CN110516666A - The license plate locating method combined based on MSER and ISODATA - Google Patents
The license plate locating method combined based on MSER and ISODATA Download PDFInfo
- Publication number
- CN110516666A CN110516666A CN201910618457.1A CN201910618457A CN110516666A CN 110516666 A CN110516666 A CN 110516666A CN 201910618457 A CN201910618457 A CN 201910618457A CN 110516666 A CN110516666 A CN 110516666A
- Authority
- CN
- China
- Prior art keywords
- license plate
- rectangle frame
- isodata
- mser
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23211—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with adaptive number of clusters
-
- 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
- G06V10/255—Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/54—Surveillance or monitoring of activities, e.g. for recognising suspicious objects of traffic, e.g. cars on the road, trains or boats
-
- 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
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
- Character Input (AREA)
Abstract
The invention belongs to pattern-recognitions and technical field of image processing, disclose a kind of license plate locating method combined based on MSER and ISODATA, image to be detected is converted into gray level image, it extracts the maximum stable extremal region MSER in gray level image and fits minimum rectangle frame, rectangle frame is screened according to characters on license plate feature, obtains candidate characters rectangle frame；To obtained candidate characters rectangle frame upper left corner ordinate be iterated self-organizing processing ISODATA, it is after cluster as a result, find cluster after containing rectangle frame at most a sample set, be determined as candidate license plate character rectangle frame；License plate up-and-down boundary is determined according to the position of rectangle frame；According to the position of rectangle frame, by sliding window method, the window most comprising rectangle frame is found out, determines right boundary.License plate locating method of the invention bad weather, illumination can be overcome poor etc. unfavorable conditions are accurately positioned, strong robustness.
Description
Technical field
The invention belongs to pattern-recognitions and technical field of image processing, more particularly to one kind to be based on MSER and ISODATA phase
In conjunction with license plate locating method.
Background technique
As the rapid development of economy and traffic causes motor vehicles significantly to increase, this has undoubtedly aggravated the negative of traffic
Load.Therefore, the safety management of automobile can not be ignored, and also promote people to the R&D intensity of automobile identifying system.Up to now,
Vehicle number plate recognition extensive application in terms of parking lot, highway, outpost vehicle detection, safety monitoring, and be intelligence
It can mostly important one of the supervision method of traffic system.Wherein, vehicle license location technique is the key that license plate recognition technology, accurately
License Plate plays the role of subsequent identification process very important.
License plate locating method traditional at present is broadly divided into two major classes, and one kind is the textural characteristics based on characters on license plate, base
Need first to detect the texture of license plate in the positioning of textural characteristics, common skin texture detection method has: edge detection method, Corner Detection
Method, Wavelet Transform, Fourier transform etc..After skin texture detection, it is aided with marginal density scanning, Gray Level Jump, shape
State processing etc. extracts license plate area, is finally accurately positioned according to vehicle license plate characteristic priori knowledge, obvious suitable for license plate,
Background scene changes in little scene, and the different, complicated fields containing a large amount of texture articles different for shooting angle, far and near
Scape image, often there is biggish limitations for first method.Another kind of is the color characteristic based on license plate.Based on color spy
The License Plate of sign depends primarily on the color characteristic of license plate, by certain color space extract license plate color region come
Positioning licence plate.In addition to this, Adaboost, genetic algorithm, neural network etc. can be also applied in License Plate, be only applicable to
The case where image clearly, the scene of uniform illumination, often can not reach satisfactory effect for bad weather, uneven illumination
Fruit.
Such as License Plate of the first kind based on edge detection, limitation is often to want shooting picture
Ask excessively harsh.When picture shooting angle or apart from it is different when, or be in picture containing a large amount of textures when, utilize edge extracting
The method of feature can all lead to edge blurry, can not detect the case where edge or a large amount of edge features can not screen.At this
In the case of sample, will lead to License Plate becomes very difficult.For example application publication number is CN108960243A, title is
The patent application of " license plate locating method ", disclose it is a kind of original license plate image is successively pre-processed, edge extracting, acquisition
The license plate locating method of profile diagram.This method has mainly used edge extracting to obtain profile diagram, then extracts feature to profile diagram,
The method positioning licence plate being finally compared with the character in the license plate classifier prestored output information simultaneously.This method is in spy
Recognition accuracy is high under fixed scene, anti-interference small.But due to the complicated multiplicity of vehicle shooting picture, also it is unable to ensure use
The scene of algorithm, so such detection method and impracticable, in formal use, accuracy rate can be reduced.
