carrying out Character segmentation to picture to be identified；According to character mean breadth, determines in the character divided and obtained and meet whether the character of preset requirement is 7；If it is, determining that the picture to be identified is license plate picture, and license plate number identification will be carried out from 7 character input Car license recognition models being partitioned into the picture to be identified；If it is not, then determining that the picture to be identified is non-license plate picture.The present invention solves the problems, such as often to have unusual character in existing Car license recognition model or is that non-license plate picture is as input, has reached the technical effect that the non-license plate picture of reduction enters the probability of Car license recognition model.
Description
Technical field
The present invention relates to intelligent identification technology field, in particular to a kind of licence plate recognition method and device.
Background technique
Currently, the security control etc. that automobile passes in and out in parking automatic charging, cell, generally all passes through automatic identification license plate
Mode realize.For example, since a vehicle will do it Car license recognition in inlet, to determine whether being fair entering toll zone
Permitted, in the case where determining allows into, to record the license plate number of the vehicle and the time of entrance, and the vehicle of letting pass into interior vehicle
, it when vehicle is driven out to out of region, again identifies that the license plate number, then determines time of the vehicle in region, realization stops
Calculating of fare etc..
In this process, it is necessary to realize the automatic identification to license plate number.Currently, in the identification process to license plate,
General is exactly after being split to character, to be input to artificial mind as the character to be identified by from left to right 7 characters of number
Through carrying out the specific identification of character content in network (Artificial Neural Network, abbreviation ANN) model, to realize
Identification to license plate.Wherein, ANN be with the activity of mathematical model imictron, be based on imitate cerebral nerve network structure and
Function and a kind of information processing system established.
However, above-mentioned identification method has the following problems:
1) occasionally there are some defects for the character after dividing, for example, some character contents have incompleteness；
2) there are non-license plates to be divided into several characters, into being known in the model of the artificial neural network of Car license recognition
Not.
In view of the above-mentioned problems, not yet putting forward effective solutions.
Summary of the invention
The embodiment of the invention provides a kind of licence plate recognition methods, enter Car license recognition model to reduce non-license plate picture
Probability, this method comprises:
Character segmentation is carried out to picture to be identified；
According to character mean breadth, determines in the character divided and obtained and meet whether the character of preset requirement is 7；
If it is, determining that the picture to be identified is license plate picture, and 7 will be partitioned into from the picture to be identified
License plate number identification is carried out in a character input Car license recognition model；
If it is not, then determining that the picture to be identified is non-license plate picture.
In one embodiment, after carrying out Character segmentation to picture to be identified, the above method can also include:
It deletes character width in the character that segmentation obtains and is less than the mean breadth half, either, white pixel
Height is less than the character of character height one third.
In one embodiment, it according to character mean breadth, determines in the character divided and obtained and meets preset requirement
Whether character is 7, comprising:
The character for being less than the mean breadth to character width merges processing；
After determining merging treatment, whether the character for meeting preset requirement is 7.
In one embodiment, the character for being less than the mean breadth to character width merges processing, comprising:
The character obtained to segmentation is from left to right successively handled as follows:
Determine whether the width of current character is less than the mean breadth；
It is less than the mean breadth, and the current character and right phase in the width of the right adjacent character of the current character
It is in the case that the centre distance of adjacent character is less than 1.2 times of the mean breadth, the current character is adjacent with the right side
Character merge into a character.
In one embodiment, by the current character and the adjacent character in the right side merge into a character it
Afterwards, the method also includes:
Determine whether the width of the character after merging is less than the mean breadth；
The width of the right adjacent character of character after the merging is less than the mean breadth, and the word after the merging
It, will in the case where according with 1.2 times that are less than the mean breadth with the centre distance of the right adjacent character of the character after described merge
Character after the merging merges into a character with the right adjacent character of the character after described merge.
In one embodiment, it according to character mean breadth, determines in the character divided and obtained and meets preset requirement
Whether character is 7, comprising:
Each character obtained to segmentation is handled as follows:
Determine whether the width of current character is less than 0.5 times of the mean breadth；
In the case where determining that the width of the current character is less than 0.5 times of the mean breadth, by the current word
The width of symbol extends to the mean breadth.
