CN104318233A - Method for horizontal tilt correction of number plate image - Google Patents
Method for horizontal tilt correction of number plate image Download PDFInfo
- Publication number
- CN104318233A CN104318233A CN201410561462.0A CN201410561462A CN104318233A CN 104318233 A CN104318233 A CN 104318233A CN 201410561462 A CN201410561462 A CN 201410561462A CN 104318233 A CN104318233 A CN 104318233A
- Authority
- CN
- China
- Prior art keywords
- connected domain
- character
- straight line
- license plate
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000012937 correction Methods 0.000 title abstract description 15
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 230000015572 biosynthetic process Effects 0.000 claims description 10
- 238000009826 distribution Methods 0.000 claims description 6
- 238000007621 cluster analysis Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 2
- 230000011218 segmentation Effects 0.000 abstract description 3
- 238000005516 engineering process Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/146—Aligning or centring of the image pick-up or image-field
- G06V30/1475—Inclination or skew detection or correction of characters or of image to be recognised
- G06V30/1478—Inclination or skew detection or correction of characters or of image to be recognised of characters or characters lines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/158—Segmentation of character regions using character size, text spacings or pitch estimation
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Character Input (AREA)
Abstract
The invention discloses a method for horizontal tilt correction of a number plate image. Due to the reasons of a shooting angle and the like of the number plate image, a number plate is tilted so that character segmentation is not facilitated, and the number plate needs to be corrected. After a series of image enhancement operations, obtained number plate edge images are good in quality; based on the edge images, the character law is utilized to search a legal character connected domain of the edge images, linear fitting is performed on an upper endpoint and a lower endpoint of the connected domain so as to find out the upper boundary and the lower boundary of a character zone, the slope is obtained, the tilt correction rotation angle is calculated, and the number plate is rotated. A removing method of a pseudo connected domain and a connected domain searching method are improved, and the robustness of the tilt correction method is greatly improved.
Description
Technical field
The present invention relates to a kind of method realized based on character connected domain in license plate image, particularly relate to a kind of license plate image horizontal tilt bearing calibration.
Background technology
In intelligent transportation system (ITS, Intelligent transportation system), automatic Car license recognition (LPR, LicensePlateRecognition) plays important role.Usually, LPR is made up of following step: Image Acquisition, License Plate, License Plate Character Segmentation and identification.But due to reasons such as shooting angle, in image, car plate there will be the problem of inclination, as Fig. 1.This can cause the problem such as connection and destruction between characters on license plate, is unfavorable for Character segmentation and identification.Therefore, the importance of license plate image slant correction is apparent.
More existing bearing calibrations are generally by trying to achieve the frame straight line of car plate at present, then calculate pitch angle.But car plate has its complicacy, on the one hand presumable car plate frame off-gauge straight line, the image that location algorithm obtains on the other hand not has frame.In addition, also have some correcting algorithms, or be exactly calculate consuming time and corrected rate is low, or be exactly by image background interference greatly.Given this, the present invention is from the characteristic of character itself, i.e. character marshalling in car plate, upper and lower end points is respectively on two straight lines.Be optimized research for the slant correction algorithm based on character connected domain, mainly improve the minimizing technology of pseudo-connected domain and the lookup method of connected domain, the robustness of slant correction algorithm is improved, and is successfully applied in the Car license recognition in the multiple state of the U.S..
Summary of the invention
In view of above problem, the present invention proposes a kind of license plate image horizontal tilt bearing calibration.
Concrete scheme of the present invention is:
A kind of license plate image horizontal tilt bearing calibration comprises the steps:
1) on car plate edge image, search all connected domains, connected domain comprises positional information, connected domain is mapped on original license plate image;
2) from character self character, from step 1) search believable character connected domain the connected domain that checks in, search the method for believable character connected domain for first searching believable character connected domain based on edge image, when the quantity of the credible character connected domain found based on edge image is more than or equal to 3, carry out subsequent step; When the quantity of the credible character connected domain found based on edge image is less than 3, turns to and search believable character connected domain based on bianry image;
3) utilize least square method respectively up-and-down boundary point to be fitted to straight line, and select character zone up-and-down boundary point accurately;
4) try to achieve the slope of character zone up-and-down boundary straight line respectively, obtain the anglec of rotation, and the position of the up-and-down boundary of character zone after calculating rotation according to the anglec of rotation.