And the localization method for the second class based on license plate color feature, it is limited in that scene change not robust.This
Class algorithm compared with first kind algorithm for, requirement for shooting angle reduces, but the requirement for license plate color but becomes
It is very high.In the case that picture is in the cloudy rainy day or is any uneven illumination, the license plate color of shooting can all had larger
Variation, so, the feature extraction based on color can become difficult so that positioning accuracy decline.Such as application publication number
For a kind of license plate locating method based on complicated panorama sketch of CN102999753, positioning accuracy is high under conditions of uniform illumination,
Identification is accurate, but this method has two, and first, when uneven illumination is even, irradiated on license plate by sunlight
Local color can be more bright-coloured, causes color feature extracted difficult；Second, when vehicle color is identical as license plate, for example for
Blue or yellow, green vehicle, can be by License Plate to full vehicle using color characteristic, and such detection effect needs further
Improve.
In conclusion problem of the existing technology is:
(1) prior art is different for shooting angle based on the textural characteristics of characters on license plate, distance is different, contains a large amount of lines
The complex scene image of article is managed, there is biggish limitations.
(2) the case where prior art is only applicable to image clearly, the scene of uniform illumination based on the color characteristic of license plate, it is right
Satisfactory effect often can not be reached in bad weather, uneven illumination.
Solve the difficulty of above-mentioned technical problem: first, to comprehensively consider different shooting angle, different illumination conditions with
And picture different clarity the case where, the prior art can only solve a certain specific condition mostly, be unable to satisfy requirement.The
Two, need to find other stronger features of more stable and robustness of license plate in the case where condition is limited, to overcome complexity
Scene change problem, to realize the accurate positionin of license plate.
Solve the meaning of above-mentioned technical problem: the prior art has the textural characteristics and readability of license plate picture very high
Requirement, to meet simultaneously more than condition need to put into a large amount of fund and construct complete license plate picture shooting system, it appears
And it is unrealistic.Therefore, the quality for improving algorithm is one of the best way to solve the above problems now.Moreover, the above problem
Solution can effectively increase the applicable scene of license plate locating method, it is not limited to which a certain situation, practicability is stronger, is applicable in
Property is wider.
Summary of the invention
In view of the problems of the existing technology, the present invention provides a kind of license plates combined based on MSER and ISODATA
Localization method.
The invention is realized in this way a kind of license plate locating method combined based on MSER and ISODATA, described to be based on
License plate locating method that MSER and ISODATA are combined the following steps are included:
Image to be detected is converted to gray level image by step 1, extracts the maximum stable extremal region in gray level image
MSER simultaneously fits minimum rectangle frame, is screened according to characters on license plate feature to rectangle frame, obtains candidate characters rectangle frame；
Step 2 is iterated self-organizing processing to the candidate characters rectangle frame upper left corner ordinate that step 1 obtains
ISODATA is after being clustered as a result, a sample set most containing rectangle frame after finding cluster, is determined as candidate license plate
Character rectangle frame；
Step 3 determines license plate up-and-down boundary according to the position of rectangle frame；
Step 4, by sliding window method, finds out the window most comprising rectangle frame according to the position of rectangle frame, determines left
Right margin.
Further, image to be detected is converted to gray level image by the step 1, and the maximum extracted in gray level image is steady
Determine extremal region MSER and fit minimum rectangle frame, rectangle frame is screened according to characters on license plate feature, obtains candidate word
According with rectangle frame includes following sub-step:
(1) image gray processing is handled；
(2) MSER extracts maximum stable extremal region and obtains its corresponding rectangle frame region；
(3) its basic statistics feature is screened according to characters on license plate feature, obtains candidate characters rectangle frame.
Further, gray processing processing meets formula:
Gray (m, n)=0.299 × R (m, n)+0.587 × G (m, n)+0.114 × B (m, n)；
Wherein (m, n) indicate m row n column location of pixels, R (m, n), G (m, n), B (m, n) represent original image it is red, it is green,
Pixel value of blue three channels at (m, n)；
Basic statistics feature includes minimum circumscribed rectangle the ratio of width to height, rectangle frame height.