In one embodiment, the width of the current character is extended into the mean breadth, comprising:
Obtain the pore size in the left side of the current character and the pore size on right side；
Determine the pore size in the left side of the current character and the pore size on right side, if enough by the current word
The width of symbol extends to the mean breadth；
In the case where determining inadequate situation, the width of the current character is extended into the current character and adjacent character
Boundary.
In one embodiment, it according to character mean breadth, determines in the character divided and obtained and meets preset requirement
Whether character is 7, comprising:
It since intermediate character, searches for the left, searches the spacing between the first quantity adjacent character and be less than described put down
1.25 times of character of equal width, is searched for the right, searches the spacing between the second quantity adjacent character less than described average
In the case where 1.25 times of character of width, if first quantity and second quantity and be 6, it is determined that it is described to
Identification picture is license plate picture, if first quantity and second quantity and be not 6, it is determined that the figure to be identified
The non-license plate picture of piece.
In one embodiment, the determination first quantity and second quantity and in the case where being 5, really
Determine the leftmost side and meet the character of condition whether being in boundary, if be in, fills out a character width to the left, then cut
Filled character is taken, and identifies whether the character of the filling is characters on license plate by Car license recognition model, if it is, determining
The picture to be identified is license plate picture.
The embodiment of the invention also provides a kind of license plate recognition devices, enter Car license recognition model to reduce non-license plate picture
Probability, which includes:
Divide module, for carrying out Character segmentation to picture to be identified；
First determining module, for determining in the character divided and obtained and meeting preset requirement according to character mean breadth
Whether character is 7；
Input module, in the case that the character for meeting preset requirement in determining the character that segmentation obtains is 7, really
The picture to be identified is determined for license plate picture, and the 7 character input Car license recognitions that will be partitioned into from the picture to be identified
License plate number identification is carried out in model；
Second determining module, the character for meeting preset requirement in determining the character that segmentation obtains are not 7 feelings
Under condition, determine that the picture to be identified is non-license plate picture.
In embodiments of the present invention, it is contemplated that license plate is usually 7 characters, and can based on character mean breadth determine to
In identification image therefore the number of character can determine the word met the requirements in images to be recognized by the mean breadth of character
Whether symbol is 7, so that it is determined that whether picture to be identified is license plate picture, to filter out non-license plate picture, to solve
Often there is a unusual character in existing Car license recognition model or be non-license plate picture problem as input, has reached reduction
Non- license plate picture enters the technical effect of the probability of Car license recognition model.
Detailed description of the invention
The drawings described herein are used to provide a further understanding of the present invention, constitutes part of this application, not
Constitute limitation of the invention.In the accompanying drawings:
Fig. 1 is the method flow diagram of licence plate recognition method according to an embodiment of the present invention；
Fig. 2 is the method flow diagram of the specific example of licence plate recognition method according to an embodiment of the present invention；
Fig. 3 is the structural block diagram of license plate recognition device according to an embodiment of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, right below with reference to embodiment and attached drawing
The present invention is described in further details.Here, exemplary embodiment and its explanation of the invention is used to explain the present invention, but simultaneously
It is not as a limitation of the invention.
It is first as follows to some concept explanations being related to before being illustrated to licence plate recognition method of the invention, with
Just the application is more fully understood:
1) ANN model is a kind of classifier of neural network detection or identification.
2) characters on license plate, the character one in license plate share 71 characters.
3) ANN model identifies score, that is, a score is had after identifying by ANN model, score is generally arrived -1.05
Between 1.05, scores nearer it is to 1, then character to be identified be the character a possibility that it is higher.
4) character external world rectangle refers to character actually external minimum rectangle.
During existing Car license recognition, there are there is improper character in the character of input, or incomplete
Character, or be that character in non-license plate picture etc. is input in identification model as input content.
For this purpose, in view of license plate is usually 7 characters, and word in images to be recognized can be determined based on character mean breadth
Therefore the number of symbol can determine whether the character met the requirements in images to be recognized is 7 by the mean breadth of character,
So that it is determined that whether picture to be identified is license plate picture, to filter out non-license plate picture.Based on this, provide in this example
A kind of licence plate recognition method, as shown in Figure 1, may include steps of:
Step 101: Character segmentation is carried out to picture to be identified；
In view of that can search and delete dot or not meet the character of character requirement, to delete in picture to be identified
Illegal character, to improve the accuracy of subsequent Car license recognition.It in one embodiment, can be by the character after segmentation
In, character width is less than the half of average character, and white pixel height is less than the one third of oneself height, white pixel area
The character occupied lower than a quarter is deleted, that is, using such character as unusual character.For example, carrying out word
After symbol segmentation, character width in the character that segmentation obtains can be deleted and be less than the mean breadth half, it is either, white
Color pixel height is less than the character of character height one third.
