CN109190621A - Greasy weather automatic license plate recognition method - Google Patents
Greasy weather automatic license plate recognition method Download PDFInfo
- Publication number
- CN109190621A CN109190621A CN201811039768.4A CN201811039768A CN109190621A CN 109190621 A CN109190621 A CN 109190621A CN 201811039768 A CN201811039768 A CN 201811039768A CN 109190621 A CN109190621 A CN 109190621A
- Authority
- CN
- China
- Prior art keywords
- image
- license plate
- carried out
- follows
- value
- 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
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000012545 processing Methods 0.000 claims abstract description 27
- 230000009466 transformation Effects 0.000 claims abstract description 16
- 229910052704 radon Inorganic materials 0.000 claims abstract description 14
- SYUHGPGVQRZVTB-UHFFFAOYSA-N radon atom Chemical compound [Rn] SYUHGPGVQRZVTB-UHFFFAOYSA-N 0.000 claims abstract description 14
- 230000000877 morphologic effect Effects 0.000 claims abstract description 10
- 230000008569 process Effects 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims abstract description 9
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000006243 chemical reaction Methods 0.000 claims abstract description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 24
- 239000004615 ingredient Substances 0.000 claims description 14
- 239000003595 mist Substances 0.000 claims description 11
- 230000001186 cumulative effect Effects 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 238000005260 corrosion Methods 0.000 claims description 3
- 230000007797 corrosion Effects 0.000 claims description 3
- 238000009795 derivation Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 claims description 3
- 230000003628 erosive effect Effects 0.000 claims description 3
- 238000005530 etching Methods 0.000 claims description 3
- 238000003702 image correction Methods 0.000 claims description 3
- 238000013178 mathematical model Methods 0.000 claims description 3
- 238000011084 recovery Methods 0.000 claims 1
- 230000006870 function Effects 0.000 description 10
- 230000000694 effects Effects 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 230000004807 localization Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000032258 transport Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
-
- 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
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/243—Aligning, centring, orientation detection or correction of the image by compensating for image skew or non-uniform image deformations
-
- 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
- G06V10/30—Noise filtering
-
- 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
- G06V30/153—Segmentation of character regions using recognition of characters or words
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20068—Projection on vertical or horizontal image axis
-
- 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)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of greasy weather automatic license plate recognition methods, have following steps: S1, image defogging；S2, image preprocessing: gray processing processing is carried out to the enhanced image that step S1 is obtained, median filter process is carried out to the gray level image that gray processing is handled, canny edge extracting is carried out to the image that median filter process obtains, obtains the marginal information of image；S3, License Plate: Morphological scale-space is carried out to the image after edge extracting；The boundary of license plate is determined by sciagraphy；S4, binary conversion treatment is carried out to the license plate image on determining license plate boundary；S5, rectification is carried out to license plate using Radon transformation；S6, Car license recognition: the identification of characters on license plate is carried out using Tesseract character recognition engine and jTessBoxEditor training tool.Very good solution of the present invention discrimination speed under automatic license plate identification system under haze weather is slow, discrimination low problem.
Description
Technical field
The present invention relates to machine vision intelligent transportation field application, especially for the image in the greasy weather, taken
A kind of greasy weather automatic license plate recognition method under the conditions of ropy.
Background technique
Electronic monitoring and control system has been obtained in field of traffic be widely applied at present, however electronic monitoring and control system is to weather
Condition is extremely sensitive.Mist is a kind of very common weather phenomenon, since, there are many suspended particulates, low visibility makes in air
It is fuzzy to obtain camera installation acquired image, contrast is low, affects the quality of license auto-recognition system in traffic administration.Therefore it is
It realizes all weather operations of vision system, robustness and reliability of the system in the greasy weather is improved, so that the license plate in greasy weather is automatic
Recognition methods becomes a key component of intelligent transportation field research.
Current Car license recognition mainly has indirect method and direct method two ways.Indirect method, which refers to, is mounted on vapour by identification
The license board information that stores in IC card or bar code on vehicle reaches the identification to license plate.The recognition accuracy of IC card technology is high, transports
Row stablizes, can be with all weather operations, but equipment is expensive, and maintenance cost height is not suitable for strange land operation.Bar codes technique has
The advantages that recognition speed is fast, accuracy is high, highly reliable and at low cost, but it is more demanding to scanner, and vehicle can not be checked,
Whether bar code is consistent, this causes difficulty to popularization in a short time.
Direct method is the license plate recognition technology based on image.It mainly include that Image Acquisition is carried out to movement or static automobile
It is identified with to collected vehicle static image.This method saves equipment placement and substantial contribution, is suitable for strange land
Operation.This method mainly includes Image Acquisition, and image preprocessing, License Plate, Character segmentation, character recognition etc. is main to be walked
Suddenly.But this method is sensitive to weather condition, such as rain and snow weather, the case where dust and sand weather, the low visibility such as haze weather
It will affect the accuracy rate of Car license recognition.Therefore the real-time and accuracy of Car license recognition under severe weather conditions are improved and is known as vehicle
An important link in board identification.
