CN103186762A - License plate character recognition method based on SURF matching algorithm - Google Patents
License plate character recognition method based on SURF matching algorithm Download PDFInfo
- Publication number
- CN103186762A CN103186762A CN2011104486829A CN201110448682A CN103186762A CN 103186762 A CN103186762 A CN 103186762A CN 2011104486829 A CN2011104486829 A CN 2011104486829A CN 201110448682 A CN201110448682 A CN 201110448682A CN 103186762 A CN103186762 A CN 103186762A
- Authority
- CN
- China
- Prior art keywords
- license plate
- characters
- character
- surf
- recognition method
- 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
Images
Abstract
The invention discloses a simple and effective license plate character recognition method based on an SURF (Speeded-up Robust Features) matching algorithm. The license plate character recognition is conducted by secondary classification of license plate character images, description of SURF keypoints and matching of features. The license plate character recognition method based on the SURF matching algorithm has good affine invariance, and good adaptability to scaling, visual angle variation, illumination difference, shielding and noise, and has higher computing speed compared with the current mainstream feature recognition method. According to the method, license plate characters are integrally classified for the first time, and the characters of the same class are individually subjected to secondary classification, and such classification mode allows the difference of similar characters to be greater and further increases the recognition rate.
Description
Technical field
The present invention relates to pattern-recognition and technical field of image processing, especially a kind of license plate character recognition method based on SURF characteristic matching algorithm.
Background technology
In intelligent transportation system, video technique has become the major technology means.Based on the Vehicle License Plate Recognition System of video technique, obtained development fast in recent years.Car plate identification (License Plate Recognition is called for short LPR) is an indispensable part in the intelligent transportation system.Every field in traffic is widely used, and as detecting in violation of rules and regulations at expressway tol lcollection, high speed overspeed detection, city vehicle and road management violating the regulations fields such as (make a dash across the red light, exceed the speed limit, drive in the wrong direction), car plate identification plays the important and pivotal role.And the increase day by day along with automobile quantity, Vehicle License Plate Recognition System also is applied in public security bayonet socket (vehicle registration system), theft vehicle detection, residential quarter managing system of car parking, various looking in the systems such as truck system, the role who in life, has played the part of " electronic police ", be traffic administration, fighting crime all plays a good role.
Recognition of License Plate Characters is a car plate core technology of identification automatically.Character identifying method commonly used at present has template matches, probability statistics, geometry classification, small echo computing, support vector machine (SVM), analysis of neural network, characteristic matching etc.Every kind of method has the merits and demerits of oneself, and some method implements fairly simple, can obtain recognition effect preferably under some environment.But because the environment for use of Vehicle License Plate Recognition System is complicated and changeable, be difficult to obtain extensive acceptable recognition effect.Though obtain certain raising on some method recognition effect, method is too complicated, or recognition speed is difficult to reach requirement.The character that reality is separated in the license plate image is often because the influence of some extraneous factors, has fracture, distortion, distortion, problem such as stained, brings very big difficulty for the identification of character.Because automatic license plate identification system is in outdoor pick-up, be subjected to damaged, the influence of polluting and blocking etc. factor of travel speed, the car plate of model specification, the vehicle of weather, illumination, background, car plate, exist uncertain.These all can cause interference to real-time processing, the pattern-recognition of system.Discrimination is the important performance indexes of recognition system in the Vehicle License Plate Recognition System in addition, and the working environment of real-time has determined the requirement to recognition speed, and the recognition speed of system and the stability of system, discrimination are the unity of contradictions bodies.How in the unity of contradictions body, to utilize existent method better identification character be to be badly in need of the key issue that solves in the present Recognition of License Plate Characters system.
Summary of the invention
The technical problem to be solved in the present invention provides a kind of simple and effective license plate character recognition method based on SURF characteristic matching algorithm.
