CN108717529A - A kind of extremely low quality license plate character recognition method and system based on vehicle - Google Patents
A kind of extremely low quality license plate character recognition method and system based on vehicle Download PDFInfo
- Publication number
- CN108717529A CN108717529A CN201810476220.XA CN201810476220A CN108717529A CN 108717529 A CN108717529 A CN 108717529A CN 201810476220 A CN201810476220 A CN 201810476220A CN 108717529 A CN108717529 A CN 108717529A
- Authority
- CN
- China
- Prior art keywords
- image
- extremely low
- vehicle
- license plate
- low quality
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/625—License plates
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Character Discrimination (AREA)
- Image Processing (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a kind of extremely low quality license plate character recognition method and system based on vehicle, the method is first according to vehicle model and color acquisition standard picture in extremely low quality license plate image, secondly Fourier transformation processing is carried out respectively to extremely low mass x vehicle image and standard picture, then point spread function is calculated, obtain deconvolution image restoration model, extremely low mass x vehicle image is restored according to the deconvolution image restoration model finally, obtains clearly license plate image relatively.The system comprises image capture module, image processing module, computing module and image restoration modules.Deconvolution Algorithm Based on Frequency and point spread function are combined and are applied in the identification of extremely low quality license plate image character by the present invention, have obtained preferable recognition effect, detect for intelligent transportation and public security organ's criminal investigation and figure and provide scientific basis, have larger practical value.
Description
Technical field
The present invention relates to a kind of extremely low quality license plate character recognition method and system based on vehicle, belongs to artificial intelligence figure
As identification technology field.
Background technology
Vehicle is the vehicles that today's society generally uses, and along with expanding economy, car ownership sharply increases.
Along with detected for break in traffic rules and regulations, information of vehicles identification, vehicle Flow Detection etc. intelligent transportation system be widely applied,
Car license recognition has received widespread attention and studies as the important topic of intelligent transportation system.
Car license recognition is to obtain it using technologies such as image procossing, pattern-recognitions to image, the video in video equipment
In vehicle license information.Usual Car license recognition is broadly divided into two kinds, and one is higher to the quality obtained in video equipment
License plate image is identified, and the automatic detection of such speed, accuracy rate and license plate area for focusing on identification is mainly used for vehicle
Room entry/exit management, automatic charging, licence plate registration；Another kind is relatively low to picture quality, human eye is difficult to the license plate image judged, is used
Different image processing method restore or improve image can identification be mainly used for public security to obtain better recognition capability
Organ's criminal investigation and figure are detectd.In many criminal cases, offender uses vehicle as the vehicles, and the number-plate number is as identification
The most effective approach of vehicle is one of the clue of most worthy during cracking of cases.But in the acquisition process of image by
In factors such as the inaccurate, hypertelorisms of focusing so that the number-plate number is difficult to distinguish, directly influences public security department to case-involving vehicle
Assert, prodigious difficulty is brought to the detection work of case.Therefore, for the identification of the relatively low car plate of picture quality, have extensively
With urgent application demand.
Car license recognition has had a large amount of scholars since proposition and studies this, and has achieved many achievements, many identification systems
System is all ripe and comes into operation, and discrimination has reached 99% or more.But these systems are typically only capable to processing high-definition camera
The high quality license plate image that head obtains, it is often helpless for the too low image recognition of picture quality.In recent years, about low-quality
Amount image restoration gradually becomes the hot spot studied both at home and abroad.Image restoration refers to removal or weakens in image acquisition procedures
The quality degradation of generation is recovered to the process for the original image that do not degenerate as far as possible.Liftering, Wiener filtering, Constrained
Least-squares algorithm and Lucy-Richardson are still currently used Deconvolution Method as classic algorithm.