Further, the candidate characters rectangle frame upper left corner ordinate obtained to step 1 is iterated self-organizing processing
ISODATA is after being clustered as a result, a sample set most containing rectangle frame after finding cluster, is determined as candidate license plate
Character rectangle frame the following steps are included:
(1) self-organizing processing ISODATA is iterated to the ordinate of obtained all candidate characters rectangle frames；
(2) sample set most containing sample size after iteration self-organizing is handled is found, candidate license plate word is determined as
Accord with rectangle frame.
Further, the position according to rectangle frame determine license plate up-and-down boundary the following steps are included:
(1) average value of the candidate license plate character rectangle frame upper left corner ordinate calculated and on being determined as according to it
Boundary；
(2) average value of the candidate license plate character rectangle frame lower right corner ordinate calculated and under being determined as according to it
Boundary；
(3) according to the draw value in the upper left corner and bottom right angular coordinate, the up-and-down boundary coordinate of license plate is determined:
Coboundary coordinate:
Lower boundary coordinate:
In formula,For the average value of rectangle frame upper left corner ordinate,For the average value of rectangle frame lower right corner ordinate.
Further, the position according to rectangle frame the window most comprising rectangle frame is found out, really by sliding window method
Determine right boundary the following steps are included:
(1) image comprising license plate is intercepted out from original image according to the ordinate of obtained up-and-down boundary；
(2) length and width of sliding window are determined according to obtained up-and-down boundary:
Sliding window height: wh=y2-y1；
Sliding window width: hh=| y2-y1|*3.85；
In formula, y1For the ordinate of license plate coboundary, y2For the ordinate of license plate lower boundary；
(3) sliding window operation is carried out to obtained interception image with obtained sliding window, and the rectangle frame in record window
Number；
(4) license plate area will be determined as comprising the most sliding window of rectangle frame, determines right boundary.
Another object of the present invention is to provide the License Plates combined described in a kind of application based on MSER and ISODATA
The License Plate Automatic Recognition System of method.
Another object of the present invention is to provide a kind of parking lots for being equipped with the License Plate Automatic Recognition System.
Another object of the present invention is to provide a kind of safety defense monitoring systems for being equipped with the License Plate Automatic Recognition System.
Another object of the present invention is to provide a kind of outpost vehicle detection systems for being equipped with the License Plate Automatic Recognition System
System.
In conclusion advantages of the present invention and good effect are as follows: the present invention extracts text rectangle frame by MSER algorithm,
And clustered rectangle frame ordinate by ISODATA algorithm, find the most region of sample in class (license plate area) progress
Positioning.License plate can be accurately located under different scenes, applied widely, strong robustness.
The present invention provides a kind of license plate locating methods combined based on MSER with ISODATA clustering algorithm.In MSER
After obtaining rectangle frame, the present invention no longer carries out the processing of Pixel-level to image, but according to rectangle frame information to license plate position
Judged, not only reduce the memory usage of CPU but also effectively reduces algorithm complexity.The present invention is suitable for complexity
Scene License Plate, poor illumination condition, atrocious weather condition and lower picture quality institute band can be overcome
The influence come, locating accuracy is high, strong robustness.
Detailed description of the invention
Fig. 1 is the license plate locating method flow chart provided in an embodiment of the present invention combined based on MSER and ISODATA.
Fig. 2 is the realization stream of the license plate locating method provided in an embodiment of the present invention combined based on MSER and ISODATA
Cheng Tu.
Fig. 3 is the original image of input required positioning provided in an embodiment of the present invention.
Fig. 4 is the MSER rectangle block diagram of extraction provided in an embodiment of the present invention.
Fig. 5 is the candidate characters rectangle block diagram after the Feature Selection provided in an embodiment of the present invention by basic statistics.
Fig. 6 is the candidate license plate character square provided in an embodiment of the present invention after iteration self-organizing handles (ISODATA)
Shape block diagram.
Fig. 7 is the license plate image after determining up-and-down boundary provided in an embodiment of the present invention.
Fig. 8 is the license plate image after determining right boundary provided in an embodiment of the present invention, i.e. License Plate result.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
In view of the problems of the existing technology, the present invention provides a kind of license plates combined based on MSER and ISODATA
Localization method is with reference to the accompanying drawing explained in detail the present invention.