Step 102: according to character mean breadth, determine meet in character that segmentation obtains preset requirement character whether be
7；
In view of some characters in license plate, such as " Hunan ", the either characters such as " saliva ", it is made of multiple character fragments
, wherein " Hunan " is to consist of three parts from left to right, " saliva " from left to right consists of two parts.Carry out Character segmentation when
It waits, Hunan may be divided into three characters, and saliva is divided into two characters.
Based on this, in one embodiment, processing can be merged to character, that is, the three of the same word will be belonged to
A part is combined into a word.I.e., it is possible to which the character for being less than the mean breadth to character width merges processing, based on merging
Afterwards as a result, whether the character for meeting preset requirement is 7.
For example, the character that can be obtained to segmentation is from left to right successively handled as follows, to realize that character merges:
S1: determine whether the width of current character is less than the mean breadth；
S2: being less than the mean breadth in the width of the right adjacent character of the current character, and the current character with
In the case that the centre distance of right adjacent character is less than 1.2 times of the mean breadth, by the current character and the right side
Adjacent character merges into a character.
That is, the centre distance between the character length and adjacent character that pass through each character determines the need for carrying out character
Merge.
In view of there is the character of three part compositions sometimes, therefore, the current character is adjacent with the right side
Character merge into after a character, can also continue to determine whether the width of the character after merging is less than the average width
Degree；The width of the right adjacent character of character after the merging is less than the mean breadth, and the character after the merging with
It, will be described in the case that the centre distance of the right adjacent character of character after the merging is less than 1.2 times of the mean breadth
Character after merging merges into a character with the right adjacent character of the character after described merge.
In one embodiment, it according to character mean breadth, determines in the character divided and obtained and meets preset requirement
Whether character is 7, may include: that the obtained each character of segmentation is handled as follows: determine current character width whether
Less than 0.5 times of the mean breadth；The case where the width for determining the current character is less than 0.5 times of the mean breadth
Under, the width of the current character is extended into the mean breadth.I.e., it is possible to the character, such as " 1 " inadequate to width etc. into
Row expansion processing.
When carrying out expanding processing, there is a situation where that left and right gap is smaller sometimes, even if left and right gap is all used
On preset mean breadth is also not achieved, then the character can be extended to the boundary of adjacent character.I.e., it is possible to obtain described work as
The pore size in the left side of preceding character and the pore size on right side；Determine pore size and the right side in the left side of the current character
Pore size, if the width of the current character is extended into the mean breadth enough；Determining inadequate situation
Under, the width of the current character is extended to the boundary of the current character and adjacent character.
Step 103: if it is, determining that the picture to be identified is license plate picture, and will be from the picture to be identified
License plate number identification is carried out in 7 character input Car license recognition models being partitioned into；
Step 104: if it is not, then determining that the picture to be identified is non-license plate picture.
In view of when carrying out picture recognition, it is thus necessary to determine that going out 7 effective characters therefore can be from middle word
Symbol starts, and searches for the left, searches spacing between the first quantity adjacent character and is less than 1.25 times of the mean breadth
Character is searched for the right, and 1.25 times of the word that the spacing between the second quantity adjacent character is less than the mean breadth is searched
In the case where symbol, if first quantity and second quantity and be 6, it is determined that the picture to be identified is license plate figure
Piece, if first quantity and second quantity and be not 6, it is determined that the non-license plate picture of picture to be identified.
Sometimes can also exist and identify situation of problems, in one embodiment, can determine first number
Amount and second quantity and whether be in boundary in the case where 5, to determine that the leftmost side meets the character of condition, if place
In then filling out a character width to the left, then intercept filled character, and identify that this is filled out by Car license recognition model
Whether the character filled is characters on license plate, if it is, determining that the picture to be identified is license plate picture.
