CN102682299A - Character overall feature-based rapid number plate verification method - Google Patents

Abstract

The invention discloses a character overall feature-based rapid number plate verification method, which comprises the following steps of: performing number plate extraction and number plate character division on an input static color vehicle image; calculating a background field assigned value of a number plate character image, and extracting the convex and concave features of the character image; verifying number plate information to be verified by using the convex and concave features of the number plate character image to obtain a verification result of each character; and combining the verification results of the number plate characters by using a decision fusion technology, and judging the accurate of the number plate information to be verified. Direct number plate character recognition is avoided, the image is not required to be complexly preprocessed, and the method is high in recognition rate and recognition speed, and can be reliably used for vehicle identity legality identification.

Description

A kind of quick car plate verification method based on the character global feature

Technical field

The verification method and the system of vehicle legitimacy the present invention relates to quick and precisely to confirm in technical field, the especially intelligent transportation of car plate checking to wait to conclude the business.

Background technology

Electronic charging system without parking (Electronic Toll Collection; ETC) be a kind of electronic toll system that is applied to road, bridge and tunnel of making great efforts popularization in the world; The ETC system can realize the non-parking charge of vehicle, is to improve the highway traffic capacity and solve the block up effective means of problem of charge station.Standard ETC system relies on the microwave communication between vehicle and the charge station to realize automatic vehicle identification, once by the licence plate automatic identification technology of capturing based on image of broad research because its discrimination was low, recognition time is long and be not used.

During the ETC operate as normal, be installed in board units (On-Board Unit, OBU on the automobile; Store the information relevant with testing vehicle register) (Road-Side Unit RSU), carries out the fast data exchange with microwave communication to the roadside unit other with being installed in the lane in which the drivers should pay fees; Carry out disbursement and sattlement through the bank of networking or the stored value card of advance payment in advance again; If Transaction Success, then the electronics railing lifts, and vehicle is through the ETC track.Track surveillance camera record is prepared against inspection in the future through vehicle.When existing two or many to wait to conclude the business vehicle in the communication range of ETC system simultaneously, RSU possibly misread the OBU of other vehicles rather than current vehicle, and Successful Transaction, causes the disorder of ETC system.Wherein be called with car and disturb, be called monkey chatter from the interference of adjacent track vehicle OBU from the interference of back car OBU on the same track.Interference problem is applied the major issue of the middle solution of needing badly for the ETC system.With the essence that car disturbs and monkey chatter exists all is the correctness owing to the vehicle identification that has its dependence microwave communication of other technologies check to realize in the existing ETC system.A kind of feasible solution is: from the vehicle image that track supervision picture pick-up device is captured, extract License Plate Image, use image information that the vehicle identity information that RSU obtains is verified.Vehicle image always derives from current correct vehicle, and whether the car plate data that obtain through checking RSU are consistent with the license plate image of correct vehicle, and just whether the decidable vehicle of waiting to conclude the business is legal.

Existing license plate image verification method adopts the method for direct license plate identification mostly, does not solve the problem of its length consuming time all the time, is not also adopted by the ETC standard.

Summary of the invention

The technical matters that (one) will solve

The objective of the invention is to; Overcome the shortcoming and defect of prior art; The problem of, consuming time length low to direct car plate accuracy of identification, research and development a kind of quick car plate verification method based on the character global feature, the correctness of the vehicle identity information that checking ETC system obtains through RSU; Whether continuous business provides clear and definite signal to RSU and current OBU, solves disturbing and the monkey chatter problem with car of ETC system.Satisfy high precision, the real-time requirement of ETC system simultaneously.

(2) technical scheme

For achieving the above object, the invention provides a kind of quick car plate verification method based on the character global feature, comprising: the static colored vehicle image of input is carried out car plate extracts and characters on license plate is cut apart; Calculate characters on license plate image background field assignment, extract the convex and concave feature of character picture; Use the convex and concave feature of characters on license plate image that license board information to be confirmed is verified, provide the checking result of every character; And the checking result who uses the every characters on license plate of Decision Fusion technical combinations, judge the correctness of license board information to be confirmed.