Described character property is character marshalling in car plate, and upper and lower end points is respectively on two straight lines.
Described search believable character connected domain based on edge image and be specially:
A) judge whether the height of two connected domains, horizontal level and spacing are less than setting threshold value, two connected domains being less than setting threshold value are merged;
B) judge the character area of connected domain, character depth-width ratio, delete character area, character depth-width ratio do not belong to the connected domain of setting range;
C) according to characters on license plate all integrated distribution this features in certain altitude range, isolated connected domain is deleted;
D) to be substantially consistent this characteristic according to the character height of characters on license plate own, connected domain height average is obtained by cluster analysis, if the difference of certain connected domain height and this mean value is greater than setting threshold value, then delete this connected domain, otherwise retain this connected domain;
The slope of the straight line e) formed according to connected domain central point between two, calculates the absolute value of difference between these slopes, according to absolute difference size, deletes the connected domain that two ends do not satisfy condition;
F) according to the spacing of connected domain, the connected domain being greater than 3 times of connected domain mean breadths is considered as the connected domain peeled off, and deletes the connected domain peeled off.
Described step c) be specially: setting straight line L, scan car plate edge image from top to bottom, find straight line L ', make its number of passing connected domain maximum, delete coboundary L ' below or lower boundary in the connected domain of more than L '.
Described step e) be specially: first straight slope formed with the center of penultimate connected domain is a, the straight slope of second and last connected domain central point formation is b, calculate abs (a-b) if be less than setting threshold value, be then considered as satisfying condition; If be greater than setting threshold value, just think that two ends have a connected domain to need to delete, then the straight slope c of first and last connected domain central point formation is calculated, the relatively size of abs (a-c) and abs (b-c), if abs (a-c) >abs (b-c), then delete last connected domain, otherwise, delete first connected domain; Step e is repeated to remaining connected domain) judgement, until satisfy condition.
Described search believable character connected domain method based on bianry image and be specially:
A) in bianry image, all connected domains are searched;
B) judge whether the height of two connected domains, horizontal level and spacing are less than setting threshold value, two connected domains being less than setting threshold value are merged;
C) judge the character area of connected domain, character depth-width ratio, delete character area, character depth-width ratio do not belong to the connected domain of setting range;
D) according to characters on license plate all integrated distribution this features in certain altitude range, setting straight line H, scans car plate bianry image from top to bottom, finds straight line H ', make its number of passing connected domain maximum, delete coboundary H ' below or lower boundary in the connected domain of more than H ';
E) to be substantially consistent this characteristic according to the character height of characters on license plate own, connected domain height average is obtained by cluster analysis, if the difference of connected domain height and this mean value is greater than setting threshold value, then delete this connected domain, otherwise retain this connected domain;
The slope of the straight line F) formed according to connected domain central point between two, calculates the absolute value of difference between these slopes, according to absolute difference size, deletes the connected domain that two ends do not satisfy condition;
G) according to the spacing of connected domain, the connected domain being greater than 3 times of connected domain mean breadths is considered as the connected domain peeled off, and deletes the connected domain peeled off.
H) if the credible character connected domain quantity found in bianry image satisfies condition, continue to perform subsequent step; Otherwise be judged as not containing license board information in license plate image, and terminate bearing calibration.
Described step F) be specially: first straight slope a formed with the center of penultimate connected domain, the straight slope b of second and last connected domain central point formation, calculate abs (a-b) if be less than setting threshold value, be then considered as satisfying condition; If be greater than setting threshold value, just think that two ends have a connected domain to need to delete, then the straight slope c of first and last connected domain central point formation is calculated, the relatively size of abs (a-c) and abs (b-c), if abs (a-c) >abs (b-c), then delete last connected domain, otherwise, delete first connected domain; Step F is repeated to remaining connected domain) judgement, until satisfy condition.