Wherein, liftering is multiple for digital picture as a kind of common Deconvolution Method since the sixties in last century
It is former.In the ideal case, any one width blurred picture can be regarded as a width clear image and point spread function (Point
Spread Function, abbreviation PSF) convolution result.Nathan is obtained with two-dimentional liftering method to handle extraterrestrial exploration satellite
Image, contemporaneity, Harris, Mcglamery et al. also use liftering caused by atmospheric perturbation obscure handles,
Hereafter liftering becomes a kind of classic algorithm of image restoration.But due to ignoring noise, liftering method is there is no noises
In the case of can accurate restored image seriously affect recovery effect if there are noise.1967, it is contemplated that most of
Image neighbor pixel has high correlation, while reducing noise jamming, and Wiener filtering is used for image restoration by Helstrom.
1973, it is similar with Wiener filtering that Hunt B.R using circular matrix model propose least square with equality constraint method, both
Noise and the priori of PSF are needed, the difference is that the latter, which passes through, adds constraints so that restoration result is unique.Lucy-
Richardson algorithms still can obtain preferably restoring knot in the case where not knowing original image and noise prior information
Fruit.This method assumes that blurred picture obeys Poisson distribution, and restoration result is obtained by iteratively solving maximal possibility estimation.Although
Lucy-Richardson algorithms only need iteration to be achieved with relatively satisfactory restoration result several times, but do not provide specific change
For end condition.Experiment shows that in a certain range, iterations are more, and deblurring effect is better.But iterations reach one
After fixed number amount, even if being further added by iterations, deblurring effect will not be significantly improved, ringing can be aggravated instead and made an uproar
The influence of sound.
But in actual conditions, the convolution kernel that degrades of image is often unknown, is recovered using low-quality image original
Image, i.e. blindly restoring image.Even to this day, it has been proposed that many method for blindly restoring image.1988, Avers and Daintv were carried
Go out the iterative blind deconvolution algorithm (IDB) based on Fast Fourier Transform, its main feature is that required calculation amount is little, but
Major defect is a lack of reliability, and the uniqueness and convergence of solution not can determine that.Later, many documents propose on this basis
Improved method, such as Lane conjugate gradient methods reduce and the unstable related problem of algorithm, it may be said that algorithm is still
A kind of now popular blind restoration algorithm.1996, the propositions such as Kundur had nonnegativity and support region constraints
Recursive inverse filtering (NAS-RIF) algorithm.One important advantage of the algorithm is that cost function is convex function, thus algorithm has very well
Convergence, and solve the problems, such as that Algorithm Convergence is bad and algorithm is computationally intensive, but the algorithm shortcomings are to noise-sensitive.
2000, Galatsanos, Mesarovic, Katsaggelos et al. the feelings in known portions obscuring image information
Under condition, the blind recovery that blurred picture is carried out with EA (evidence analysis) algorithm of condition Bayes, its essence are proposed
It is also a kind of iterative algorithm, but calculation amount is also very big.
In conclusion the identification for extremely low quality characters on license plate, there is presently no very effective methods, can reduce
Algorithm calculation amount, and the recognition effect of extremely low quality license plate image character can be improved.
Invention content
For the deficiency for solving in the above-mentioned prior art, the present invention provides a kind of extremely low quality characters on license plate based on vehicle
Recognition methods and system can not only reduce algorithm calculation amount, and can improve the knowledge of extremely low quality license plate image character
Other effect.
The present invention solves its technical problem and adopts the technical scheme that：
On the one hand, a kind of extremely low quality license plate character recognition method based on vehicle provided in an embodiment of the present invention, it is wrapped
Include following steps：
Step 1：According to vehicle model and color acquisition standard picture in extremely low quality license plate image；
Step 2：Fourier transformation processing is carried out respectively to extremely low mass x vehicle image and standard picture；
Step 3：Point spread function is calculated, deconvolution image restoration model is obtained；
Step 4：Extremely low mass x vehicle image is restored according to the deconvolution image restoration model, is obtained relatively clear
Clear license plate image.