In the such scheme; Said static colour image to input carries out in car plate extracts and characters on license plate is cut apart the step; Be to utilize the textural characteristics of license plate area and combine gray scale and edge projection information location car plate; Utilize the location of the edge characteristics correction car plate of characters on license plate again, finally obtain the characters on license plate image of binaryzation.

In the such scheme; Said calculating characters on license plate image background field assignment; Extracting in the step of convex and concave feature of character picture, is to utilize license board information to be confirmed to be Given information, chooses digit order number in the number-plate number as discrimination bit; The background dot of different digital calculative ambient field assignment is also different, is obtained the concavo-convex global feature of character picture by the ambient field assignment.

In the such scheme; The convex and concave feature of said use characters on license plate image is verified license board information to be confirmed; Providing in checking result's the step of every character, is to rely on the characters on license plate image whether to have the pairing convex and concave feature of license board information to be confirmed, judges the correctness of characters on license plate to be confirmed; The checking result of every character representes that with consistent, inconsistent and uncertain three judgement hypothesis the assignment size of judgement hypothesis has been represented the degree of support to such court verdict.

5, the quick car plate verification method based on the character global feature according to claim 1; It is characterized in that; The checking result of the every characters on license plate of said use Decision Fusion technical combinations; Judge in the step of correctness of license board information to be confirmed that the Decision Fusion technology that is adopted is the D-S evidence theory.

(3) beneficial effect

The present invention compared with prior art has following advantage and beneficial effect:

1, the quick car plate verification method based on the character global feature provided by the invention has solved with car and has disturbed and the monkey chatter problem.In the communication range of ETC antenna; There are two or many automobiles to be concluded the business simultaneously; Through license plate image to the license plate information that microwave communication obtains verify judge whether it derives from current correct vehicle, whether continuous business provides clear and definite signal to RSU and current OBU; In time detect illegal vehicle, stop illegal transaction.

2, the quick car plate verification method based on the character global feature provided by the invention, vehicle pass-through speed is high.The license board information to be confirmed that obtains with RSU is a Given information, with license plate image it is verified, and is as the Flame Image Process of checking means, much quicker than directly carrying out image recognition; Simultaneously, choose convex and concave feature, need not image is done complicated pretreatment such as refinement, inclination rectification, improved the speed of checking as diagnostic characteristics.When solution ETC is with car interference and monkey chatter problem, the vehicle speed of crossing in ETC track is accelerated greatly.

3, the quick car plate verification method based on the character global feature provided by the invention, the scrap build quantities is little.Adopt the ETC system of image authentication technology; Need not to increase the equipment of standard ETC system, also need not change the RSU and the OBU hardware designs of ETC system, only need the track surveillance equipment be moved forward; Cooperate relevant software, just can solve disturbing and the monkey chatter problem of ETC system conscientiously with car.

4, the quick car plate verification method based on the character global feature provided by the invention, system reliability is high.Adopt the D-S evidence that the checking result of each character is merged, obtain final car plate checking result.The comprehensive multiple information of this court verdict obtains, and than the evidence of any single information higher confidence level is arranged, and has improved the correctness to the testing vehicle register judgement.

Description of drawings

Fig. 1 is the process flow diagram based on the quick car plate verification method of character global feature according to the embodiment of the invention;

Fig. 2 be according to the circle of the numerical character of the embodiment of the invention with about the synoptic diagram that caves in;

Fig. 3 is the synoptic diagram according to the image of the input of the embodiment of the invention;

Fig. 4 is the car plate bianry image that extracts according to the embodiment of the invention;

Fig. 5 is checking decision tree and the court verdict assignment according to the numeral 6 of the embodiment of the invention;

Fig. 6 is the Decision Fusion model according to the characters on license plate checking result of the embodiment of the invention.

Embodiment

For making the object of the invention, technical scheme and advantage clearer, below in conjunction with specific embodiment, and with reference to accompanying drawing, to further explain of the present invention.