Above-mentioned licence plate recognition method is illustrated below with reference to a specific embodiment, it is important to note, however, that should
Specific embodiment does not constitute the specific restriction to the application merely to the application is better described.
As shown in Fig. 2, the identification process of license plate may comprise steps of:
S1: the mean breadth of characters on license plate is calculated；
S2: small particles and abnormal character in license plate are deleted；
S3: processing is merged to the Chinese character for being divided into several parts；
S4: dividing processing and reparation again is carried out to character；
S5: using the inside character rule of license plate, the non-license plate picture and one, left side character for deleting doubtful license plate are lost
Chase after and look for processing.
These steps are specifically described below as follows:
S1: the mean breadth of characters on license plate is calculated:
Selection for picture can choose the picture of a scheduled long wide scope as picture to be identified, and control
The piece right boundary that charts is no more than the width of three characters, while picture being allowed to be non-license plate, and institute is had already passed through using picture
Horizontal correction and vertical correction, are the pictures that there is no problem.
Based on images to be recognized, the object that can limit statistics is limited in character width between 8-22, wide for character
The case where degree is less than 2 characters can make at non-license plate this picture because not meeting the number requirement of characters on license plate
Reason.This is primarily due to using average character width, when a character, loses average meaning, it is possible to will only count on
The license plate of one character is considered as non-license plate.
Specifically, can be averaging by overall width, the character number of statistics character to the two, to obtain average width
Degree, which is exactly the character mean breadth of the license plate acquired.
S2: small particles and abnormal character in license plate are deleted:
The step is primarily to suppressing exception character, that is, searches and deletes dot or do not meet the word of character requirement
Symbol.In one embodiment, character width can be less than to the half of average character, white pixel height is less than oneself height
One third, white pixel area occupies the character lower than a quarter and deleted, that is, using such character as
Unusual character.
S3: processing is merged to the Chinese character for being divided into several parts:
The step allows for some Chinese characters with discontinuity, therefore, can to divided Chinese character into
Row merging treatment.
Specifically, the character that first average character width can be less than to character width is handled, for example, from left to right according to
Secondary traversal, when there is a character for being less than average character width, then character width adjacent to its right side judges, if
Adjacent character width is less than average character width, and two character center distances are less than 1.2 times of average character width, then can
The two characters are merged into a character.It is then possible to the width size of new composite characters be judged again, if after merging
Character still less than average character width, then, then the right side adjacent character of the character after the merging is judged, if it is less than
Character mean breadth, and adjacent character centre distance is less than the 1.2 of average character width on the right side of newly synthesized character center language
Times, then, it merges again.
Wherein, merging number at most twice, this is primarily due to some characters in Chinese character and is at most divided into three,
Through the above way from left to right, the character that may be partitioned from is handled, successively to realize accurately identifying for Chinese.
S4: dividing processing and reparation again is carried out to character:
After carrying out above-mentioned merging, dividing processing can be re-started, specifically, in view of in two neighboring character,
The right side for the left character of two neighboring character having is adjacent with the left border of right character, with the presence of the certain gap of meeting.For example,
Adjacent several 1 will have that mutual void ratio is larger, in order to enable subsequent more square when be filled
Just, text filling can be carried out in advance.In view of that there can be the reason of very big gap between adjacent character, it is primarily due to propose
Just realize therefore, there can be gap when character by boundary rectangle mode.
Character is carried out dividing processing again and repairs can be to realize in the following way, first, it is determined that character is wide
Whether degree is less than 0.8 times of character width, if it is less than the half of character width (for example, character " 1 " or some cutting away one
Partial character), then expanding the character both sides.Specifically, can first judge character left and right side when expanding
How many each gap, then according to left or right side gap how much, subject to the width to the character width that is averaged of escape character, if
Gap cannot be met the requirements, then adjacent character boundary is subject to, until extending to boundary.
It, can be in the hope of the center spacing between adjacent character, if the center spacing of adjacent character is less than after having expanded
Gap is divided into two parts then calculate the gap distance of the two characters by 1.2 times of average character width, and adjacent two
Each half gap in side, if the center spacing of adjacent character is greater than or equal to 1.2 times of average character width, to the two
Gap between adjacent character is without processing.
By way of this separating character again, incomplete character can be modified, so that subsequent character recognition is more held
Easily.