Summary of the invention
In order to overcome the shortcomings of in above-mentioned technology, for the foggy weather in bad weather, the present invention proposes a kind of greasy weather
Automatic license plate recognition method is combined quick based on dark primary prior image fogging method and Tesseract character recognition engine
Licence plate recognition method.The technological means that the present invention uses is as follows:
A kind of greasy weather automatic license plate recognition method has following steps:
S1, image defogging:
Dark primary priori fogging method carries out defogging processing to the greasy weather automobile image taken, and successively carries out image increasing
By force, color image restoration, image correction and image enhancement obtain enhanced image；
S2, image preprocessing:
Gray processing processing, the gray level image handled gray processing are carried out to the enhanced image that step S1 is obtained
Median filter process (reducing the noise in image) is carried out, canny edge extracting is carried out to the image that median filter process obtains,
Obtain the marginal information of image；
S3, License Plate:
S31, Morphological scale-space is carried out to the image after edge extracting；
S32, the boundary that license plate is determined by sciagraphy；
S4, binary conversion treatment is carried out (using improved localization Binarization methods to the license plate image on determining license plate boundary
Bernsen algorithm)；
S5, rectification is carried out to license plate using Radon transformation；
S6, Car license recognition:
The identification of characters on license plate is carried out using Tesseract character recognition engine and jTessBoxEditor training tool.
The specific steps of the step S1 are as follows:
S11, the color lump that the greasy weather automobile image of input is divided into 15*15 seek part and global dark primary figure；
S12, assume atmosphere light ingredient it is known that estimate atmospheric transmissivity by dark primary priori theoretical, the mist figure used
At model are as follows:
I (y)=J (y) t (x)+A (1-t (x))
Wherein t (x) is atmospheric transmissivity, and I is the foggy image that camera takes, and J is image after defogging, and A is atmosphere
Light ingredient；
Pass through the calculation formula of dark primary priori theoretical and the available atmospheric transmissivity discreet value of mathematical derivation are as follows:
Wherein Ω (x) is the regional area centered on x, and ω is a constant, takes 0.95, Ic(y) it is shot for camera
One in a certain Color Channel of R, G, B of the foggy image I (y) arrived, AcIt is logical for a certain color of R, G, B of atmosphere light ingredient A
One in road；
S13, dark primary prior estimate atmosphere light ingredient is utilized:
Preceding 0.1% pixel is taken to take then in these positions in camera by brightness size from dark channel diagram
Foggy image I in find estimated value of the value of the corresponding point with maximum brightness as atmosphere light ingredient A；
S14, fog free images are restored:
By known I, A, t substitute into mist figure at mathematical model in, can be in the hope of:
Wherein t0It is the lower limit value of transmissivity setting, as t < t0When, enable t=t0, (avoiding the occurrence of the excessively bright situation of J leads to figure
Image distortion) take t0=0.1；
S15, image enhancement and color image restoration are carried out to fog free images are restored using multi-Scale Retinex Algorithm, it is more
Dimension Retinex algorithm is as follows:
Wherein RMSR(x, y) represents the output of multiple dimensioned Retinex, which Color Channel i represents, due to being RGB figure
Picture, therefore N=3, WkExpression and FkRelevant weight coefficient, Ii(x, y) is i-th of the Color Channel or i-th of wave of input picture
Section, * is convolution algorithm, and K represents the number of scaling function, and as K=1, multiple dimensioned MSR is converted into single scale SSR, Fk(x,y)
It indicates k-th surround function, takes Gauss surround function, and meet ∫ ∫ Fk(x, y)=dxdy=1；
S16, image J after defogging is decomposed into tri- width gray level image of R, G, B, and successively by each pixel in this three width grayscale image
The data type of point gray value becomes double type, is then once operated to three width images respectively:
Determine Gaussian environment function formula are as follows:
Choose three different standard deviation sigmaskRespectively 13,73,130, and according to ∫ ∫ Fk(x, y)=dxdy=1 finds out Ck；
Using MSR method, treated that image pixel will appear negative value, and will appear beyond display indication range
Therefore situation is modified image using gain/offset method, then the gray value of revised image is projected to ash
It spends in range (0~255), specific algorithm is as follows:
Rxz(x, y)=G × RMSR(x,y)+offset
Wherein RMSR、Rxz、RoThe respectively gray value of input picture, the gray value after correcting, the ash exported after projection
Angle value, gain coefficient G and offset take 3 and 50, r respectivelymaxAnd rminRespectively RxzMinimum and maximum gray value；
S17, image R, image G and image B are carried out to the above operation, available three width image carries out enhanced figure
Then picture is merged image R, image G and image B by multi-Scale Retinex Algorithm, obtain that camera takes has mist figure
As the enhanced image of I.