As shown in Figure 1, the license plate locating method provided in an embodiment of the present invention combined based on MSER and ISODATA includes
Following steps:
S11: being converted to gray level image for image to be detected, extracts the maximum stable extremal region in gray level image
(MSER) and minimum rectangle frame is fitted, then rectangle frame is screened according to characters on license plate feature, obtain candidate characters square
Shape frame；
S102: self-organizing processing (ISODATA) is iterated to obtained candidate characters rectangle frame upper left corner ordinate, is obtained
It is after to cluster as a result, a sample set most containing rectangle frame after finding cluster, is determined as candidate license plate character rectangle frame；
S103: license plate up-and-down boundary is determined according to the position of rectangle frame；
S104: according to the position of rectangle frame, by sliding window method, the window most comprising rectangle frame is found out, determines left and right
Boundary.
Technical solution of the present invention is further described with reference to the accompanying drawing.
As shown in Fig. 2, the license plate locating method provided in an embodiment of the present invention combined based on MSER and ISODATA is specific
The following steps are included:
Step 1, MSER character candidates extracted region.
Fig. 3 be input want positioning licence plate original image
Original color image shown in Fig. 3 is converted gray level image by (1a), as follows:
Given RGB image Src (m, n), three of them color component are respectively R (m, n), G (m, n), B (m, n), wherein (m,
N) location of pixels of m row n column is indicated.Carry out following greyscale transformation:
Gray (m, n)=0.299 × R (m, n)+0.587 × G (m, n)+0.114 × B (m, n)；
(1b) MSER extracts maximum stable extremal region and obtains its corresponding rectangle frame region, extracts result such as Fig. 4 institute
Show.Specific step is as follows:
According to the size of the gray value of the gray level image after gray processing, all pixels in image are ranked up.
Seek the connected domain in image.Two-value is carried out to gray level image with 256 different threshold values in gray scale interval [0,255]
Change；Enable QiThe a certain connected region in the corresponding bianry image of binarization threshold i is indicated, when binarization threshold becomes i+ Δ by i
When with i- Δ, connected domain QiAlso Q has been becomei+ΔWith Qi-Δ。
Acquire MSER.Calculating threshold value is i area ratio q (i) (such as following formula), works as QiArea changes with binarization threshold i and is sent out
When raw small change, that is, q (i) is local minimum, QiFor maximum extreme value stability region:
Wherein, | Qi| indicate connected region QiArea, | Qi+Δ+Qi-Δ| indicate Qi+ΔSubtract Qi-ΔRemaining area face afterwards
Product；
(1c) screens rectangle frame according to characters on license plate feature, obtains candidate characters rectangle frame, the specific steps are as follows:
The rectangle frame extracted above is screened according to characters on license plate geometrical characteristic.It is main high by the width of rectangle frame
Than being screened with basic statistics features such as shape frame height degree, it is as follows that candidate connected domain meets condition:
Wherein, fRIndicate the ratio of width to height f of rectangle frameR=h/w.H is the height of rectangle frame, and w is the width of rectangle frame, screening
Afterwards shown in attached drawing 5；
ISODATA cluster, i.e. iteration self-organizing data analysis algorithm are that a kind of mode of non-supervisory Dynamic Clustering Algorithm is known
Other method.ISODATA analyzes the similarity of data itself using Euclidean distance, and core concept is that Euclidean distance is smaller, phase
It is bigger like spending, the high data of similarity are flocked together automatically.
Step 2, ISODATA algorithm determine candidate license plate character rectangle frame.
(2a) clusters the ordinate of all candidate characters rectangle frames with ISODATA algorithm to it；
The basic step of ISODATA clustering algorithm is as follows:
1) enabling clusters number is c0, maximum number of iterations T0, maximum kind internal standard difference is θs, min cluster central cluster
For θc, random initializtion clustering prototype enables the number of iterations t=0.
The random initializtion clustering prototype refers to: randomly selecting c0A ordinate value zj, j=1,2 ..., c0, by institute
There is rectangle frame to give cluster centre z according to minimal distance principle according to its upper left corner ordinatej, form cluster set Sj, wherein minimum
Distance principle refers to that the Euclidean distance of rectangle frame upper left corner ordinate is minimum.