For addressing the above problem, a kind of license plate character recognition method based on SURF characteristic matching algorithm of the present invention may further comprise the steps:
1, at first the characters on license plate image is carried out whole first subseries, the characters on license plate image is divided into Chinese character, letter, numeral and character and digit four big classes;
2, according to the characters on license plate characteristics, first classification results is carried out individual secondary classification again, make the character picture with more common ground be classified as a class;
3, at extraction SURF key point feature of all categories, and key point is carried out feature describe;
4, carry out the coupling of the character picture in the class according to the feature descriptor of describing feature;
5, carry out Recognition of License Plate Characters according to matching result.
The first sorting technique of characters on license plate image adopts neural network or SVM clustering algorithm in the described step 1.
The secondary classification method of characters on license plate image adopts neural network or SVM clustering algorithm in the described step 2.
Earlier on the result of secondary classification, its common ground of character picture that will have common ground is got rid of in the described step 3, the extracting section SURF key point feature that character picture is not excluded again, and key point is carried out feature describe.
Adopt the license plate character recognition method based on SURF characteristic matching algorithm of the present invention, not only has good affine unchangeability, and yardstick convergent-divergent, visual angle change, illumination difference, block, the noise aspect all has good adaptive faculty, has higher computing velocity than present main flow characteristic recognition method.Among the present invention, at first characters on license plate has been carried out whole first subseries, and the character of same classification has been adopted the mode of individual secondary classification again, this mode classification makes the otherness of close character higher, can further improve discrimination.
Description of drawings
Fig. 1 is the process flow diagram of the license plate character recognition method based on SURF characteristic matching algorithm of the present invention;
Fig. 2 is the classification synoptic diagram of the license plate character recognition method based on SURF characteristic matching algorithm of the present invention;
Embodiment
In order to make those skilled in the art person understand technical solution of the present invention better, the present invention is described in further detail below in conjunction with drawings and embodiments.
As shown in Figure 1, 2, the license plate character recognition method based on SURF characteristic matching algorithm of the present invention may further comprise the steps:
Step 1: adopt neural network or SVM clustering algorithm that the characters on license plate image is carried out whole first subseries, the characters on license plate image is divided into Chinese character, letter, numeral and character and digit four big classes.
Motor vehicle accession designation number in the car plate has 7 characters, is formed by the particular order combination by Chinese character, English alphabet and arabic numeral, and the numbering of preceding car plate is horizontal single file and arranges.The front two of numbering is motor vehicle registration unit code name, formed by Chinese character and English alphabet respectively, Chinese character is the abbreviation of each provinces, autonomous regions and municipalities: capital, Tianjin, Shanxi, Ji, illiteracy, the Liao Dynasty, Ji, black, Shanghai, Soviet Union, Zhejiang, Anhui, Fujian, Jiangxi, Shandong, Henan, Hubei Province, Hunan, Guangdong, osmanthus, fine jade, river, expensive, cloud, Tibetan, Shan, sweet, blue or green, peaceful, new, change totally 31 Chinese characters; Other field car plate Chinese character: first, second, third, oneself, heptan, suffering, the ninth of the ten Heavenly Stems, the third of the twelve Earthly Branches, occasion, not, the Shen, defend, the noon, make, lead, learn, alert, about more than 50 Chinese characters altogether.Second character of car plate is English capitalization: " A " arrives " Z "; The 3rd character arrives " Z " for the English capitalization " A " except " I " and " 0 " or digital " 0 " arrives " 9 ", totally 34 characters.The 4th character only was that " 0 " arrives " 9 " these ten numerals in the past.According to the design feature of characters on license plate the training sample of whole sorter is divided into four big classes: Chinese characters kind: i.e. all characters on license plate Chinese characters; Alphabetic class: A, C, E, F, G, H, J, K, L, M, N, P, R, U, V, W, X, Y.Numeric class: 1,3,4,6,9; Letter and number class: 0, D, Q, 2, Z, 8, B, 5, S, 7, T.Train whole sorter with the good training sample of classifying, obtain preferable sorter model, then characters on license plate to be identified is carried out the integral body classification.