Specific step is as follows by the step S31:
Etching operation (for eliminating the uncorrelated noise in image) is carried out to the image after edge extracting, erosion algorithm expression
Formula is as follows:
Wherein B is structural element, and A` is the image after edge extracting；
Expansive working (for eliminating target area noise) and filling cavity, expansion algorithm table are carried out to the image after corrosion
It is as follows up to formula:
Specific step is as follows by the step S32:
Determine the up-and-down boundary of license plate, specific steps are as follows:
A1, the projection that horizontal direction is done to the image after Morphological scale-space, and set the threshold value of license plate pixel as T, and take the T to be
50；
A2, the place of pixel highest point is scanned downwards line by line from projection, when the quantity of scanning to white pixel is greater than threshold
When value T, then it is assumed that be license plate coboundary, and be recorded as LPT；
A3, continue scanning downwards, when scanning to white pixel quantity is less than threshold value T, then it is assumed that it is license plate lower boundary, and
It is recorded as LPB；
License plate up-and-down boundary is such as not detected, then enables T=T-3, and repeat step A1-A3；
Determine the right boundary of license plate, specific steps are as follows:
B1, the projection that vertical direction is done to the image after Morphological scale-space, and license plate pixel threshold is set as T1；
B2, the most left place of pixel is scanned to the right from projection, when scanning to white pixel point number is greater than threshold value T1,
Then it is considered license plate left margin, and is recorded as LPL；
B3, continuation scan to the right, when scanning to white pixel number is less than threshold value T1Shi Ze is considered license plate right margin, and remembers
Record is LPR；
License plate right boundary is such as not detected, then enables T1=T1- 1, and repeat step B1-B3.
Specific step is as follows by the step S4:
S41, the license plate image progress gray processing processing to license plate boundary is determined；
Threshold value T after the gray processing processing that S42, calculating step S41 are obtained at the pixel f (x, y) of license plate1` (x, y),
Wherein window value ω selection 10, parameter lambda selection 0.48:
S43, the median filtering that template W is 3 × 3 is carried out to pixel f (x, y):
G (x, y)=med { f (x-k, y-l), k, l ∈ W }
S44, the threshold value T for calculating G (x, y) point after gaussian filtering2` (x, y):
S45, the binaryzation point-by-point to f (x, y), F (i, j) indicate final binarization result, proportionality coefficient α=0.1:
In the image that actually photographed, license plate both horizontally and vertically can all have a different degrees of inclination, but because
Inclination for license plate in vertical angle is not obvious, and has no effect on last identification process, therefore Radon is used only and converts to vehicle
Board carries out rectification.
To any angle β, the formula of the Radon transformation of function f (x, y) are as follows:
Wherein:
The specific steps of step S5 are as follows:
S51, carried out in [- 20 °, 20 °] section Radon transformation (by license plate known to priori knowledge inclination angle always [-
20 °, 20 °] between)；
S52, ask Radon transformation results in step S51 the absolute value of first derivative cumulative respectively and, have maximum cumulative
The angle of sum is tilt angle.
S53, the level correction of vehicle license plate, rotation transformation are realized to license plate rotation β angle are as follows:
Under automatic license plate identification system under haze weather discrimination speed is slow for very good solution of the present invention, and discrimination is low
The problem of.
The present invention can be widely popularized in fields such as intelligent transportation based on the above reasons.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to do simply to introduce, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without any creative labor, can be with
It obtains other drawings based on these drawings.
Fig. 1 is the flow chart of greasy weather automatic license plate recognition method in a specific embodiment of the invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
As shown in Figure 1, a kind of greasy weather automatic license plate recognition method, it is characterised in that have following steps:
S1, image defogging:
Dark primary priori fogging method carries out defogging processing to the greasy weather automobile image taken, and successively carries out image increasing
By force, color image restoration, image correction and image enhancement obtain enhanced image；
S2, image preprocessing:
Gray processing processing, the gray level image handled gray processing are carried out to the enhanced image that step S1 is obtained
Median filter process is carried out, canny edge extracting is carried out to the image that median filter process obtains, obtains the edge letter of image
Breath；
S3, License Plate:
S31, Morphological scale-space is carried out to the image after edge extracting；
S32, the boundary that license plate is determined by sciagraphy；
S4, binary conversion treatment is carried out (using improved localization Binarization methods to the license plate image on determining license plate boundary
Bernsen algorithm)；
S5, rectification is carried out to license plate using Radon transformation；
S6, Car license recognition:
The identification of characters on license plate is carried out using Tesseract character recognition engine and jTessBoxEditor training tool.