S5: using the inside character rule of license plate, the non-license plate picture and one, left side character for deleting doubtful license plate are lost
Chase after and look for processing.
In view of there are scheduled setting rules for character in license plate, for example, character pitch can not be more than that average character is wide
1.25 times of degree, if it does, then not being long license plate.Based on this rule, the non-word of characters on license plate or so can be deleted
The case where symbol, finds the right boundary of license plate.
Specifically, can determine whether to meet license plate rule in the following way:
Since an intermediate character, to two-sided search, the distance between each adjacent two character is calculated, if apart from big
In 1.2 times, then stop search.Specifically, being started with an intermediate character, search for the left, if search n character meet away from
From requiring, the distance between (n+1)th character and n-th of character are greater than 1.25 times of average character width, then stop search,
Record searches n character.Started with one character in centre, is searched for the right, if searching m character meets required distance,
The distance between m+1 character and m-th of character be greater than average character width 1.2 times (due to two characters it is maximum away from
From in the left side of license plate, no longer right side), then stopping search, record searches m character.
By above-mentioned search step, the available m+n+1 characters conformed to, if m+n+1 is equal to 7, this
Character number meets characters on license plate rule, shows that this picture is license plate picture, to carry out next step Recognition of License Plate Characters behaviour
Make.If m+n+1 is greater than 7, show that this character number does not meet characters on license plate rule, it can be by extra m+n-6 word out
Symbol is identified using ANN model, m+n-6 character of identification is successively carried out from left to right, if recognition result score is both greater than
0.85 point, then showing that these characters are all characters on license plate.Then, this m+n-6 character then is from right to left identified, if known
Other result score is both greater than 0.85 point, then showing that these characters are all characters on license plate, if the character finally identified
Number then illustrates that this is not license plate more than 7.If m+n-6, left side and m+n-6, right side character are not characters after identification,
So character number does not meet character requirement less than 7, then this this doubtful license plate picture is not license plate.If m+n+1
Equal to 6, if first character is in boundary position, it can according to need and expansion picture is carried out to doubtful license plate, according to
Average character width, is estimated a character width character to the left, intercepts out character, identified using ANN model, if
Divide and be greater than 0.9 point, then being exactly to determine that this character is the character required supplementation with, if it is less than 0.9 point, then not being just symbol
Desired character is closed, this doubtful license plate picture is also not license plate picture.If m+n+1, less than 6, these characters are corresponding
Doubtful license plate picture is not license plate picture.
In upper example, by deleting abnormal character, merging Chinese character and repairing incomplete character, and according to word in license plate
Accord with the doubtful license plate of redundant rule elimination in non-license plate picture, to left margin lack a character license plate picture carry out supplement detection trace
Character is lacked, license plate picture is accurately identified to realize, so that being input to the picture of final license plate number identification model more
Rationally, a possibility that avoiding the input of wrong picture.
Based on the same inventive concept, a kind of license plate recognition device is additionally provided in the embodiment of the present invention, such as following implementation
Described in example.Since the principle that license plate recognition device solves the problems, such as is similar to licence plate recognition method, the reality of license plate recognition device
The implementation that may refer to licence plate recognition method is applied, overlaps will not be repeated.It is used below, term " unit " or " mould
The combination of the software and/or hardware of predetermined function may be implemented in block ".Although device described in following embodiment is preferably with soft
Part is realized, but the realization of the combination of hardware or software and hardware is also that may and be contemplated.Fig. 3 is of the invention real
A kind of structural block diagram of the license plate recognition device of example is applied, as shown in figure 3, may include: segmentation module 301, the first determining module
302, input module 303 and the second determining module 304, are below illustrated the structure.
Divide module 301, for carrying out Character segmentation to picture to be identified；
First determining module 302, for determining in the character divided and obtained and meeting preset requirement according to character mean breadth
Character whether be 7；
Input module 303, the situation that the character for meeting preset requirement in determining the character that segmentation obtains is 7
Under, determine that the picture to be identified is license plate picture, and 7 character input license plates that will be partitioned into from the picture to be identified
License plate number identification is carried out in identification model；
Second determining module 304 is not 7 for meeting the character of preset requirement in determining the character that segmentation obtains
In the case of, determine that the picture to be identified is non-license plate picture.