Described step 3) be specially: the highs and lows finding each connected domain, these points are considered as the up-and-down boundary point of connected domain, by least square method, linear fit is carried out respectively to all up-and-down boundary points, judge up-and-down boundary point middle distance separately fitting a straight line point farthest whether be less than or equal to threshold value from the distance of fitting a straight line, if be less than or equal to threshold value, then continue subsequent step, if be greater than threshold value, then delete this point farthest, and again least-squares algorithm linear fitting is carried out to the point of remainder, until point farthest after matching is less than or equal to threshold value apart from the distance of fitting a straight line, finally obtain up-and-down boundary straight line.
Described step 4) in obtain the anglec of rotation and be specially: if the absolute value of the difference of the slope of up-and-down boundary straight line is less than setting threshold value, so the slope of up-and-down boundary straight line is added and averages, obtain the anglec of rotation of car plate; Otherwise select the slope of the maximum fitting a straight line of the number of reservation frontier point to calculate the anglec of rotation of car plate.
Described calculate rotation according to the anglec of rotation after the position of up-and-down boundary of character zone be specially:
The computing method of position, coboundary choose to be positioned at straight line upper end, coboundary and the nearest point of distance fitting a straight line, according to the anglec of rotation obtained, calculates this and put position in image after rotation;
Lower boundary position calculating method chooses to be positioned at lower boundary straight-line lower end and the nearest point of distance fitting a straight line, according to the anglec of rotation obtained, calculates this and put position in image after rotation;
Its computing formula is as follows:
x＝Xcosα-Ysinα
y＝Xsinα+Ycosα
Wherein (X, Y) is the coordinate before the point nearest apart from fitting a straight line rotates, and (x, y) is the coordinate after the point nearest apart from fitting a straight line rotates, and α is the anglec of rotation.
Innovative point of the present invention is the characteristic from character itself, i.e. character marshalling in car plate, and upper and lower end points is respectively on two straight lines.By searching the up-and-down boundary point of each character connected domain, utilize least square method respectively up-and-down boundary point to be fitted to two straight lines, the slope of straight line can reflect license plate sloped angle.In addition, present invention improves over the place to go method of pseudo-connected domain and the lookup method of connected domain, substantially increase the robustness of slant correction algorithm, and be successfully applied in the Car license recognition in the multiple state of the U.S..
Accompanying drawing explanation
Fig. 1 (a) is model during license plate image horizontal direction inclined angle alpha >0;
Fig. 1 (b) is model during license plate image horizontal direction inclined angle alpha <0;
Fig. 2 (a) is model during license plate image vertical direction tiltangleθ >0;
Fig. 2 (b) is model during license plate image vertical direction tiltangleθ <0;
Figure 3 shows that top boundary point and the bottom boundary point of character zone in car plate;
Figure 4 shows that and use least square method to carry out fitting a straight line to character top in car plate and bottom boundary point respectively;
Fig. 5 (a) is license plate grey level image G1;
Fig. 5 (b) is the license plate image after histogram equalization;
Fig. 5 (c) for gaussian filtering level and smooth after license plate image;
Fig. 5 (d) is the license plate image G2 after laplacian spectral radius;
Fig. 5 (e) license plate binary image B1;
Fig. 5 (f) is car plate Canny edge-detected image C1;
Fig. 6 (a) is mapped to original image for connected domain;
Connected domain in Fig. 6 (b) character " X " is merged;
Fig. 6 (c) is illegal, and connected domain is deleted;
Fig. 6 (d) for the connected domain in character " 8 " deleted;
Fig. 6 (e) for the connected domain in character " H ", " 2 " deleted;
Fig. 6 (f) deletes the license plate image before last connected domain;
Fig. 6 (g) deletes the license plate image after last connected domain;
Fig. 6 (h) deletes rightmost and to peel off the license plate image before connected domain;
Fig. 6 (i) deletes rightmost and to peel off the license plate image after connected domain;
Fig. 7 (a) is car plate edge image; Fig. 7 (b) is that the connected domain detected based on edge image is mapped to original image;
Fig. 7 (c) maps on the original image for the legal connected domains of two of detecting;
Fig. 7 (d) car plate bianry image;
Fig. 7 (e) is that the connected domain detected based on bianry image is mapped to original image;
Fig. 7 (f) is mapped on original image for the legal connected domain detected;
Fig. 8 (a) is for carry out the license plate image after linear fit by least square method respectively to the highs and lows of each connected domain;
License plate image after the point that Fig. 8 (b) is delete character " 6 " top and bottom;
Fig. 9 (a) is for carrying out the license plate image after linear fit by least square method respectively to residue up-and-down boundary point;
Fig. 9 (b) is the image after car plate horizontal tilt correction;
Figure 10 shows that image level slant correction algorithm flow chart.