The specific steps of the step S1 are as follows:
S11, the color lump that the greasy weather automobile image of input is divided into 15*15 seek part and global dark primary figure；
S12, assume atmosphere light ingredient it is known that estimate atmospheric transmissivity by dark primary priori theoretical, the mist figure used
At model are as follows:
I (y)=J (y) t (x)+A (1-t (x))
Wherein t (x) is atmospheric transmissivity, and I is the foggy image that camera takes, and J is image after defogging, and A is atmosphere
Light ingredient；
Pass through the calculation formula of dark primary priori theoretical and the available atmospheric transmissivity discreet value of mathematical derivation are as follows:
Wherein Ω (x) is the regional area centered on x, and ω is a constant, takes 0.95, Ic(y) it is shot for camera
One in a certain Color Channel of R, G, B of the foggy image I (y) arrived, AcIt is logical for a certain color of R, G, B of atmosphere light ingredient A
One in road；
S13, dark primary prior estimate atmosphere light ingredient is utilized:
Preceding 0.1% pixel is taken to take then in these positions in camera by brightness size from dark channel diagram
Foggy image I in find estimated value of the value of the corresponding point with maximum brightness as atmosphere light ingredient A；
S14, fog free images are restored:
By known I, A, t substitute into mist figure at mathematical model in, can be in the hope of:
Wherein t0It is the lower limit value of transmissivity setting, as t < t0When, enable t=t0, take t0=0.1；
S15, image enhancement and color image restoration are carried out to fog free images are restored using multi-Scale Retinex Algorithm, it is more
Dimension Retinex algorithm is as follows:
Wherein RMSR(x, y) represents the output of multiple dimensioned Retinex, which Color Channel i represents, due to being RGB figure
Picture, therefore N=3, WkExpression and FkRelevant weight coefficient, Ii(x, y) is i-th of the Color Channel or i-th of wave of input picture
Section, * is convolution algorithm, and K represents the number of scaling function, and as K=1, multiple dimensioned MSR is converted into single scale SSR, Fk(x,y)
It indicates k-th surround function, takes Gauss surround function, and meet ∫ ∫ Fk(x, y)=dxdy=1；
S16, image J after defogging is decomposed into tri- width gray level image of R, G, B, and successively by each pixel in this three width grayscale image
The data type of point gray value becomes double type, is then once operated to three width images respectively:
Determine Gaussian environment function formula are as follows:
Choose three different standard deviation sigmaskRespectively 13,73,130, and according to ∫ ∫ Fk(x, y)=dxdy=1 finds out Ck；
Image is modified using gain/offset method, then the gray value of revised image is projected into gray scale
In range (0~255), specific algorithm is as follows:
Rxz(x, y)=G × RMSR(x,y)+offset
Wherein RMSR、Rxz、RoThe respectively gray value of input picture, the gray value after correcting, the ash exported after projection
Angle value, gain coefficient G and offset take 3 and 50, r respectivelymaxAnd rminRespectively RxzMinimum and maximum gray value；
S17, image R, image G and image B are carried out to the above operation, available three width image carries out enhanced figure
Then picture is merged image R, image G and image B by multi-Scale Retinex Algorithm, obtain that camera takes has mist figure
As the enhanced image of I.
Specific step is as follows by the step S31:
Etching operation is carried out to the image after edge extracting, erosion algorithm expression formula is as follows:
Wherein B is structural element, and A` is the image after edge extracting；
Expansive working is carried out to the image after corrosion and filling cavity, expansion algorithm expression formula are as follows:
Specific step is as follows by the step S32:
Determine the up-and-down boundary of license plate, specific steps are as follows:
A1, the projection that horizontal direction is done to the image after Morphological scale-space, and set the threshold value of license plate pixel as T, and take the T to be
50；
A2, the place of pixel highest point is scanned downwards line by line from projection, when the quantity of scanning to white pixel is greater than threshold
When value T, then it is assumed that be license plate coboundary, and be recorded as LPT；
A3, continue scanning downwards, when scanning to white pixel quantity is less than threshold value T, then it is assumed that it is license plate lower boundary, and
It is recorded as LPB；
License plate up-and-down boundary is such as not detected, then enables T=T-3, and repeat step A1-A3；
Determine the right boundary of license plate, specific steps are as follows:
B1, the projection that vertical direction is done to the image after Morphological scale-space, and license plate pixel threshold is set as T1；
B2, the most left place of pixel is scanned to the right from projection, when scanning to white pixel point number is greater than threshold value T1,
Then it is considered license plate left margin, and is recorded as LPL；
B3, continuation scan to the right, when scanning to white pixel number is less than threshold value T1Shi Ze is considered license plate right margin, and remembers
Record is LPR；
License plate right boundary is such as not detected, then enables T1=T1- 1, and repeat step B1-B3.
Specific step is as follows by the step S4:
S41, the license plate image progress gray processing processing to license plate boundary is determined；
Threshold value T after the gray processing processing that S42, calculating step S41 are obtained at the pixel f (x, y) of license plate1` (x, y),
Wherein window value ω selection 10, parameter lambda selection 0.48:
S43, the median filtering that template W is 3 × 3 is carried out to pixel f (x, y):
G (x, y)=med { f (x-k, y-l), k, l ∈ W }
S44, the threshold value T for calculating G (x, y) point after gaussian filtering2` (x, y):
S45, the binaryzation point-by-point to f (x, y), F (i, j) indicate final binarization result, proportionality coefficient α=0.1:
To any angle β, the formula of the Radon transformation of function f (x, y) are as follows:
Wherein:
The specific steps of step S5 are as follows:
S51, Radon transformation is carried out in [- 20 °, 20 °] section；
S52, ask Radon transformation results in step S51 the absolute value of first derivative cumulative respectively and, have maximum cumulative
The angle of sum is tilt angle.