As a kind of possible realization method of the present embodiment, the step 1 specifically includes following steps：
Step 11, a width sample image I (x, y) is selected from extremely low quality license plate image, is determined according to the sample image
The wherein corresponding vehicle model of vehicle and vehicle color；
Step 12, according to the weather conditions shown in sample image, identical weather conditions acquisition standard picture is selected；
Step 13, it is placed on and sample image environment same model corresponding with car plate to be identified and with the vehicle of color
In, using the identical camera with shooting sample image I (x, y), pass through focusing, the clear vehicle image I ' of zoom operation one width of shooting
(x, y) is used as standard picture.
As a kind of possible realization method of the present embodiment, the identical weather conditions refer to be all fine day, rain, snow
Or the basic weather pattern of dense fog.
As a kind of possible realization method of the present embodiment, the step 2 specifically includes following steps：
Step 21：Fourier transformation processing is carried out to the sample image I (x, y) selected from extremely low mass x vehicle image,
Obtain sample image Fourier transformation result G (ω 1, ω 2)；
Step 22：Fourier transformation processing is carried out to the standard picture I ' (x, y) of shooting, obtains standard picture Fourier change
Change result F (ω 1, ω 2).
As a kind of possible realization method of the present embodiment, the step 3 specifically includes following steps：
Step 31：With sample image Fourier transformation result G (ω 1, ω 2) divided by standard picture Fourier transformation result F
(ω 1, ω 2) obtains quotient H (ω 1, ω 2)；
Step 32：Inverse Fourier transform is carried out to quotient H (ω 1, ω 2) and obtains point spread function h (x, y).
As a kind of possible realization method of the present embodiment, the step 4 specifically includes following steps：
Step 41：Fourier transformation is carried out to extremely low-quality sample image I (x, y), obtains transformation results G ' (ω 1, ω
2)；
Step 42：It is gone with the result G ' (ω 1, ω 2) of extremely low quality sample image Fourier transformation divided by point spread function
Fourier transformation result H (ω 1, ω 2), obtain quotient F (ω 1, ω 2)；
Step 43：Inverse Fourier transform is carried out to quotient F (ω 1, ω 2), obtains the knot after extremely low quality sample image restoration
Fruit image f (x, y).
On the other hand, a kind of extremely low quality Recognition of License Plate Characters system based on vehicle provided in an embodiment of the present invention, it
Including：
Image capture module：Extremely low quality license plate image and standard picture are obtained,
Image processing module：Fourier transformation processing is carried out respectively to extremely low mass x vehicle image and standard picture；
Computing module：It is multiple to obtain deconvolution image for the point spread function for calculating extremely low quality license plate image and standard picture
Master mould；
Image restoration module：Extremely low mass x vehicle image is restored according to the deconvolution image restoration model, is obtained
To clearly license plate image relatively.
As a kind of possible realization method of the present embodiment, the standard picture be according to extremely low quality license plate image its
Vehicle model in middle piece image and vehicle color select identical weather conditions vehicle same position in sample image environment
The clear vehicle image of a width obtained by focusing, zoom operation shooting.
The technical solution of the embodiment of the present invention can have the advantage that as follows：
A kind of extremely low quality license plate character recognition method based on vehicle of the embodiment of the present invention technical solution, first
According to vehicle model and color acquisition standard picture in extremely low quality license plate image, secondly to extremely low mass x vehicle image and standard
Image carries out Fourier transformation processing respectively, then calculates point spread function, deconvolution image restoration model is obtained, finally to pole
Low quality vehicle image is restored according to the deconvolution image restoration model, obtains clearly license plate image relatively.This hair
It is bright that Deconvolution Algorithm Based on Frequency and point spread function are combined and are applied in the identification of extremely low quality license plate image character, obtained compared with
Good recognition effect is detectd for intelligent transportation and public security organ's criminal investigation and figure and provides scientific basis, has larger practical value.
The present invention establishes an image restoration model using Deconvolution Algorithm Based on Frequency, and low quality license plate image is restored, is obtained
To clearly license plate image relatively.Its core methed is to solve point spread function using the same model in clear image, will be anti-
Convolution algorithm is used in the recovery of extremely low quality license plate image, can be preferably extremely low-quality image restoration at relatively clear
License plate image, scientific basis is provided for subsequent License Plate Segmentation and identification, for the criminal investigation of intelligent transportation and public security organ
It is detectd with larger practical value with figure.