Step 2: according to the characters on license plate characteristics, adopt neural network or SVM clustering algorithm that the first classification results of whole sorter is carried out individual secondary classification again, make the character picture with more common ground be classified as a class.
For example common radical and the complicated and simple of structure according to Chinese character can be classified Chinese characters kind like this in Chinese characters kind, comprises Tianjin, Shanghai, Zhejiang, Hunan, Chongqing as Chinese character classification 1.Chinese character classification 2 comprises illiteracy, Soviet Union, hides.Chinese character classification 3 comprises: Anhui, Fujian, Jiangxi, Shandong, Henan, Hubei Province, Guangdong, osmanthus, fine jade, Ji, black, police, neck, Shan, make, newly.Chinese character classification 4 comprises: the third of the twelve Earthly Branches, capital, Shanxi, Ji, green grass or young crops, heptan, expensive, occasion, learn, defend.Chinese character classification 5 comprises: the Liao Dynasty, river, cloud, sweet, peaceful, first, second, third, oneself, suffering, the ninth of the ten Heavenly Stems, not, Shen, noon.Letter classification 1 comprises: A, X, Y, W.Letter classification 2 comprises: E, F, H, J, K, L, P, R, M, N.Letter classification 3 comprises: C, G.Letter classification 4 comprises: U, V.Numeral classification 1 comprises: 1,4.Numeral classification 2 comprises: 3,6,9.Letter and number classification 1 comprises: 0, D, Q. letter and number classification 2 comprise: 2, Z.3,8, B the letter and number classification comprises:.Letter and number classification 4 comprises: 5, S.Letter and number classification 5 comprises: 7, T.With the training method of whole sorter, with the good training sample training individuals sorter of classifying, obtain preferable sorter model equally, then characters on license plate to be identified is carried out individual segregation.Mode classification as shown in Figure 2.
Step 3: on the result of secondary classification, its common ground of character picture that will have common ground is got rid of, the extracting section SURF key point feature that character picture is not excluded again, and key point is carried out feature describe, obtain describing the feature descriptor of key point feature.
The overall flow of SURF algorithm is similar with SIFT, but adopted with SIFT diverse ways in the descriptor construction process: it is based on integral image and Hessian matrix, with integral image tectonic scale space, with quick Hessian operator detected characteristics point, go to describe unique point with the SURF descriptor.The SURF algorithm is the improvement to the SIFT algorithm, has brought the significantly raising of calculating under the very little situation of performance loss, thereby in Recognition of License Plate Characters using value is arranged more than SIFT algorithm.
Step 4: the coupling of carrying out the character picture in the class according to the feature descriptor of describing feature.
Carry out the coupling of the character picture in the class according to the feature descriptor of describing feature; Feature descriptor is the proper vector of describing characteristic area.After obtaining describing the proper vector of characteristic area, can adopt the Euclidean distance of proper vector as the similarity measurement of two width of cloth character picture key points.Can adopt based on the estimating of Euclidean space distance, establish X, Y is respectively two two characteristic areas that the eigenvectors representative is mated, wherein
X＝(x
1，…x
n)Y＝(y
1，…y
n)
Then
At first seek the unique point of same position according to classification under the characters on license plate, that is to say the character with same section, the key point that same section extracts does not participate in the key point coupling.Can further increase the difference between the character like this, make recognition effect further improve.Because the key point number that participates in mating is reduced, the time of key point coupling also will obtain shortening, thereby the real-time of algorithm further improves in addition.
Step 5: carry out Recognition of License Plate Characters according to matching result.
When the ratio of the unique point number of the number of Feature Points Matching and the coupling of ginseng reached certain threshold value, this threshold value was desirable 80%, can think two characters on license plate couplings, and then had reached the purpose of Recognition of License Plate Characters.