S53, the level correction of vehicle license plate, rotation transformation are realized to license plate rotation β angle are as follows:
Specific step is as follows by step S6:
S61, download Tesseract and attach it to C: Program Files Tesseract-OCR.Tesseract
Exploitation relies on leptonica Runtime Library, it is the image image procossing library write by C language, and downloading leptonica simultaneously will
It is unziped under Tesseract installation directory, then again in the environmental variance of computer addition C: Program Files
Tesseract-OCR lib, then the exploitation environment of Visual Studio 2015 is configured, VC is found in property pages
++ catalogue, it is following comprising catalogue under add following catalogue:
D:\Program Files\Tesseract-OCR\tesseract-ocr\api；
D:\Program Files\Tesseract-OCR\tesseract-ocr\ccutil；
D:\Program Files\Tesseract-OCR\tesseract-ocr\ccstruct；
D:\Program Files\Tesseract-OCR\tesseract-ocr\ccmain；
And added under library directory: C: Program Files Tesseract-OCR lib；
S62, in order to improve Car license recognition speed, using aim at OCR software design training tool jTessBoxEditor,
It is developed by Java, and using needing to install Java Virtual Machine, specific training process is as follows:
S621, in order to improve discrimination, prepare license plate image of the 100fu6 Jing Guo the processed binaryzation of above step, and
Convert tif file, name into respectively: 1.GIF, 2.GIF ... ..., 100.GIF, be stored in D: chepai under xunlian；
S622, jTessBoxEditor is opened, clicks Tools- > Merge Tiff, select previously mentioned 100 license plates
Tif file is new.fontcp.exp0.GIF and newly-generated tif is merged into new mesh the new file designation of generation is merged
Record D: chepai xunlian under new.
S623, box file is generated, executes and generates new.fontcp.exp0.box file to issue orders:
tesseract new.fontcp.exp0.GIF new.fontcp.exp0-l eng-psm
7batch.nochopmakebox
S624, modification box file, are switched to the Box Editor page of jTessBoxEditor tool, click open, open
The tiff file new.fontcp.exp0.GIF of front, tool can load corresponding box file automatically, check box data, will
The text of mistake, letter, digital manual amendment, which all checks, to be terminated and saves；
S625, font_properties is generated, executes echo order and generates font_properties；
echo fontcp 0 0 0 0 0>font_properties
S626, training file is generated, executed to issue orders, generate new.fontcp.exp0.tr training file；
tesseract new.fontcp.exp0.GIF new.fontcp.exp0-l eng-psm 7 nobatch
box.train
S627, character set file is generated, executes to issue orders, generates the character set file of entitled unicharset；
unicharset_extractor new.fontcp.exp0.box
Generate shape file；
S628, it executes to issue orders, generates shape file；
shapeclustering-F font_properties-U unicharset-O langyp.unicharset
langyp.fontyp.exp0.tr
S629, aggregation character feature file is generated；
Order is executed, 3 characteristic character files, unicharset, inttemp, pffmtable are generated
mftraining-F font_properties-U unicharset-O new.unicharset
new.fontcp.exp0.tr
S6210, character normalization tag file is generated；
Order is executed, normalization tag file normproto is generated；
cntraining new.fontcp.exp0.tr
S6211, renaming, execute order, and step S629, the tag file that step S6210 is generated is renamed；
rename normproto fontcp.normproto
rename inttemp fontcp.inttemp
rename pffmtable fontcp.pffmtable
rename unicharset fontcp.unicharset
rename shapetable fontcp.shapetable
S6212, merge training file, execute to issue orders, generate fontcp.traineddata file；
combine_tessdata fontyp.
S6213, license plate are without letter O and letter I, and license plate is only capitalization and number, therefore in order to avoid knowing
Mistake occurs in, the white list that program can be set only identifies capitalization and number, be added in program blacklist letter O with
Alphabetical I.
Export license plate recognition result.
Experimental result
In order to verify reliability of the invention, the automobile image 583 for acquiring slight haze weather respectively is opened, moderate haze
Lower automobile image 612 is opened, and automobile image 605 is opened under severe haze, carries out experimental verification.By verifying, in slight haze condition
Under, it can correctly identify 552, discrimination 94.68%.548 can be correctly identified under moderate haze, discrimination is
89.54%.Under severe haze, 485 can be identified, discrimination 80.17%.From the point of view of recognition result, it is based on dark primary
The defogging algorithm that priori and Retinex algorithm combine can reach to the good defog effect of image, know to the license plate in greasy weather
Accuracy not with higher.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；To the greatest extent
Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to
So be possible to modify the technical solutions described in the foregoing embodiments, or to some or all of the technical features into
Row equivalent replacement；And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The range of scheme.