By carrying out image restoration to blurred picture deconvolution, key technology is to calculate point spread function the present invention.
Although characters on license plate in extremely low quality image is difficult to, but can recognize with ease that out the color and vehicle of vehicle, that
The vehicle that a same color and model can be placed under same position and similar weather condition, for obtaining a width
Equally fuzzy vehicle image is obtaining the clear image of a panel height by zoom, the operations such as further, can be with according to this two images
Point spread function is calculated, so as to carry out deconvolution recovery to original extremely low quality license plate image.
A kind of extremely low quality Recognition of License Plate Characters system based on vehicle of the embodiment of the present invention technical solution, it includes
Image capture module, image processing module, computing module and image restoration module, described image acquisition module are extremely low for obtaining
Quality license plate image and standard picture, described image processing module be used for extremely low mass x vehicle image and standard picture respectively into
The processing of row Fourier transformation；The computing module is used to calculate the point spread function of extremely low quality license plate image and standard picture,
Obtain deconvolution image restoration model；Described image restoration module is used for extremely low mass x vehicle image according to the deconvolution figure
As restoration model is restored, clearly license plate image relatively is obtained.The present invention mutually ties Deconvolution Algorithm Based on Frequency with point spread function
Merge in the identification applied to extremely low quality license plate image character, obtained preferable recognition effect, is intelligent transportation and public security
Organ's criminal investigation and figure, which are detectd, provides scientific basis, has larger practical value.
Description of the drawings
Fig. 1 is a kind of extremely low quality license plate character recognition method based on vehicle shown according to an exemplary embodiment
Flow chart；
Fig. 2 (a), Fig. 2 (b) and Fig. 2 (c) are three extremely low quality license plate images；
Fig. 3 (a), Fig. 3 (b) and Fig. 3 (c) are respectively to extremely low quality car plate figure shown in Fig. 2 (a), Fig. 2 (b) and Fig. 2 (c)
Result schematic diagram as carrying out Fourier transformation；
Fig. 4 (a), Fig. 4 (b) and Fig. 4 (c) are a kind of extremely low quality Recognition of License Plate Characters based on vehicle using the present invention
Method respectively carries out extremely low quality license plate image shown in Fig. 2 (a), Fig. 2 (b) and Fig. 2 (c) result schematic diagram of image restoration.
Specific implementation mode
In order to clarify the technical characteristics of the invention, below by specific implementation mode and combining its attached drawing to the present invention
It is described in detail.Following disclosure provides many different embodiments or example is used for realizing the different structure of the present invention.
In order to simplify disclosure of the invention, hereinafter the component of specific examples and setting are described.In addition, the present invention can be not
With repeat reference numerals in example and/or letter.This repetition is for purposes of simplicity and clarity, itself not indicate to be begged for
By the relationship between various embodiments and/or setting.It should be noted that illustrated component is painted not necessarily to scale in the accompanying drawings
System.Present invention omits the descriptions to known assemblies and treatment technology and process to avoid the present invention is unnecessarily limiting.
By carrying out image restoration to blurred picture deconvolution, key technology is to calculate point spread function the present invention.
Although characters on license plate in extremely low quality image is difficult to, but can recognize with ease that out the color and vehicle of vehicle, that
The vehicle that a same color and model can be placed under same position and similar weather condition, for obtaining a width
Equally fuzzy vehicle image is obtaining the clear image of a panel height by zoom, the operations such as further, can be with according to this two images
Point spread function is calculated, so as to carry out deconvolution recovery to original extremely low quality license plate image.