2) cluster centre of following formula amendment clustering prototype is utilized:
In formula, zjFor jth class cluster set cluster centre, NjFor the number of rectangle frame in jth class Clustering Domain, SjIt birdss of the same feather flock together for jth
Class set, x are cluster set SjInterior rectangle frame upper left corner ordinate value, c0For clusters number.
3) the following parameter of clustering prototype after correcting cluster centre is calculated:
Average distance in total class:
Average distance in class:
Class internal standard is poor:
In formula, j=1,2 ..., c0, N is the number of all rectangle frames, NjFor the number of the rectangle frame in jth class cluster set
Mesh, SjFor jth class cluster set, x is cluster set SjInterior rectangle frame upper left corner ordinate value, zjIn cluster for jth class Clustering Domain
The heart, c0For clusters number；
4) by class internal standard difference djWith maximum kind internal standard difference θsIt is compared, if dj> θsIt thens follow the steps 5), otherwise
Jump procedure 6)；
5) to SjCarry out splitting operation；
5.1) to SjCarry out splitting operation；
By SjAveragely split into positive cluster set Sj +With negative cluster set Sj -, calculate positive cluster set Sj +Cluster centre zj +With it is negative
Cluster set zj -Cluster centre:
Positive cluster set: zj +=zj-|zj+1-zj-1|/6；
Negative cluster set: zj -=zj+|zj+1-zj-1|/6；
In formula, j=1,2 ... c0- 1, zj-1It is the cluster centre of -1 class cluster set of jth, zjIt is the cluster of jth class cluster set
Center, zj+1It is the cluster centre of+1 class cluster set of jth, c0For clusters number；
5.2) clusters number: c is updated0=c0+ 1, return step 2)；
6) j-th of cluster set S in clustering prototype is calculatedjWith+1 cluster set S of jthj+1Cluster centre distance Dj,j+1:
Dj,j+1=| zj+1-zj|, j=1,2 ... c0- 1, in formula, zjIt is the cluster centre of jth class cluster set, zj+1It is+1 class of jth cluster
The cluster centre of collection；
7) the distance D for the cluster centre for obtaining step 6)j,j+1With min cluster centre distance θcIt compares, if Dj,j+1
<θcIt thens follow the steps 8), otherwise, jump procedure 9)；
8) to jth class cluster set SjWith+1 class cluster set S of jthj+1Merge operation:
8.1) by jth class cluster set SjWith+1 class cluster set S of jthj+1It is merged into jth j class cluster set Sjj, calculate SjjIt is poly-
Class center zjj: zjj=(zj+1+zj)/2, j=1,2 ... c0-1；In formula, zjIt is the cluster centre of jth class cluster set, zj+1It is
The cluster centre of j+1 class cluster set, c0For clusters number；
8.2) clusters number: c is updated0=c0- 1, return step 2)；The number of iterations t is compared with maximum number of iterations,
If t=T0, export optimum cluster number c0=c0, iterative operation is terminated, t=t+1, return step 2 are otherwise enabled)；
(2b) is focused to find out a sample set most comprising rectangle frame in all samples, and the rectangle in the sample set
Frame is set as candidate license plate character rectangle frame, as shown in Figure 6.
Step 3 determines license plate up-and-down boundary:
(3a) calculates the average value of candidate license plate character rectangle frame upper left corner ordinate obtained in (2b).
Upper left corner ordinate draw value:
In formula, j=1,2 ..., c0, N is the number of all rectangle frames, NjFor the number of the rectangle frame in jth class cluster set
Mesh, SjFor jth class cluster set, x1For cluster set SjInterior rectangle frame upper left corner ordinate value；
(3b) calculates the average value of candidate license plate character rectangle frame lower right corner ordinate obtained in (2b).
Lower right corner ordinate draw value:
In formula, j=1,2 ..., c0, N is the number of all rectangle frames, NjFor the number of the rectangle frame in jth class cluster set
Mesh, SjFor jth class cluster set, x2For cluster set SjInterior rectangle frame lower right corner ordinate value；
(3c) determines the longitudinal coordinate of license plate according to the draw value in the upper left corner and bottom right angular coordinate, and specific practice is as follows:
Coboundary coordinate:
Lower boundary coordinate:
In formula,For the average value of rectangle frame upper left corner ordinate,For the average value of rectangle frame lower right corner ordinate,
Image after up-and-down boundary determines is as shown in Figure 7.