Claims (6)
1. a kind of greasy weather automatic license plate recognition method, it is characterised in that have following steps:
S1, image defogging:
Dark primary priori fogging method carries out defogging processing to the greasy weather automobile image taken, and successively carries out image enhancement, coloured silk
The recovery of chromatic graph picture, image correction and image enhancement, obtain enhanced image；
S2, image preprocessing:
Gray processing processing is carried out to the enhanced image that step S1 is obtained, the gray level image that gray processing is handled is carried out
Median filter process carries out canny edge extracting to the image that median filter process obtains, obtains the marginal information of image；
S3, License Plate:
S31, Morphological scale-space is carried out to the image after edge extracting；
S32, the boundary that license plate is determined by sciagraphy；
S4, binary conversion treatment is carried out to the license plate image on determining license plate boundary；
S5, rectification is carried out to license plate using Radon transformation；
S6, Car license recognition:
The identification of characters on license plate is carried out using Tesseract character recognition engine and jTessBoxEditor training tool.
2. greasy weather automatic license plate recognition method according to claim 1, it is characterised in that: the specific steps of the step S1
Are as follows:
S11, the color lump that the greasy weather automobile image of input is divided into 15*15 seek part and global dark primary figure；
S12, assume atmosphere light ingredient it is known that estimate atmospheric transmissivity by dark primary priori theoretical, the mist figure Cheng Mo used
Type are as follows:
I (y)=J (y) t (x)+A (1-t (x))
Wherein t (x) is atmospheric transmissivity, and I is the foggy image that takes of camera, and J is image after defogging, A be atmosphere light at
Point；
Pass through the calculation formula of dark primary priori theoretical and the available atmospheric transmissivity discreet value of mathematical derivation are as follows:
Wherein Ω (x) is the regional area centered on x, and ω is a constant, takes 0.95, Ic(y) have for what camera took
One in a certain Color Channel of R, G, B of mist image I (y), AcFor in a certain Color Channel of R, G, B of atmosphere light ingredient A
One；
S13, dark primary prior estimate atmosphere light ingredient is utilized:
Preceding 0.1% pixel is taken to have then in these positions what camera took by brightness size from dark channel diagram
Estimated value of the value of the corresponding point with maximum brightness as atmosphere light ingredient A is found in mist image I；
S14, fog free images are restored:
By known I, A, t substitute into mist figure at mathematical model in, can be in the hope of:
Wherein t0It is the lower limit value of transmissivity setting, as t < t0When, enable t=t0, take t0=0.1；
S15, image enhancement and color image restoration, various dimensions are carried out to fog free images are restored using multi-Scale Retinex Algorithm
Retinex algorithm is as follows:
Wherein RMSR(x, y) represents the output of multiple dimensioned Retinex, which Color Channel i represents, due to being RGB image, therefore N
=3, WkExpression and FkRelevant weight coefficient, Ii(x, y) is i-th of the Color Channel or i-th of wave band of input picture, and * is volume
Product operation, K represents the number of scaling function, and as K=1, multiple dimensioned MSR is converted into single scale SSR, Fk(x, y) indicates k-th
Surround function takes Gauss surround function, and meets ∫ ∫ Fk(x, y)=dxdy=1；
S16, image J after defogging is decomposed into tri- width gray level image of R, G, B, and successively by pixel ash each in this three width grayscale image
The data type of angle value becomes double type, is then once operated to three width images respectively:
Determine Gaussian environment function formula are as follows:
Choose three different standard deviation sigmaskRespectively 13,73,130, and according to ∫ ∫ Fk(x, y)=dxdy=1 finds out Ck；
Image is modified using gain/offset method, then the gray value of revised image is projected into tonal range
In (0~255), specific algorithm is as follows:
Rxz(x, y)=G × RMSR(x,y)+offset
Wherein RMSR、Rxz、RoThe respectively gray value of input picture, the gray value after correcting, the gray scale exported after projection
Value, gain coefficient G and offset take 3 and 50, r respectivelymaxAnd rminRespectively RxzMinimum and maximum gray value；
S17, image R, image G and image B being carried out to the above operation, available three width image carries out enhanced image,
Then image R, image G and image B are merged by multi-Scale Retinex Algorithm, obtains the foggy image I that camera takes
Enhanced image.
3. greasy weather automatic license plate recognition method according to claim 1, it is characterised in that: the specific step of the step S31
It is rapid as follows:
Etching operation is carried out to the image after edge extracting, erosion algorithm expression formula is as follows:
Wherein B is structural element, and A` is the image after edge extracting；
Expansive working is carried out to the image after corrosion and filling cavity, expansion algorithm expression formula are as follows:
4. greasy weather automatic license plate recognition method according to claim 3, it is characterised in that: the specific step of the step S32
It is rapid as follows:
Determine the up-and-down boundary of license plate, specific steps are as follows:
A1, the projection that horizontal direction is done to the image after Morphological scale-space, and the threshold value of license plate pixel is set as T, and taking T is 50；
A2, the place of pixel highest point is scanned downwards line by line from projection, when the quantity of scanning to white pixel is greater than threshold value T
When, then it is assumed that it is license plate coboundary, and is recorded as LPT；
A3, continue scanning downwards, when scanning to white pixel quantity is less than threshold value T, then it is assumed that be license plate lower boundary, and record
For LPB；
License plate up-and-down boundary is such as not detected, then enables T=T-3, and repeat step A1-A3；
Determine the right boundary of license plate, specific steps are as follows:
B1, the projection that vertical direction is done to the image after Morphological scale-space, and license plate pixel threshold is set as T1；
B2, the most left place of pixel is scanned to the right from projection, when scanning to white pixel point number is greater than threshold value T1, then recognize
To be license plate left margin, and it is recorded as LPL；
B3, continuation scan to the right, when scanning to white pixel number is less than threshold value T1Shi Ze is considered license plate right margin, and is recorded as
LPR；
License plate right boundary is such as not detected, then enables T1=T1- 1, and repeat step B1-B3.