In one embodiment, above-mentioned license plate recognition device further includes removing module, for picture to be identified into
After line character segmentation, deletes character width in the character for determining and dividing and obtaining and be less than the mean breadth half, or
It is that white pixel height is less than the character of character height one third.
In one embodiment, the first determining module 302 specifically can be less than the mean breadth to character width
Character merges processing；After determining merging treatment, whether the character for meeting preset requirement is 7.
In one embodiment, the character that the first determining module 302 can obtain segmentation from left to right successively carries out
Following processing: determine whether the width of current character is less than the mean breadth；In the right adjacent character of the current character
Whether width is less than the mean breadth, and the current character and the centre distance of right adjacent character are less than the average width
In the case where 1.2 times of degree, the current character and the adjacent character in the right side are merged into a character.
In one embodiment, the first determining module 302 is closed by the adjacent character of the current character and the right side
And for that can also determine whether the width of the character after merging is less than the mean breadth after character；In the merging
Whether the width of the right adjacent character of rear character is less than the mean breadth, and the character after the merging with after described merge
The right adjacent character of character centre distance be less than 1.2 times of the mean breadth in the case where, by the word after the merging
It accords with the character adjacent with the right side of the character after described merge and merges into a character.
In one embodiment, each character that the first determining module 302 can specifically obtain segmentation is located as follows
Reason: determine whether the width of current character is less than 0.5 times of the mean breadth；It is less than in the width for determining the current character
In the case where 0.5 times of the mean breadth, the width of the current character is extended into the mean breadth.
In one embodiment, the gap in the left side of the specific available current character of the first determining module 302
The pore size of size and right side；Determine the pore size in the left side of the current character and the pore size on right side, if foot
It is enough that the width of the current character is extended into the mean breadth；In the case where determining inadequate situation, by the current character
Width extend to the boundary of the current character and adjacent character.
In one embodiment, the first determining module 302 can search for the left since intermediate character, search
Spacing between one quantity adjacent character is less than 1.25 times of character of the mean breadth, searches for the right, searches second
In the case that spacing between quantity adjacent character is less than 1.25 times of character of the mean breadth, if first number
Amount with second quantity and be 6, it is determined that the picture to be identified be license plate picture, if first quantity with it is described
Second quantity and not be 6, it is determined that the non-license plate picture of picture to be identified.
In one embodiment, the first determining module 302 determine first quantity and second quantity and
In the case where 5, it can determine that the leftmost side meets the character of condition and whether is in boundary, if be in, fill out to the left
Then one character width intercepts filled character, and identifies whether the character of the filling is vehicle by Car license recognition model
Board character, if it is, determining that the picture to be identified is license plate picture.
In another embodiment, a kind of software is additionally provided, the software is for executing above-described embodiment and preferred reality
Apply technical solution described in mode.
In another embodiment, a kind of storage medium is additionally provided, above-mentioned software is stored in the storage medium, it should
Storage medium includes but is not limited to: CD, floppy disk, hard disk, scratch pad memory etc..
It can be seen from the above description that the embodiment of the present invention realizes following technical effect: in view of license plate is general
It is 7 characters, and can determines the number of character in images to be recognized based on character mean breadth, therefore, character can be passed through
Mean breadth determine whether the character met the requirements in images to be recognized is 7, so that it is determined that picture to be identified whether be
License plate picture, to filter out non-license plate picture, thus solve often have in existing Car license recognition model unusual character or
Person is non-license plate picture problem as input, has reached the technology that the non-license plate picture of reduction enters the probability of Car license recognition model
Effect.
Obviously, those skilled in the art should be understood that each module of the above-mentioned embodiment of the present invention or each step can be with
It is realized with general computing device, they can be concentrated on a single computing device, or be distributed in multiple computing devices
On composed network, optionally, they can be realized with the program code that computing device can perform, it is thus possible to by it
Store and be performed by computing device in the storage device, and in some cases, can be held with the sequence for being different from herein
The shown or described step of row, perhaps they are fabricated to each integrated circuit modules or will be multiple in them
Module or step are fabricated to single integrated circuit module to realize.In this way, the embodiment of the present invention be not limited to it is any specific hard
Part and software combine.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field
For art personnel, the embodiment of the present invention can have various modifications and variations.All within the spirits and principles of the present invention, made
Any modification, equivalent substitution, improvement and etc. should all be included in the protection scope of the present invention.