Fig. 1 is a kind of extremely low quality license plate character recognition method based on vehicle shown according to an exemplary embodiment
Flow chart.As shown in Figure 1, a kind of extremely low quality license plate character recognition method based on vehicle of the embodiment, may include with
Lower step：
Step 1：According to vehicle model and color acquisition standard picture in extremely low quality license plate image；
Step 2：Fourier transformation processing is carried out respectively to extremely low mass x vehicle image and standard picture；
Step 3：Point spread function is calculated, deconvolution image restoration model is obtained；
Step 4：Extremely low mass x vehicle image is restored according to the deconvolution image restoration model, is obtained relatively clear
Clear license plate image.
In one possible implementation, the step 1 specifically includes following steps：
Step 11, a width sample image I (x, y) is selected from extremely low quality license plate image, is determined according to the sample image
The wherein corresponding vehicle model of vehicle and vehicle color；
Step 12, according to the weather conditions shown in sample image, identical weather conditions acquisition standard picture is selected；
Step 13, it is placed on and sample image environment same model corresponding with car plate to be identified and with the vehicle of color
In, using the identical camera with shooting sample image I (x, y), pass through focusing, the clear vehicle image I ' of zoom operation one width of shooting
(x, y) is used as standard picture.
In step 1, according to the vehicle model and vehicle color in the wherein piece image of extremely low quality license plate image, choosing
It selects identical weather conditions vehicle same position in sample image environment and passes through focusing, the zoom operation shooting clear vehicle figure of one width
As being used as standard picture.Wherein, the identical weather conditions refer to be all fine day, rain, snow or the basic weather pattern of dense fog.
In one possible implementation, the step 2 specifically includes following steps：
Step 21：Fourier transformation processing is carried out to the sample image I (x, y) selected from extremely low mass x vehicle image,
Obtain sample image Fourier transformation result G (ω 1, ω 2)；
Step 22：Fourier transformation processing is carried out to the standard picture I ' (x, y) of shooting, obtains standard picture Fourier change
Change result F (ω 1, ω 2).
In one possible implementation, the step 3 specifically includes following steps：
Step 31：With sample image Fourier transformation result G (ω 1, ω 2) divided by standard picture Fourier transformation result F
(ω 1, ω 2) obtains quotient H (ω 1, ω 2), and calculation formula is as shown in Equation 1：
Step 32：Inverse Fourier transform is carried out to quotient H (ω 1, ω 2) and obtains point spread function h (x, y), calculation formula
As shown in Equation 2：
I (x, y)=I'(x, y) * h (x, y) (2)
Any one extremely low-quality vehicle image can be approximately considered be its standard picture and point spread function h (x,
Y) convolution as a result, according to h (x, y) to extremely low mass x vehicle image carry out deconvolution can be obtained clear image.
In one possible implementation, the step 4 specifically includes following steps：
Step 41：Fourier transformation is carried out to extremely low-quality sample image I (x, y), obtains transformation results G ' (ω 1, ω
2), calculation formula is as shown in Equation 3：
G'(ω1,ω2)=F (I (x, y)) (3)
Step 42：It is gone with the result G ' (ω 1, ω 2) of extremely low quality sample image Fourier transformation divided by point spread function
Fourier transformation result H (ω 1, ω 2), obtain quotient F (ω 1, ω 2), calculation formula is as shown in Equation 4：
Step 43：Inverse Fourier transform is carried out to quotient F (ω 1, ω 2), obtains the knot after extremely low quality sample image restoration
Fruit image f (x, y), calculation formula are as shown in Equation 5：
F (x, y)=F-1(F(ω1,ω2) (5)
The result images after any one extremely low mass x vehicle image restoration can be obtained according to the method for the present invention.
The present embodiment is first according to vehicle model and color acquisition standard picture in extremely low quality license plate image, secondly to pole
Low quality vehicle image and standard picture carry out Fourier transformation processing respectively, then calculate point spread function, obtain deconvolution
Image restoration model finally restores extremely low mass x vehicle image according to the deconvolution image restoration model, obtains phase
To clearly license plate image.Deconvolution Algorithm Based on Frequency and point spread function are combined and are applied to extremely low quality license plate image by the present invention
In the identification of character, obtained preferable recognition effect, for intelligent transportation and public security organ's criminal investigation and figure detect the science of providing according to
According to having larger practical value.