Embodiment
There are two kinds of situations in the direction that license plate area tilts: horizontal tilt, as Fig. 1; Vertical bank, as Fig. 2.In Fig. 1, only have tilting zone axis X ' and horizontal axis X between angle α be unknown, once try to achieve angle α, whole license plate image only needs rotation-α, just can correct.In Fig. 2, only have the angle theta in vertical direction unknown.In actual applications, there is the inclination on two kinds of directions in most of car plate simultaneously.Therefore, the inclination first on level of corrections direction, then correct the inclination in vertical direction.Algorithm in the present invention only relates to the correction in horizontal direction.
In car plate, the color of character is the same or close, so in the bianry image of car plate, a complete independently character will become a connected domain.On the other hand, owing to there is relatively high contrast between character color and background color, the edge of character can be obtained by Boundary extracting algorithm, as Canny algorithm.Complete character edge is also a connected domain.Some strategies can be utilized, the connected domain of only reserved character.Mention above, in license plate image, the top boundary point of character zone and bottom boundary point (being called unique point) can reflect the slope trend of character zone, as Fig. 3.After the top boundary point obtaining these character zones and bottom boundary point, respectively fitting a straight line is carried out to top boundary point and bottom boundary point by least square method, thus can in the hope of the pitch angle in licence plate horizontal direction, as Fig. 4.Therefore, license plate sloped angle information can be obtained from the trend of these some distributions.
To process license plate grey level image G1 (see Fig. 5 (a)) for example, set forth specific embodiments.Before slant correction, need to carry out some pretreatment operation to license plate grey level image.Pretreatment operation comprises histogram equalization (see Fig. 5 (b)), with Gaussian filter (size 5x5, δ=1.1) filtering (see Fig. 5 (c)), the laplacian spectral radius then carried out based on second-order differential operates (see Fig. 5 (d)).After three step pretreatment operation, obtain a secondary license plate grey level image G2.Observe and find, gray level image G2 has more or less resisted noise, and marginal information wherein have also been obtained enhancing.Finally, binary image B1 (see Fig. 5 (e)) and Canny edge-detected image C1 (see Fig. 5 (f)) is obtained.In B1, character is white, and background is black.Equally, in C1, edge is white, and background is black.
It is as follows that license plate image horizontal tilt corrects concrete steps:
(1) on car plate edge image, search all connected domains, connected domain contains positional information, connected domain is mapped on original image, as shown in Fig. 6 (a).
(2) search believable character connected domain based on edge image, mainly contain following step:
1) the character connected domain be separated is merged, because some character is as during " M ", " X ", " H " may occur
Between fracture be divided into two parts, in order to retain connected domain as much as possible, need the character connected domain to these may be separated to merge, combination principle judges that whether two connected domains meet highly close, coboundary and be less than certain threshold value in same level position and spacing.As shown in Fig. 6 (b), the connected domain in character " X " is merged.
2) delete illegal connected domain, by some prioris, such as character area, character the ratio of width to height delete the connected domain do not satisfied condition, and obtain Fig. 6 (c), can find out that most of illegal connected domain is deleted.
3) isolated connected domain is deleted, characters on license plate be distributed with certain features, that be exactly characters on license plate all integrated distribution in certain altitude range, therefore as shown in Fig. 6 (c), set a red line L, scan image finds through that maximum position of L connected domain number from top to bottom, deletes those coboundaries in below L or the lower boundary connected domain at more than L, as Fig. 6 (d), the connected domain in character " 8 " is deleted.