5. greasy weather automatic license plate recognition method according to claim 4, it is characterised in that: the specific steps of the step S4
It is as follows:
S41, the license plate image progress gray processing processing to license plate boundary is determined；
Threshold value T after the gray processing processing that S42, calculating step S41 are obtained at the pixel f (x, y) of license plate1` (x, y), wherein window
Mouth value ω selection 10, parameter lambda selection 0.48:
S43, the median filtering that template W is 3 × 3 is carried out to pixel f (x, y):
G (x, y)=med { f (x-k, y-l), k, l ∈ W }
S44, the threshold value T for calculating G (x, y) point after gaussian filtering2` (x, y):
S45, the binaryzation point-by-point to f (x, y), F (i, j) indicate final binarization result, proportionality coefficient α=0.1:
6. greasy weather automatic license plate recognition method according to claim 5, it is characterised in that:
To any angle β, the formula of the Radon transformation of function f (x, y) are as follows:
Wherein:
The specific steps of step S5 are as follows:
S51, Radon transformation is carried out in [- 20 °, 20 °] section；
S52, ask Radon transformation results in step S51 the absolute value of first derivative cumulative respectively and, there is maximum cumulative sum
Angle is tilt angle.
S53, the level correction of vehicle license plate, rotation transformation are realized to license plate rotation β angle are as follows:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811039768.4A CN109190621A (en) | 2018-09-06 | 2018-09-06 | Greasy weather automatic license plate recognition method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811039768.4A CN109190621A (en) | 2018-09-06 | 2018-09-06 | Greasy weather automatic license plate recognition method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109190621A true CN109190621A (en) | 2019-01-11 |
Family
ID=64914914
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811039768.4A Pending CN109190621A (en) | 2018-09-06 | 2018-09-06 | Greasy weather automatic license plate recognition method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109190621A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110490163A (en) * | 2019-08-26 | 2019-11-22 | 西南交通大学 | A kind of railway video intelligent data processing method and device |
| CN110634222A (en) * | 2019-08-27 | 2019-12-31 | 河海大学 | Bank bill information identification method |
| CN111297396A (en) * | 2020-02-24 | 2020-06-19 | 华东师范大学 | Ultrasonic pleural ray automatic identification and positioning method based on Radon transformation |
| CN111582268A (en) * | 2020-04-13 | 2020-08-25 | 浙江大华技术股份有限公司 | License plate image processing method and device and computer storage medium |
| CN111639542A (en) * | 2020-05-06 | 2020-09-08 | 中移雄安信息通信科技有限公司 | License plate recognition method, device, equipment and medium |
| CN111784601A (en) * | 2020-06-27 | 2020-10-16 | 浙江同善人工智能技术有限公司 | Image defogging method |
| CN112598002A (en) * | 2020-12-07 | 2021-04-02 | 南京航空航天大学 | License plate recognition method under influence of fog and noise |
| CN114166167A (en) * | 2021-12-07 | 2022-03-11 | 中国铁道科学研究院集团有限公司 | Contour shape detection system |
| CN114414660A (en) * | 2022-03-18 | 2022-04-29 | 盐城工学院 | Method for identifying axle number and cracks of railway vehicle wheel set |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103065138A (en) * | 2012-12-06 | 2013-04-24 | 新疆公众信息产业股份有限公司 | Recognition method of license plate number of motor vehicle |
| CN103093202A (en) * | 2013-01-21 | 2013-05-08 | 信帧电子技术（北京）有限公司 | Car logo locating method and car logo locating device |
| CN103870803A (en) * | 2013-10-21 | 2014-06-18 | 北京邮电大学 | Vehicle license plate recognition method and system based on coarse positioning and fine positioning fusion |
| CN107358795A (en) * | 2017-07-17 | 2017-11-17 | 陈剑桃 | A kind of effective Vehicle License Plate Recognition System |
-
2018
- 2018-09-06 CN CN201811039768.4A patent/CN109190621A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103065138A (en) * | 2012-12-06 | 2013-04-24 | 新疆公众信息产业股份有限公司 | Recognition method of license plate number of motor vehicle |
| CN103093202A (en) * | 2013-01-21 | 2013-05-08 | 信帧电子技术（北京）有限公司 | Car logo locating method and car logo locating device |
| CN103870803A (en) * | 2013-10-21 | 2014-06-18 | 北京邮电大学 | Vehicle license plate recognition method and system based on coarse positioning and fine positioning fusion |
| CN107358795A (en) * | 2017-07-17 | 2017-11-17 | 陈剑桃 | A kind of effective Vehicle License Plate Recognition System |
Non-Patent Citations (3)
| Title |
|---|
| LIHONG ZHENG 等: "An algorithm for accuracy enhancement of license plate recognition", 《ELSEVIER》 * |