Claims (10)
1. a kind of licence plate recognition method characterized by comprising
Character segmentation is carried out to picture to be identified；
According to character mean breadth, determines in the character divided and obtained and meet whether the character of preset requirement is 7；
If it is, determining that the picture to be identified is license plate picture, and 7 words that will be partitioned into from the picture to be identified
License plate number identification is carried out in symbol input Car license recognition model；
If it is not, then determining that the picture to be identified is non-license plate picture.
2. the method as described in claim 1, which is characterized in that after carrying out Character segmentation to picture to be identified, the side
Method further include:
It deletes character width in the character that segmentation obtains and is less than the mean breadth half, either, white pixel height
Less than the character of character height one third.
3. the method as described in claim 1, which is characterized in that according to character mean breadth, determine in the character divided and obtained
Whether the character for meeting preset requirement is 7, comprising:
The character for being less than the mean breadth to character width merges processing；
After determining merging treatment, whether the character for meeting preset requirement is 7.
4. method as claimed in claim 3, which is characterized in that the character for being less than the mean breadth to character width closes
And it handles, comprising:
The character obtained to segmentation is from left to right successively handled as follows:
Determine whether the width of current character is less than the mean breadth；
It is less than the mean breadth in the width of the right adjacent character of the current character, and the current character and the right side are adjacent
In the case that the centre distance of character is less than 1.2 times of the mean breadth, by the current character and the adjacent word in the right side
Symbol merges into a character.
5. method as claimed in claim 4, which is characterized in that merge by the current character character adjacent with the right side
After a character, the method also includes:
Determine whether the width of the character after merging is less than the mean breadth；
The width of the right adjacent character of character after the merging is less than the mean breadth, and the character after the merging with
It, will be described in the case that the centre distance of the right adjacent character of character after the merging is less than 1.2 times of the mean breadth
Character after merging merges into a character with the right adjacent character of the character after described merge.
6. the method as described in claim 1, which is characterized in that according to character mean breadth, determine in the character divided and obtained
Whether the character for meeting preset requirement is 7, comprising:
Each character obtained to segmentation is handled as follows:
Determine whether the width of current character is less than 0.5 times of the mean breadth；
In the case where determining that the width of the current character is less than 0.5 times of the mean breadth, by the current character
Width extends to the mean breadth.
7. method as claimed in claim 6, which is characterized in that the width of the current character is extended to the average width
Degree, comprising:
Obtain the pore size in the left side of the current character and the pore size on right side；
Determine the pore size in the left side of the current character and the pore size on right side, if enough by the current character
Width extends to the mean breadth；
In the case where determining inadequate situation, the width of the current character is extended to the side of the current character and adjacent character
Boundary.
8. the method as described in claim 1, which is characterized in that according to character mean breadth, determine in the character divided and obtained
Whether the character for meeting preset requirement is 7, comprising:
It since intermediate character, searches for the left, searches the spacing between the first quantity adjacent character and be less than the average width
1.25 times of character of degree, is searched for the right, and the spacing searched between the second quantity adjacent character is less than the mean breadth
1.25 times of character in the case where, if first quantity and second quantity and be 6, it is determined that it is described to be identified
Picture is license plate picture, if first quantity and second quantity and be not 6, it is determined that the picture to be identified is non-
License plate picture.
9. method according to claim 8, which is characterized in that determination first quantity and second quantity and
In the case where 5, determines that the leftmost side meets the character of condition and whether is in boundary, if be in, fill out one to the left
Then character width intercepts filled character, and identifies whether the character of the filling is license plate word by Car license recognition model
Symbol, if it is, determining that the picture to be identified is license plate picture.