The present invention establishes an image restoration model using Deconvolution Algorithm Based on Frequency, and low quality license plate image is restored, is obtained
To clearly license plate image relatively.Its core methed is to solve point spread function using the same model in clear image, will be anti-
Convolution algorithm is used in the recovery of extremely low quality license plate image, can be preferably extremely low-quality image restoration at relatively clear
License plate image, scientific basis is provided for subsequent License Plate Segmentation and identification, for the criminal investigation of intelligent transportation and public security organ
It is detectd with larger practical value with figure.
A kind of extremely low quality Recognition of License Plate Characters system based on vehicle of the present invention, it includes：
Image capture module：Extremely low quality license plate image and standard picture are obtained,
Image processing module：Fourier transformation processing is carried out respectively to extremely low mass x vehicle image and standard picture；
Computing module：It is multiple to obtain deconvolution image for the point spread function for calculating extremely low quality license plate image and standard picture
Master mould；
Image restoration module：Extremely low mass x vehicle image is restored according to the deconvolution image restoration model, is obtained
To clearly license plate image relatively.
In one possible implementation, the standard picture is the wherein width figure according to extremely low quality license plate image
Vehicle model as in and vehicle color, select identical weather conditions vehicle same position in sample image environment to pass through tune
The clear vehicle image of a width that burnt, zoom operation shooting obtains.
Deconvolution Algorithm Based on Frequency and point spread function are combined and are applied to extremely low quality license plate image character by the present embodiment
In identification, preferable recognition effect has been obtained, has detectd for intelligent transportation and public security organ's criminal investigation and figure and provides scientific basis, have
Larger practical value.
In addition, the application range of the present invention is not limited to the technique, mechanism, system of the specific embodiment described in specification
It makes, material composition, means, method and step.From the disclosure, will be easy as those skilled in the art
Ground understands, for current technique that is existing or will developing later, mechanism, manufacture, material composition, means, method or
Step, the knot that wherein they execute the function being substantially the same with the corresponding embodiment of the invention described or acquisition is substantially the same
Fruit can apply them according to the present invention.Therefore, appended claims of the present invention are intended to these techniques, mechanism, system
It makes, material composition, means, method or step are included in its protection domain.
Claims (8)
1. a kind of extremely low quality license plate character recognition method based on vehicle, characterized in that include the following steps：
Step 1：According to vehicle model and color acquisition standard picture in extremely low quality license plate image；
Step 2：Fourier transformation processing is carried out respectively to extremely low mass x vehicle image and standard picture；
Step 3：Point spread function is calculated, deconvolution image restoration model is obtained；
Step 4：Extremely low mass x vehicle image is restored according to the deconvolution image restoration model, is obtained relatively clearly
License plate image.
2. a kind of extremely low quality license plate character recognition method based on vehicle according to claim 1, characterized in that described
Step 1 specifically includes following steps：
Step 11, a width sample image I (x, y) is selected from extremely low quality license plate image, is determined wherein according to the sample image
The corresponding vehicle model of vehicle and vehicle color；
Step 12, according to the weather conditions shown in sample image, identical weather conditions acquisition standard picture is selected；
Step 13, same model corresponding with car plate to be identified and with the vehicle of color be placed on in sample image environment,
Using with shooting sample image I (x, y) identical camera, by focusing, the clear vehicle image I ' of zoom operation one width of shooting (x,
Y) it is used as standard picture.
3. a kind of extremely low quality license plate character recognition method based on vehicle according to claim 2, characterized in that described
Identical weather conditions refer to be all fine day, rain, snow or the basic weather pattern of dense fog.
4. a kind of extremely low quality license plate character recognition method based on vehicle according to claim 2, characterized in that described
Step 2 specifically includes following steps：
Step 21：Fourier transformation processing is carried out to the sample image I (x, y) selected from extremely low mass x vehicle image, is obtained
Sample image Fourier transformation result G (ω 1, ω 2)；
Step 22：Fourier transformation processing is carried out to the standard picture I ' (x, y) of shooting, obtains standard picture Fourier transformation knot
Fruit F (ω 1, ω 2).