Characteristics such as circle in the image and recessed zone are called the convex and concave feature of image, and convex and concave feature is the global feature of character picture.Character picture concavity and convexity feature extracting method comprises following two steps: the assignment background is calculated in (1).Background dot (pixel value is 0 point in the binary picture) to character picture; From this point; To the right, upper right, last, upper left, left a, left side down, down, 8 directions in bottom right send 8 rays, are the assignment of background dot with the crossing number of rays in foreground point (character).(2) extract concavity and convexity.By circle and the sunk area of assignment background in can the process decision chart picture.The background assignment is 8, representes that this background pixel is surrounded by foreground pixel, and the zone at its place is circle.Assignment is that 5,6,7 background area is the depressed area, and wherein the right of each background dot all can find the depression of foreground point to be called left side depression; The left side of each background dot all can find the depression of foreground point to be called right depression.

The circle of 10 numerals with about the depression as shown in Figure 2.Black part is divided into circle, and horizontal line is filled to left depressed area, and perpendicular line is filled to right depressed area.Choose behind the licence plate numeral in 5 as the foundation of license plate validation.The license board information to be confirmed that obtains with RSU is verified it with correspondence position licence plate character picture as Given information from right to left.Whether have the concavo-convex characteristic identical according to the licence plate character picture and provide the differentiation result, as the input of next step decision networks with known characters on license plate.

Quick car plate verification method based on the character global feature provided by the invention be to choose the foundation that the digit order number in the licence plate character is differentiated as licence plate, and the concavo-convex characteristic of selection digital picture is as its diagnostic characteristics.Because digit recognition is typical small size character set identification, with respect to the identification of Chinese character with letter, digit recognition is simple much accurate.The convex and concave feature of character is the global feature of character simultaneously; Choose convex and concave feature as diagnostic characteristics; Need not image is done complicated pretreatment such as refinement, inclination rectification, the image of same character is even shape bit is equipped with certain deviation; Still have identical global feature, thereby for realizing that license plate validation provides theoretical foundation quickly and accurately.

Fig. 1 is the process flow diagram of the method for detecting phases based on the character global feature provided by the invention, and this method may further comprise the steps:

Step 101: the static colored vehicle image of input carried out car plate extracts and characters on license plate is cut apart;

Step 102: calculate characters on license plate image background field assignment, extract the convex and concave feature of character picture;

Step 103: use the convex and concave feature of characters on license plate image that license board information to be confirmed is verified, provide the checking result of every character;

Step 104: use the checking result of the every characters on license plate of Decision Fusion technical combinations, judge the correctness of license board information to be confirmed.

Embodiment below in conjunction with concrete describes.

Step 1: car plate extracts and characters on license plate is cut apart

This step is to carry out the preceding processing of car plate checking, comprises original image gray processing, edge extracting, car plate location, Character segmentation several steps.Utilize textural characteristics other and the combination gray scale and the edge projection information location car plate of license plate area, and utilize the location of the edge characteristics correction car plate of characters on license plate, finally obtain the characters on license plate image of binaryzation.Need not to image do refinement, complicated pretreatment such as the rectification of tilting.Fig. 3 is an original image, the binaryzation car plate of Fig. 4 for from original image, extracting.

Step 2: calculate the ambient field assignment.

Choose behind the licence plate numeral in 5 as the foundation of license plate validation.The license board information to be confirmed that obtains with RSU is calculated the ambient field assignment of correspondence position licence plate character picture part background dot from right to left as Given information.The ambient field assignment of calculating all background dots is still more consuming time, only needs the characteristics according to different digital in the reality, selects the most representative position to carry out the ambient field assignment and calculates.Like numeral 6 checking, whether have 6 corresponding circle and recess feature on the 1/4 height level's line (Pos1) that only need search corresponding character picture and the 3/4 height level's line (Pos2), be specially at image whether circle is arranged on Pos1, whether right the depression is arranged on Pos2.

Step 3: provide character checking result and confidence level.