10. a kind of license plate recognition device characterized by comprising
Divide module, for carrying out Character segmentation to picture to be identified；
First determining module, for determining the character for meeting preset requirement in the character divided and obtained according to character mean breadth
It whether is 7；
Input module determines institute in the case that the character for meeting preset requirement in determining the character that segmentation obtains is 7
Picture to be identified is stated as license plate picture, and the 7 character input Car license recognition models that will be partitioned into from the picture to be identified
Middle progress license plate number identification；
In the case that second determining module for meeting the character of preset requirement in determining the obtained character of segmentation is not 7,
Determine that the picture to be identified is non-license plate picture.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710428442.XA CN109034157A (en) | 2017-06-08 | 2017-06-08 | Licence plate recognition method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710428442.XA CN109034157A (en) | 2017-06-08 | 2017-06-08 | Licence plate recognition method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109034157A true CN109034157A (en) | 2018-12-18 |
Family
ID=64629209
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710428442.XA Pending CN109034157A (en) | 2017-06-08 | 2017-06-08 | Licence plate recognition method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109034157A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111598104A (en) * | 2020-06-30 | 2020-08-28 | 成都鹏业软件股份有限公司 | License plate character recognition method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102722707A (en) * | 2012-06-11 | 2012-10-10 | 复旦大学 | License plate character segmentation method based on connected region and gap model |
| CN104715252A (en) * | 2015-03-12 | 2015-06-17 | 电子科技大学 | License plate character segmentation method with combination of dynamic template and pixel points |
| CN105426891A (en) * | 2015-12-14 | 2016-03-23 | 广东安居宝数码科技股份有限公司 | Image-based vehicle license plate character segmentation method and system |
| CN105894004A (en) * | 2014-12-16 | 2016-08-24 | 中防通用电信技术有限公司 | Complement technology of locating incomplete license plate aiming at license plate identification system |
-
2017
- 2017-06-08 CN CN201710428442.XA patent/CN109034157A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102722707A (en) * | 2012-06-11 | 2012-10-10 | 复旦大学 | License plate character segmentation method based on connected region and gap model |
| CN105894004A (en) * | 2014-12-16 | 2016-08-24 | 中防通用电信技术有限公司 | Complement technology of locating incomplete license plate aiming at license plate identification system |
| CN104715252A (en) * | 2015-03-12 | 2015-06-17 | 电子科技大学 | License plate character segmentation method with combination of dynamic template and pixel points |
| CN105426891A (en) * | 2015-12-14 | 2016-03-23 | 广东安居宝数码科技股份有限公司 | Image-based vehicle license plate character segmentation method and system |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111598104A (en) * | 2020-06-30 | 2020-08-28 | 成都鹏业软件股份有限公司 | License plate character recognition method and system |
| CN111598104B (en) * | 2020-06-30 | 2023-05-12 | 成都鹏业软件股份有限公司 | License plate character recognition method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110335277A (en) | Image processing method, device, computer readable storage medium and computer equipment | |
| US6741755B1 (en) | System and method providing mixture-based determination of opacity | |
| US7889947B2 (en) | Image completion | |
| CN106960195A (en) | A kind of people counting method and device based on deep learning | |
| CN107330373A (en) | A kind of parking offense monitoring system based on video | |
| US20180053048A1 (en) | Text image processing method and apparatus | |
| CN112200117B (en) | Form identification method and device | |
| CN111480183B (en) | Light field image rendering method and system for generating perspective effect | |
| JP2000112997A (en) | Method for automatically classifying picture into event | |
| CN111951154B (en) | Picture generation method and device containing background and medium | |
| CN109145906B (en) | Target object image determination method, device, equipment and storage medium | |
| CN115565207B (en) | Occlusion scene downlink person detection method with feature simulation fused | |
| CN114519819A (en) | Remote sensing image target detection method based on global context awareness | |
| CN101369312A (en) | Method and equipment for detecting intersection in image | |
| CN109034157A (en) | Licence plate recognition method and device | |
| CN112446375A (en) | License plate recognition method, device, equipment and storage medium | |
| CN112101323B (en) | Method, system, electronic device and storage medium for identifying title list | |
| CN104992176B (en) | A kind of Chinese character extracting method towards an inscription on a tablet | |
| Cahn et al. | Segmentation of cervical cell images. | |
| CN106650732A (en) | License plate identification method and apparatus | |
| CN110135416A (en) | A kind of licence plate recognition method and device | |
| CN113011131B (en) | Typesetting method based on picture electronic book, electronic equipment and storage medium | |
| JPH05334490A (en) | Table recognizing device | |
| CN114917590B (en) | Virtual reality game system | |
| WO2017096670A1 (en) | Fingerprint security protection method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181218 |
|
| RJ01 | Rejection of invention patent application after publication |