Claims (4)
1. license plate character recognition method based on SURF characteristic matching algorithm is characterized in that: may further comprise the steps:
(1) at first the characters on license plate image is carried out whole first subseries, the characters on license plate image is divided into Chinese character, letter, numeral and character and digit four big classes;
(2) according to the characters on license plate characteristics, first classification results is carried out individual secondary classification again, make the character picture with more common ground be classified as a class;
(3) at extraction SURF key point feature of all categories, and key point is carried out feature describe;
(4) carry out the coupling of the character picture in the class according to the feature descriptor of describing feature;
(5) carry out Recognition of License Plate Characters according to matching result.
2. according to claim 1 based on the license plate character recognition method of SURF characteristic matching algorithm, it is characterized in that: the first sorting technique of characters on license plate image adopts neural network or SVM clustering algorithm in the described step (1).
Described in claim 1 or 2 based on the license plate character recognition method of SURF characteristic matching algorithm, it is characterized in that: the secondary classification method of characters on license plate image adopts neural network or SVM clustering algorithm in the described step (2).
As described in the claim 3 based on the license plate character recognition method of SURF characteristic matching algorithm, it is characterized in that: elder generation is on the result of secondary classification in the described step (3), its common ground of character picture that will have common ground is got rid of, the extracting section SURF key point feature that character picture is not excluded again, and key point is carried out feature describe.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011104486829A CN103186762A (en) | 2011-12-28 | 2011-12-28 | License plate character recognition method based on SURF matching algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011104486829A CN103186762A (en) | 2011-12-28 | 2011-12-28 | License plate character recognition method based on SURF matching algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103186762A true CN103186762A (en) | 2013-07-03 |
Family
ID=48677925
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2011104486829A Pending CN103186762A (en) | 2011-12-28 | 2011-12-28 | License plate character recognition method based on SURF matching algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103186762A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106127205A (en) * | 2016-06-22 | 2016-11-16 | 山东鲁能智能技术有限公司 | A kind of recognition methods of the digital instrument image being applicable to indoor track machine people |
| CN106709867A (en) * | 2016-11-23 | 2017-05-24 | 电子科技大学 | Medical image registration method based on improved SURF and improved mutual information |
| CN108898137A (en) * | 2018-05-25 | 2018-11-27 | 黄凯 | A kind of natural image character identifying method and system based on deep neural network |
| CN110991467A (en) * | 2019-12-06 | 2020-04-10 | 西安广源机电技术有限公司 | Intelligent license plate recognition method |
| CN111783766A (en) * | 2020-07-10 | 2020-10-16 | 上海淇毓信息科技有限公司 | Method and device for recognizing image characters step by step and electronic equipment |
| CN114863692A (en) * | 2022-04-18 | 2022-08-05 | 广州天长信息技术有限公司 | Vehicle pattern recognition fee evasion checking method based on local feature alignment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101329734A (en) * | 2008-07-31 | 2008-12-24 | 重庆大学 | License plate character recognition method based on K-L transform and LS-SVM |
| CN101807257A (en) * | 2010-05-12 | 2010-08-18 | 上海交通大学 | Method for identifying information of image tag |
| CN101937508A (en) * | 2010-09-30 | 2011-01-05 | 湖南大学 | License plate localization and identification method based on high-definition image |
| US8582889B2 (en) * | 2010-01-08 | 2013-11-12 | Qualcomm Incorporated | Scale space normalization technique for improved feature detection in uniform and non-uniform illumination changes |
-
2011
- 2011-12-28 CN CN2011104486829A patent/CN103186762A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101329734A (en) * | 2008-07-31 | 2008-12-24 | 重庆大学 | License plate character recognition method based on K-L transform and LS-SVM |