Each numeral of ten numerals of 0-9 is set up a discrimination tree, and the ambient field assignment of calculating according to step 2 and the convex and concave feature of extraction obtain to characters on license plate image and RSU that license board information is whether consistent to provide the checking result, and provide this result's credit assignment.Consistance court verdict hypothesis is represented with 3 set; Promptly to represent to adjudicate image character consistent with Given information for

A; Both are inconsistent in

expression judgement, and expression is uncertain to court verdict.The degree that the size representative of judgement hypothesis assignment is supported this hypothesis is used for the input of Decision Fusion network.The decision tree of numeral 6 is as shown in Figure 5.Divide two kinds of situation that consistance judgement hypothesis is carried out assignment:

1): court verdict is positioned at lobus sinister of decision tree, and it is consistent promptly to judge both.This moment, the character number of not getting rid of (comprising the alphabetic character with identical convex and concave feature) was confirmed the court verdict assignment of corresponding hypothesis according to the digital number of having got rid of.

$m{\left(A\right)}^{\′}=\frac{X-Y+1}{10+Z}.---\left(1\right)$

$m{\left(\stackrel{\‾}{A}\right)}^{\′}=\frac{Y+Z-1}{10+Z}.---\left(2\right)$

$m{\left(\right\{A,\stackrel{\‾}{A}\left\}\right)}^{\′}=\frac{Y+Z}{10+Z}.---\left(3\right)$

Wherein X representes to get rid of the number of numeral, and Y representes not get rid of the number of numeral, and Z representes to satisfy the alphabetical number of possibility of corresponding convex and concave feature.Again with m (A) ';

carries out normalization, promptly obtains the degree of belief assignment of corresponding judgement hypothesis.As in the decision tree of character 6, be consistent if judge both, the numeral (0 that got rid of this moment; 1; 2,3,4; 7; 8,9) number is 8, does not get rid of numeral (5; 6) number is 2; Possible letter (G) number is 1, by formula (1)-(3), and m (A) '=0.64;

normalization gets m (A)=0.58,

$m\left(\right\{A,\stackrel{\‾}{A}\left\}\right)=0.25.$

2): court verdict is positioned at lobus dexter of decision tree, judges that promptly both are inconsistent.Carry out assignment in the number of plies of decision tree to the judgement hypothesis according to court verdict this moment.

m(A)＝0.05. (4)

$m{\left(\stackrel{\‾}{A}\right)}_k=0.95-0.05\·K.---\left(5\right)$

$m{\left(\right\{A,\stackrel{\‾}{A}\left\}\right)}_k=0.05\·K.---\left(6\right)$

Wherein K is judgement node number of plies in decision tree, and the regulation root node is at the 0th layer.As in the decision tree of character 6, suppose to judge that both are inconsistent, this moment, court verdict was positioned at lobus dexter of decision tree.When the lobus dexter period of the day from 11 p.m. to 1 a.m that is ground floor; The lower part Wu Quan of character picture; By formula (4)-(6); M (A)=0.05; When the lobus dexter period of the day from 11 p.m. to 1 a.m that is the second layer, there is circle the character picture lower part to

, but does not cave in top; Court verdict is still for negating still uncertain degree increase.By formula (4)-(6); M (A)=0.05; Other digital differentiation of

is similar with numeral 6, repeats no more.

Step 4: Decision Fusion, the legitimacy of differentiation vehicle.The rule of combination that adopts the D-S evidence theory to provide; The checking result of every car plate numerical character that step 3 is obtained carries out Decision Fusion; Select the maximum hypothesis of support as final judging result according to certain decision rule; Judge whether the vehicle identity information that RSU obtains derives from current correct vehicle, whether continuous business provides clear and definite signal to RSU and current OBU.The comprehensive multidigit discriminative information of this court verdict obtains, and than the evidence of any single information higher confidence level is arranged, and has improved the correctness to the testing vehicle register judgement.Fig. 6 is the Decision Fusion model.

The Decision Fusion technology that is adopted in embodiments of the present invention is the D-S evidence theory, is used for a plurality of licence plate numerical character checking results are merged.The D-S evidence theory is a kind of Theory of Information Fusion based on uncertain inference; Can describe and hold the not intellectual and uncertainty of problem than traditional probability theory better; Information a large amount of different aspects, uncertain are converted into the deterministic result of decision, thereby are widely used in the uncertainty decision-making is carried out in a plurality of evidence combinations.