Step 4 determines license plate right boundary:
The ordinate of (4a) up-and-down boundary according to obtained in step 3 intercepts out the image comprising license plate from original image；
(4b) up-and-down boundary according to obtained in step 3 determines the length and width of sliding window, the specific steps are as follows:
Sliding window height: wh=y2-y1；
Sliding window width: hh=| y2-y1|*4.85；
In formula, y1For the ordinate of license plate coboundary, y2For the ordinate of license plate lower boundary；
It is 2 that the sliding window that (4c) is obtained according to (4b), which carries out stride from left to right to the interception image that (4a) is obtained,
Sliding window operation, and the number of the rectangle frame in record window, the specific steps are as follows:
Calculate the abscissa of the central point of all rectangle frames:
The abscissa of central point: xc=x+width/2；
Wherein, x is rectangle frame upper left corner abscissa value, and width is the width of rectangle frame.
If the sliding window of jth time is Wj, traverse all rectangle frames and find the rectangle frame in sliding window.If a certain rectangle frame
In sliding window, then meet following relationship: xL≤xc≤xR；
X in formulaLFor the abscissa value of sliding window left side, xRFor the abscissa value of sliding window right edge, xcFor the center of rectangle frame
The abscissa value of point；
(4d) finds the sliding window most comprising rectangle frame number, using the abscissa of the side on the left of the window as a license plate left side
The abscissa on boundary, using the abscissa of the side on the right side of the window as the abscissa of license plate right margin, after right boundary determines
Image as shown in figure 8, this result is also final positioning result.
Technical effect of the invention is explained in detail below with reference to emulation.
1. simulated conditions:
The present invention be central processing unit be Intel (R) Corei7-77003.60GHZ, memory 16G, WINDOWS10 grasp
Make in system, is emulated with 2015 software of VS.
2. emulation content:
In order to compare the superiority of the present invention with other license plate locating methods, herein respectively with the positioning based on edge detection
Method with based on color localization method and method provided herein test 400 Zhang great little be 2048*1568 bayonet image,
The bayonet data of different illumination conditions, different weather conditions are contained in this 400 images, test result, which has, persuades
Power.
3. the simulation experiment result and analysis:
13 kinds of location algorithm test results of table
Table 1 be 400 containing different illumination conditions, different weather conditions complicated bayonet image in test three kinds not
The test result of same localization method, the License Plate accuracy rate of the invention in complex scene is wanted as can be seen from the test results
Far better than the License Plate accuracy rate of other two methods.Since the resolution ratio of bayonet image is generally higher, inventive algorithm
It handles the time and above two method only has slightly advantage.In summary, the present invention can overcome poor illumination condition and
Atrocious weather condition reaches the accurate positionin of license plate, strong robustness.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (10)
1. a kind of license plate locating method combined based on MSER and ISODATA, which is characterized in that it is described based on MSER and
License plate locating method that ISODATA is combined the following steps are included:
Image to be detected is converted to gray level image by step 1, extracts the maximum stable extremal region MSER in gray level image
And fit minimum rectangle frame；Rectangle frame is screened according to characters on license plate feature, obtains candidate characters rectangle frame；
Step 2 is iterated self-organizing processing ISODATA to the candidate characters rectangle frame upper left corner ordinate that step 1 obtains；
It is after being clustered as a result, a sample set most containing rectangle frame after finding cluster, is determined as candidate license plate character rectangle
Frame；
Step 3 determines license plate up-and-down boundary according to the position of rectangle frame；
Step 4, by sliding window method, finds out the window most comprising rectangle frame, determines left and right side according to the position of rectangle frame
Boundary.
2. the license plate locating method combined as described in claim 1 based on MSER and ISODATA, which is characterized in that described
Image to be detected is converted to gray level image by step 1, is extracted the maximum stable extremal region MSER in gray level image and is fitted
Minimum rectangle frame out screens rectangle frame according to characters on license plate feature, and obtaining candidate characters rectangle frame includes following sub-step
It is rapid:
(1) image gray processing is handled；
(2) MSER extracts maximum stable extremal region and obtains its corresponding rectangle frame region；
(3) its basic statistics feature is screened according to characters on license plate feature, obtains candidate characters rectangle frame.