4) highly illegal connected domain is deleted, the height of characters on license plate is consistent substantially, connected domain height average avgHeight is obtained by cluster analysis, deletion and avgHeight differ larger connected domain, as shown in Fig. 6 (e), the connected domain in character " H ", " 2 " is deleted.
5) connected domain that two ends do not satisfy condition is deleted, condition is first straight slope a formed with the center of penultimate connected domain, the straight slope b of second and last connected domain central point formation, calculate abs (a-b) if be greater than some threshold values, just think that two ends have a connected domain to need to delete, then the straight slope c of first and last connected domain central point formation is calculated, the relatively size of abs (a-c) and abs (b-c), if abs (a-c) >abs (b-c), then delete last connected domain, otherwise, delete first connected domain.As deleted in last connected domain of Fig. 6 (f), obtain Fig. 6 (g).
6) connected domain peeled off is deleted, by judging the spacing of connected domain, if the connected domain mean breadth avgWidth being greater than 3 times illustrates it is outlier, as shown in Fig. 6 (h), there is the rightmost connected domain that peels off, delete and obtain Fig. 6 (i).
(3) believable character connected domain is searched based on bianry image.In some cases, the effect of Canny rim detection is also bad, as shown in Fig. 7 (a), (b), the legal connected domain number that finds may be caused little, even do not have, as shown in Fig. 7 (c), only have the connected domain that two legal, cannot carry out slant correction, at this moment we can turn to the connected domain of searching bianry image.Therefore, when the legal connected domain number of Canny edge image is less than 3, directly turn to the method with searching believable character connected domain described in step (2) in bianry image, search credible connected domain, and Automatic adjusument binary-state threshold, make the multiple that it is arranging threshold value be respectively { 0.3,0.5,0.6,0.8,1.5,1.7,1.9, this scope inner conversion of 2.1}, until the legal connected domain quantity found in bianry image satisfies condition, continues to perform subsequent step; Otherwise be judged as not containing license board information in image, and terminate correcting algorithm.If Fig. 7 (d) is bianry image, all connected domains are as shown in Fig. 7 (e), and finally remaining legal connected domain has 3, satisfies condition, as shown in Fig. 7 (f).
(4) calculate the anglec of rotation of slant correction, find the up-and-down boundary of character zone.As shown in Fig. 8 (a), find the highs and lows of each connected domain, by least square method, linear fit is carried out respectively to the point of all top and bottom, and delete from fitting a straight line point farthest, until this maximum distance is less than some threshold values, stop deleting.As shown in Fig. 8 (b), the point of the top and bottom of delete character " 6 ".Wherein the lookup method of connected domain up-and-down boundary point is as follows:
2) suppose that first character connected domain is expressed as Rs (x0, y0, w0, h0), wherein (x0, y0) represents the coordinate of connected domain Rs upper left side point, and (w0, h0) represents that connected domain Rs's is wide and high; Suppose that last connected domain is expressed as Re (x1, y1, w1, h1), wherein (x1, y1) represents the coordinate of connected domain Re upper left side point, and (w1, h1) represents that connected domain Re's is wide and high.Select some P0 (x, y), wherein x=x0, a y=y0+h0*0.1 at connected domain Rs left margin, and select a some P1 (x ', y ') at connected domain Re left margin, wherein x '=x1, y '=y1+h1*0.1.Therefore a coboundary scounting line Lu can be obtained, as shown in Fig. 8 (a) by these 2.Analogously, P2 (x2, y2), wherein x2=x0, y2=y0+h1*0 is selected at 1 at connected domain Rs left margin, 9; And select a some P3 (x2 ', y2 ') at connected domain Re left margin, wherein x2 '=x0, y2 '=y0+h1*0.9.Therefore can obtain a lower boundary by these 2 and search for first Lb, as shown in Fig. 8 (a).