5. a kind of extremely low quality license plate character recognition method based on vehicle according to claim 4, characterized in that described
Step 3 specifically includes following steps：
Step 31：With sample image Fourier transformation result G (ω 1, ω 2) divided by standard picture Fourier transformation result F (ω 1,
ω 2), obtain quotient H (ω 1, ω 2)；
Step 32：Inverse Fourier transform is carried out to quotient H (ω 1, ω 2) and obtains point spread function h (x, y).
6. a kind of extremely low quality license plate character recognition method based on vehicle according to claim 5, characterized in that described
Step 4 specifically includes following steps：
Step 41：Fourier transformation is carried out to extremely low-quality sample image I (x, y), obtains transformation results G ' (ω 1, ω 2)；
Step 42：It is gone with the result G ' (ω 1, ω 2) of extremely low quality sample image Fourier transformation divided by Fu of point spread function
In leaf transformation result H (ω 1, ω 2), obtain quotient F (ω 1, ω 2)；
Step 43：Inverse Fourier transform is carried out to quotient F (ω 1, ω 2), obtains the result figure after extremely low quality sample image restoration
As f (x, y).
7. a kind of extremely low quality Recognition of License Plate Characters system based on vehicle, characterized in that including：
Image capture module：Extremely low quality license plate image and standard picture are obtained,
Image processing module：Fourier transformation processing is carried out respectively to extremely low mass x vehicle image and standard picture；
Computing module：The point spread function for calculating extremely low quality license plate image and standard picture obtains deconvolution image and restores mould
Type；
Image restoration module：Extremely low mass x vehicle image is restored according to the deconvolution image restoration model, obtains phase
To clearly license plate image.
8. a kind of extremely low quality Recognition of License Plate Characters system based on vehicle according to claim 7, characterized in that described
Standard picture is vehicle model and vehicle color in the wherein piece image according to extremely low quality license plate image, and selection phase is on the same day
The vaporous condition clear vehicle figure of a width that vehicle same position is obtained by focusing, zoom operation shooting in sample image environment
Picture.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810476220.XA CN108717529A (en) | 2018-05-17 | 2018-05-17 | A kind of extremely low quality license plate character recognition method and system based on vehicle |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810476220.XA CN108717529A (en) | 2018-05-17 | 2018-05-17 | A kind of extremely low quality license plate character recognition method and system based on vehicle |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108717529A true CN108717529A (en) | 2018-10-30 |
Family
ID=63899918
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810476220.XA Pending CN108717529A (en) | 2018-05-17 | 2018-05-17 | A kind of extremely low quality license plate character recognition method and system based on vehicle |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108717529A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112541386A (en) * | 2019-09-04 | 2021-03-23 | 通用汽车环球科技运作有限责任公司 | Image processing for position recognition of eyes |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080013850A1 (en) * | 2006-07-14 | 2008-01-17 | Junzou Sakurai | Image processing apparatus and image restoration method and program |
| JP2009089082A (en) * | 2007-09-28 | 2009-04-23 | Kyocera Corp | Image restoration apparatus and its method |
| CN102800053A (en) * | 2012-06-18 | 2012-11-28 | 西北工业大学 | Image deblurring method based on redundant dictionary pair joint optimization |
| CN104331871A (en) * | 2014-12-02 | 2015-02-04 | 苏州大学 | Image de-blurring method and image de-blurring device |
| CN104820969A (en) * | 2015-04-03 | 2015-08-05 | 西安交通大学 | Real-time blind image restoration method |
| CN106651791A (en) * | 2016-11-21 | 2017-05-10 | 云南电网有限责任公司电力科学研究院 | Recovery method for single motion blur image |