3. the license plate locating method combined as claimed in claim 2 based on MSER and ISODATA, it is characterised in that: gray scale
Change processing meets formula:
Gray (m, n)=0.299 × R (m, n)+0.587 × G (m, n)+0.114 × B (m, n)；
Wherein (m, n) indicates that the location of pixels of m row n column, R (m, n), G (m, n), B (m, n) represent original image red, green, blue three
Pixel value of a channel at (m, n)；
Basic statistics feature includes minimum circumscribed rectangle the ratio of width to height, rectangle frame height.
4. the license plate locating method combined as described in claim 1 based on MSER and ISODATA, which is characterized in that described
Self-organizing processing ISODATA is iterated to the candidate characters rectangle frame upper left corner ordinate that step 1 obtains, after obtaining cluster
As a result, a sample set most containing rectangle frame after finding cluster, it includes following for being determined as candidate license plate character rectangle frame
Step:
(1) self-organizing processing ISODATA is iterated to the ordinate of obtained all candidate characters rectangle frames；
(2) sample set most containing sample size after iteration self-organizing is handled is found, candidate license plate character square is determined as
Shape frame.
5. the license plate locating method combined as described in claim 1 based on MSER and ISODATA, which is characterized in that described
Determine license plate up-and-down boundary according to the position of rectangle frame the following steps are included:
(1) average value of the candidate license plate character rectangle frame upper left corner ordinate calculated is simultaneously determined as top according to it
Boundary；
(2) average value of the candidate license plate character rectangle frame lower right corner ordinate calculated is simultaneously determined as below according to it
Boundary；
(3) according to the draw value in the upper left corner and bottom right angular coordinate, the up-and-down boundary coordinate of license plate is determined:
Coboundary coordinate:
Lower boundary coordinate:
In formula,For the average value of rectangle frame upper left corner ordinate,For the average value of rectangle frame lower right corner ordinate.
6. the license plate locating method combined as described in claim 1 based on MSER and ISODATA, which is characterized in that described
According to the position of rectangle frame, by sliding window method, the window most comprising rectangle frame is found out, determines that right boundary includes following step
It is rapid:
(1) image comprising license plate is intercepted out from original image according to the ordinate of obtained up-and-down boundary；
(2) length and width of sliding window are determined according to obtained up-and-down boundary:
Sliding window height: wh=y2-y1；
Sliding window width: hh=| y2-y1|*3.85；
In formula, y1For the ordinate of license plate coboundary, y2For the ordinate of license plate lower boundary；
(3) sliding window operation is carried out to obtained interception image with obtained sliding window, and the number of the rectangle frame in record window
Mesh；
(4) license plate area will be determined as comprising the most sliding window of rectangle frame, determines right boundary.
7. it is a kind of using described in claim 1~6 any one based on MSER and the ISODATA license plate locating method combined
License Plate Automatic Recognition System.
8. a kind of parking lot for being equipped with License Plate Automatic Recognition System described in claim 7.
9. a kind of safety defense monitoring system for being equipped with License Plate Automatic Recognition System described in claim 8.