The D-S evidence theory provides a very useful combinatorial formula, can synthesize the court verdict that a plurality of evidences source provides.For

The court verdict assignment m that n evidence source on the judgement framework θ provides ₁, m ₂..., m _nThe Dempster rule of combination do

$(m_1\⊕m_2\⊕\·\·\·\⊕m_n)\left(A\right)=$

$\left\{\begin{array}{c}0,A=\mathrm{\Φ};\\ {(1-k)}^{-1}\underset{A_1\∩\·\·\·\∩A_n=A}{\mathrm{\Σ}}{m}_1\left(A_1\right)m_2\left(A_2\right)\·\·\·m_n\left(A_n\right),A\≠\mathrm{\Φ}.\end{array}\right.---\left(7\right)$

Wherein

$k=\underset{A_1\∩\·\·\·\∩A_n\mathrm{\Φ}}{\mathrm{\Σ}}{m}_1\left(A_1\right)m_2\left(A_2\right)\·\·\·m_n\left(A_n\right).---\left(8\right)$

K is called the contradiction factor, has reflected the conflict spectrum between each evidence.

The license board information of supposing the RSU acquisition is for { capital NHT680}, the vehicle image that ETC track supervision picture pick-up device obtains is as shown in Figure 3, adopts the quick car plate verification method and the D-S fusion results of the present invention's proposition as shown in table 1.According to fusion results, judge that the license board information of vehicle image car plate and RSU acquisition is inconsistent, vehicle to be concluded the business is illegal vehicle.

Checking of licence plate character and the comprehensive court verdict of D-S evidence under the inconsistent situation of table 1

Above-described specific embodiment; The object of the invention, technical scheme and beneficial effect have been carried out further explain, and institute it should be understood that the above is merely specific embodiment of the present invention; Be not limited to the present invention; All within spirit of the present invention and principle, any modification of being made, be equal to replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (5)

1. the quick car plate verification method based on the character global feature is characterized in that, comprising:

The static colored vehicle image of input carried out car plate extracts and characters on license plate is cut apart;

Calculate characters on license plate image background field assignment, extract the convex and concave feature of character picture;

Use the convex and concave feature of characters on license plate image that license board information to be confirmed is verified, provide the checking result of every character; And.

Use the checking result of the every characters on license plate of Decision Fusion technical combinations, judge the correctness of license board information to be confirmed.

2. the quick car plate verification method based on the character global feature according to claim 1; It is characterized in that; Said static colour image to input carries out in car plate extracts and characters on license plate is cut apart the step; Be to utilize the textural characteristics of license plate area and combine gray scale and edge projection information location car plate, utilize the location of the edge characteristics correction car plate of characters on license plate again, finally obtain the characters on license plate image of binaryzation.

3. the quick car plate verification method based on the character global feature according to claim 1; It is characterized in that said calculating characters on license plate image background field assignment is in the step of the convex and concave feature of extraction character picture; Be to utilize license board information to be confirmed to be Given information; Choose digit order number in the number-plate number as discrimination bit, the background dot of different digital calculative ambient field assignment is also different, is obtained the concavo-convex global feature of character picture by the ambient field assignment.

4. the quick car plate verification method based on the character global feature according to claim 1; It is characterized in that; The convex and concave feature of said use characters on license plate image is verified license board information to be confirmed; Providing in checking result's the step of every character, is to rely on the characters on license plate image whether to have the pairing convex and concave feature of license board information to be confirmed, judges the correctness of characters on license plate to be confirmed; The checking result of every character representes that with consistent, inconsistent and uncertain three judgement hypothesis the assignment size of judgement hypothesis has been represented the degree of support to such court verdict.

5. the quick car plate verification method based on the character global feature according to claim 1; It is characterized in that; The checking result of the every characters on license plate of said use Decision Fusion technical combinations; Judge in the step of correctness of license board information to be confirmed that the Decision Fusion technology that is adopted is the D-S evidence theory.

Images