| US8582889B2 (en) * | 2010-01-08 | 2013-11-12 | Qualcomm Incorporated | Scale space normalization technique for improved feature detection in uniform and non-uniform illumination changes |
| CN101807257A (en) * | 2010-05-12 | 2010-08-18 | 上海交通大学 | Method for identifying information of image tag |
| CN101937508A (en) * | 2010-09-30 | 2011-01-05 | 湖南大学 | License plate localization and identification method based on high-definition image |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106127205A (en) * | 2016-06-22 | 2016-11-16 | 山东鲁能智能技术有限公司 | A kind of recognition methods of the digital instrument image being applicable to indoor track machine people |
| CN106709867A (en) * | 2016-11-23 | 2017-05-24 | 电子科技大学 | Medical image registration method based on improved SURF and improved mutual information |
| CN108898137A (en) * | 2018-05-25 | 2018-11-27 | 黄凯 | A kind of natural image character identifying method and system based on deep neural network |
| CN108898137B (en) * | 2018-05-25 | 2022-04-12 | 黄凯 | Natural image character recognition method and system based on deep neural network |
| CN110991467A (en) * | 2019-12-06 | 2020-04-10 | 西安广源机电技术有限公司 | Intelligent license plate recognition method |
| CN111783766A (en) * | 2020-07-10 | 2020-10-16 | 上海淇毓信息科技有限公司 | Method and device for recognizing image characters step by step and electronic equipment |
| CN111783766B (en) * | 2020-07-10 | 2023-02-14 | 上海淇毓信息科技有限公司 | Method and device for recognizing image characters step by step and electronic equipment |
| CN114863692A (en) * | 2022-04-18 | 2022-08-05 | 广州天长信息技术有限公司 | Vehicle pattern recognition fee evasion checking method based on local feature alignment |
| CN114863692B (en) * | 2022-04-18 | 2023-10-27 | 广州天长信息技术有限公司 | Vehicle pattern recognition fee evasion checking method based on local feature alignment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110969160B (en) | License plate image correction and recognition method and system based on deep learning | |
| CN100357988C (en) | Identification method of mobile number plate based on three-channel parallel artificial nerve network | |
| CN101937508B (en) | License plate localization and identification method based on high-definition image | |
| CN103186762A (en) | License plate character recognition method based on SURF matching algorithm | |
| CN108090429B (en) | Vehicle type recognition method for graded front face bayonet | |
| CN102968637B (en) | Complicated background image and character division method | |
| Dehghan et al. | View independent vehicle make, model and color recognition using convolutional neural network | |
| CN103971097B (en) | Vehicle license plate recognition method and system based on multiscale stroke models | |
| CN105335702B (en) | A kind of bayonet model recognizing method based on statistical learning | |
| CN105335710A (en) | Fine vehicle model identification method based on multi-stage classifier | |
| CN105184271A (en) | Automatic vehicle detection method based on deep learning | |
| CN104463241A (en) | Vehicle type recognition method in intelligent transportation monitoring system | |
| CN103279756A (en) | Vehicle detecting analysis system and detecting analysis method thereof based on integrated classifier | |
| CN107016362B (en) | Vehicle weight recognition method and system based on vehicle front windshield pasted mark | |
| CN105740809A (en) | Expressway lane line detection method based on onboard camera | |
| Al Faqheri et al. | A real-time Malaysian automatic license plate recognition (M-ALPR) using hybrid fuzzy | |
| CN102880859A (en) | Method for recognizing number plate | |
| CN109033175A (en) | A kind of method and system to scheme to search vehicle | |
| CN101604381A (en) | License plate character recognition method based on multi-category support vector machines | |
| CN103390167A (en) | Multi-characteristic layered traffic sign identification method | |
| CN106778646A (en) | Model recognizing method and device based on convolutional neural networks | |
| CN106919939B (en) | A kind of traffic signboard tracks and identifies method and system | |
| CN102163278A (en) | Illegal vehicle intruding detection method for bus lane | |
| CN103268498A (en) | Method for area-of-interest blurred image semantic comprehension | |
| CN109191492A (en) | A kind of intelligent video black smoke vehicle detection method based on edge analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130703 |