| 刘振发: "雾约束下的车牌识别方法研究与实现", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 * |
| 赵春丽 等: "基于暗通道及多尺度Retinex的雾霾天气图像增强算法", 《激光杂志》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110490163A (en) * | 2019-08-26 | 2019-11-22 | 西南交通大学 | A kind of railway video intelligent data processing method and device |
| CN110634222A (en) * | 2019-08-27 | 2019-12-31 | 河海大学 | Bank bill information identification method |
| CN110634222B (en) * | 2019-08-27 | 2021-07-09 | 河海大学 | Bank bill information identification method |
| CN111297396A (en) * | 2020-02-24 | 2020-06-19 | 华东师范大学 | Ultrasonic pleural ray automatic identification and positioning method based on Radon transformation |
| CN111297396B (en) * | 2020-02-24 | 2021-05-07 | 华东师范大学 | Ultrasonic pleural ray automatic identification and positioning method based on Radon transformation |
| CN111582268B (en) * | 2020-04-13 | 2023-05-12 | 浙江大华技术股份有限公司 | License plate image processing method and device and computer storage medium |
| CN111582268A (en) * | 2020-04-13 | 2020-08-25 | 浙江大华技术股份有限公司 | License plate image processing method and device and computer storage medium |
| CN111639542A (en) * | 2020-05-06 | 2020-09-08 | 中移雄安信息通信科技有限公司 | License plate recognition method, device, equipment and medium |
| CN111784601A (en) * | 2020-06-27 | 2020-10-16 | 浙江同善人工智能技术有限公司 | Image defogging method |
| CN111784601B (en) * | 2020-06-27 | 2024-05-14 | 浙江同善人工智能技术有限公司 | Image defogging method |
| CN112598002A (en) * | 2020-12-07 | 2021-04-02 | 南京航空航天大学 | License plate recognition method under influence of fog and noise |
| CN112598002B (en) * | 2020-12-07 | 2024-04-30 | 南京航空航天大学 | License plate recognition method under influence of fog and noise |
| CN114166167A (en) * | 2021-12-07 | 2022-03-11 | 中国铁道科学研究院集团有限公司 | Contour shape detection system |
| CN114414660A (en) * | 2022-03-18 | 2022-04-29 | 盐城工学院 | Method for identifying axle number and cracks of railway vehicle wheel set |
| CN114414660B (en) * | 2022-03-18 | 2024-01-12 | 盐城工学院 | Axle number and crack identification method for railway vehicle wheel set |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109190621A (en) | Greasy weather automatic license plate recognition method | |
| CN109784344B (en) | Image non-target filtering method for ground plane identification recognition | |
| CN103065138B (en) | Recognition method of license plate number of motor vehicle | |
| CN113128442B (en) | Chinese character handwriting style identification method and scoring method based on convolutional neural network | |
| KR100902491B1 (en) | System for processing digit image, and method thereof | |
| Yang et al. | A novel approach for license plate character segmentation | |
| CN109993161B (en) | Text image rotation correction method and system | |
| CN110060221B (en) | Bridge vehicle detection method based on unmanned aerial vehicle aerial image | |
| CN111222507B (en) | Automatic identification method for digital meter reading and computer readable storage medium | |
| CN114387591A (en) | License plate recognition method, system, equipment and storage medium | |
| CN113449632A (en) | Vision and radar perception algorithm optimization method and system based on fusion perception and automobile | |
| CN113435407A (en) | Small target identification method and device for power transmission system | |
| CN111539980A (en) | Multi-target tracking method based on visible light | |
| CN114581914A (en) | License plate recognition method in haze environment | |
| CN110033425B (en) | Interference area detection device and method and electronic equipment | |
| CN113537037A (en) | Pavement disease identification method, system, electronic device and storage medium | |
| CN110188693B (en) | Improved complex environment vehicle feature extraction and parking discrimination method | |
| JP7264428B2 (en) | Road sign recognition device and its program | |
| Bala et al. | Image simulation for automatic license plate recognition | |
| CN116823756A (en) | Pile leg weld defect detection method | |
| CN116883868A (en) | Unmanned aerial vehicle intelligent cruising detection method based on adaptive image defogging | |
| CN107292303A (en) | Method and apparatus for license plate inspection with edge type sliding concentric window | |
| Lin et al. | Locating license plate based on edge features of intensity and saturation subimages | |
| CN112801963B (en) | Video image occlusion detection method and system | |
| Wu et al. | A detection method of road traffic sign based on inverse perspective transform |
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: 20190111 |
|
| RJ01 | Rejection of invention patent application after publication |