| CN107092876A (en) * | 2017-04-12 | 2017-08-25 | 湖南源信光电科技股份有限公司 | The low-light (level) model recognizing method combined based on Retinex with S SIFT features |
-
2018
- 2018-05-17 CN CN201810476220.XA patent/CN108717529A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080013850A1 (en) * | 2006-07-14 | 2008-01-17 | Junzou Sakurai | Image processing apparatus and image restoration method and program |
| JP2009089082A (en) * | 2007-09-28 | 2009-04-23 | Kyocera Corp | Image restoration apparatus and its method |
| CN102800053A (en) * | 2012-06-18 | 2012-11-28 | 西北工业大学 | Image deblurring method based on redundant dictionary pair joint optimization |
| CN104331871A (en) * | 2014-12-02 | 2015-02-04 | 苏州大学 | Image de-blurring method and image de-blurring device |
| CN104820969A (en) * | 2015-04-03 | 2015-08-05 | 西安交通大学 | Real-time blind image restoration method |
| CN106651791A (en) * | 2016-11-21 | 2017-05-10 | 云南电网有限责任公司电力科学研究院 | Recovery method for single motion blur image |
| CN107092876A (en) * | 2017-04-12 | 2017-08-25 | 湖南源信光电科技股份有限公司 | The low-light (level) model recognizing method combined based on Retinex with S SIFT features |
Non-Patent Citations (1)
| Title |
|---|
| 彭艳来等: "模糊图像复原技术在车牌图像增强中的应用", 《科技信息》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112541386A (en) * | 2019-09-04 | 2021-03-23 | 通用汽车环球科技运作有限责任公司 | Image processing for position recognition of eyes |
| CN112541386B (en) * | 2019-09-04 | 2024-04-16 | 通用汽车环球科技运作有限责任公司 | Image processing for position recognition of eyes |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Fast image dehazing method based on linear transformation | |
| Huang et al. | An efficient visibility enhancement algorithm for road scenes captured by intelligent transportation systems | |
| Huang et al. | Single satellite optical imagery dehazing using SAR image prior based on conditional generative adversarial networks | |
| Chen et al. | Haze removal using radial basis function networks for visibility restoration applications | |
| CN102665034A (en) | Night effect removal method for camera-collected video | |
| CN109523476B (en) | License plate motion blur removing method for video detection | |
| CN104282003A (en) | Digital blurred image blind restoration method based on gradient screening | |
| Patil et al. | Motion saliency based generative adversarial network for underwater moving object segmentation | |
| CN102236887A (en) | Motion-blurred image restoration method based on rotary difference and weighted total variation | |
| CN108614998B (en) | Single-pixel infrared target detection method | |
| CN109118440B (en) | Single image defogging method based on transmissivity fusion and adaptive atmospheric light estimation | |
| Cheng et al. | Visibility enhancement of single hazy images using hybrid dark channel prior | |
| CN113362375A (en) | Moving object detection method for vehicle | |
| Cheng et al. | A highway traffic image enhancement algorithm based on improved GAN in complex weather conditions | |
| Huang et al. | Single image snow removal using sparse representation and particle swarm optimizer | |
| CN105574826A (en) | Thin cloud removing method of remote sensing image | |
| CN111598814A (en) | Single image defogging method based on extreme scattering channel | |
| Wang et al. | A novel visibility restoration algorithm for single hazy images | |
| CN108717529A (en) | A kind of extremely low quality license plate character recognition method and system based on vehicle | |
| Anantrasirichai et al. | Mitigating the effects of atmospheric distortion using DT-CWT fusion | |
| CN112927155B (en) | Multi-angle SAR image filtering method based on super image | |
| Park et al. | Fast rain removal using rain map and temporal filtering for consumer imaging applications | |
| Liu et al. | Research on lane line segmentation algorithm based on deeplabv3 | |
| Zahra et al. | Visibility enhancement of scene images degraded by foggy weather condition: An application to video surveillance | |
| Hu et al. | Maritime video defogging based on spatial-temporal information fusion and an improved dark channel prior |
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: 20181030 |
|
| RJ01 | Rejection of invention patent application after publication |