10. a kind of outpost vehicle detecting system for being equipped with License Plate Automatic Recognition System described in claim 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910618457.1A CN110516666B (en) | 2019-07-10 | 2019-07-10 | License plate positioning method based on combination of MSER and ISODATA |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910618457.1A CN110516666B (en) | 2019-07-10 | 2019-07-10 | License plate positioning method based on combination of MSER and ISODATA |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110516666A true CN110516666A (en) | 2019-11-29 |
| CN110516666B CN110516666B (en) | 2023-01-17 |
Family
ID=68622598
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910618457.1A Active CN110516666B (en) | 2019-07-10 | 2019-07-10 | License plate positioning method based on combination of MSER and ISODATA |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110516666B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111027544A (en) * | 2019-11-29 | 2020-04-17 | 武汉虹信技术服务有限责任公司 | MSER license plate positioning method and system based on visual saliency detection |
| CN111523562A (en) * | 2020-03-20 | 2020-08-11 | 浙江大学 | Commuting mode vehicle identification method based on license plate identification data |
| CN111553251A (en) * | 2020-04-24 | 2020-08-18 | 平安科技（深圳）有限公司 | Certificate four-corner incomplete detection method, device, equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105373794A (en) * | 2015-12-14 | 2016-03-02 | 河北工业大学 | Vehicle license plate recognition method |
| CN106815583A (en) * | 2017-01-16 | 2017-06-09 | 上海理工大学 | A kind of vehicle at night license plate locating method being combined based on MSER and SWT |
| WO2018072233A1 (en) * | 2016-10-20 | 2018-04-26 | 中山大学 | Method and system for vehicle tag detection and recognition based on selective search algorithm |
-
2019
- 2019-07-10 CN CN201910618457.1A patent/CN110516666B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105373794A (en) * | 2015-12-14 | 2016-03-02 | 河北工业大学 | Vehicle license plate recognition method |
| WO2018072233A1 (en) * | 2016-10-20 | 2018-04-26 | 中山大学 | Method and system for vehicle tag detection and recognition based on selective search algorithm |
| CN106815583A (en) * | 2017-01-16 | 2017-06-09 | 上海理工大学 | A kind of vehicle at night license plate locating method being combined based on MSER and SWT |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111027544A (en) * | 2019-11-29 | 2020-04-17 | 武汉虹信技术服务有限责任公司 | MSER license plate positioning method and system based on visual saliency detection |
| CN111027544B (en) * | 2019-11-29 | 2023-09-29 | 武汉虹信技术服务有限责任公司 | MSER license plate positioning method and system based on visual saliency detection |
| CN111523562A (en) * | 2020-03-20 | 2020-08-11 | 浙江大学 | Commuting mode vehicle identification method based on license plate identification data |
| CN111523562B (en) * | 2020-03-20 | 2021-06-08 | 浙江大学 | Commuting mode vehicle identification method based on license plate identification data |
| CN111553251A (en) * | 2020-04-24 | 2020-08-18 | 平安科技（深圳）有限公司 | Certificate four-corner incomplete detection method, device, equipment and storage medium |
| CN111553251B (en) * | 2020-04-24 | 2024-05-07 | 平安科技（深圳）有限公司 | Certificate four-corner defect detection method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110516666B (en) | 2023-01-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wei et al. | Multi-vehicle detection algorithm through combining Harr and HOG features | |
| CN108108761B (en) | Rapid traffic signal lamp detection method based on deep feature learning | |
| CN109740478B (en) | Vehicle detection and identification method, device, computer equipment and readable storage medium | |
| CN109101924B (en) | Machine learning-based road traffic sign identification method | |
| CN109840521B (en) | Integrated license plate recognition method based on deep learning | |
| KR101403876B1 (en) | Method and Apparatus for Vehicle License Plate Recognition | |
| CN106650553A (en) | License plate recognition method and system | |
| CN102799859B (en) | Method for identifying traffic sign | |
| CN102663760B (en) | Location and segmentation method for windshield area of vehicle in images | |
| CN108921120B (en) | Cigarette identification method suitable for wide retail scene | |
| Zhang et al. | Study on traffic sign recognition by optimized Lenet-5 algorithm | |
| CN106529532A (en) | License plate identification system based on integral feature channels and gray projection | |
| CN105160691A (en) | Color histogram based vehicle body color identification method | |
| CN104050684B (en) | A kind of video frequency motion target sorting technique based on on-line training and system | |
| CN110516666A (en) | The license plate locating method combined based on MSER and ISODATA | |
| CN104766344B (en) | Vehicle checking method based on movement edge extractor | |
| CN107590500A (en) | A kind of color recognizing for vehicle id method and device based on color projection classification | |
| CN111915583A (en) | Vehicle and pedestrian detection method based on vehicle-mounted thermal infrared imager in complex scene | |
| CN108664969A (en) | Landmark identification method based on condition random field | |
| CN113221881A (en) | Multi-level smart phone screen defect detection method | |
| Liu et al. | A robust algorithm for text detection in color images | |
| CN110689003A (en) | Low-illumination imaging license plate recognition method and system, computer equipment and storage medium | |
| Chen et al. | License plate recognition for moving vehicles using a moving camera | |
| Shi et al. | License plate localization in complex environments based on improved GrabCut algorithm | |
| CN113344047A (en) | Platen state identification method based on improved K-means algorithm |
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 |