2) each white point on Lu and Lb is followed the tracks of respectively.Postulated point p is a white point on Lu, and upwards search can obtain the peak ph of and the same connected domain of some p.Therefore, the peak of each connected domain can be searched for by Lu and obtain.All these points can form a coboundary point set, are expressed as Sup, as shown in Fig. 8 (b).Similarly, the minimum point of each connected domain can search for acquisition downwards by Lb, and all these points are configured to lower boundary point set, are expressed as Sbtm, see Fig. 8 (b).
3) point do not satisfied condition in point set Sup and Sbtm is deleted.As shown in Fig. 8 (b), some points do not belong to any character connected domain, but meet the condition of above-mentioned frontier point.In addition, the fracture of character also can have influence on searching of frontier point.For ensureing that boundary line slope accurately can reflect the tilt angle information of car plate, these points not belonging to any character connected domain must be deleted.The concrete steps of deleting these points are as follows: i) according to least square method, with these points of straight line L1 matching; Ii) mean value and variance that these points arrive straight line L1 distance is calculated; Iii) when the mean value of distance is greater than a certain threshold value, those are deleted to L1 apart from maximum point; Iv) execution step I is returned), until mean value is less than a certain threshold value.Therefore, remaining point is qualified up-and-down boundary point.
4) utilize least square method that point remaining in Sup and Sbtm is fitted to straight line, and select character zone up-and-down boundary point accurately.Up-and-down boundary point is fitted to straight line L1 and L2, respectively as shown in Fig. 9 (a).The straight slope of L1 and L2 can be calculated, be expressed as ku and kb.If the absolute value of the difference of upper and lower two slopes is less than some threshold values, so upper and lower two slopes are added and average, obtain the anglec of rotation of car plate, otherwise by the condition such as number, slope size comparing retention point select preferably slope to calculate the anglec of rotation of car plate.While obtaining the anglec of rotation, also to obtain the up-and-down boundary in characters on license plate region, the method that obtains of coboundary chooses to be positioned at fitting a straight line upper end and the point nearest apart from fitting a straight line, then according to the anglec of rotation obtained, calculate this and put coordinate in image after rotation, this value is exactly the value of coboundary, the method that obtains of same lower boundary chooses to be positioned at fitting a straight line lower end and the point nearest apart from fitting a straight line, then according to the anglec of rotation obtained, calculate this and put coordinate in image after rotation, this value is exactly the value of lower boundary.Its computing formula is as follows:
x＝Xcosα-Ysinα
y＝Xsinα+Ycosα
Wherein (X, Y) is the coordinate before frontier point rotation, and (x, y) is the coordinate after frontier point rotates, and α is the anglec of rotation.
(5) after obtaining the anglec of rotation, according to this angle, car plate is rotated, also obtain the up-and-down boundary of character zone, simultaneously as shown in Fig. 9 (b).
Claims (10)
1. a license plate image horizontal tilt bearing calibration, is characterized in that, comprises the steps:
1) on car plate edge image, search all connected domains, connected domain comprises positional information, connected domain is mapped on original license plate image;
2) from character self character, from step 1) search believable character connected domain the connected domain that checks in, search the method for believable character connected domain for first searching believable character connected domain based on edge image, when the quantity of the credible character connected domain found based on edge image is more than or equal to 3, carry out subsequent step; When the quantity of the credible character connected domain found based on edge image is less than 3, turns to and search believable character connected domain based on bianry image;
3) utilize least square method respectively up-and-down boundary point to be fitted to straight line, and select character zone up-and-down boundary point accurately;
4) try to achieve the slope of character zone up-and-down boundary straight line respectively, obtain the anglec of rotation, and the position of the up-and-down boundary of character zone after calculating rotation according to the anglec of rotation.
2. license plate image horizontal tilt bearing calibration as claimed in claim 1, it is characterized in that described character property is character marshalling in car plate, upper and lower end points is respectively on two straight lines.
3. license plate image horizontal tilt bearing calibration as claimed in claim 1, is characterized in that described search believable character connected domain based on edge image and being specially:
A) judge whether the height of two connected domains, horizontal level and spacing are less than setting threshold value, two connected domains being less than setting threshold value are merged;
B) judge the character area of connected domain, character depth-width ratio, delete character area, character depth-width ratio do not belong to the connected domain of setting range;
C) according to characters on license plate all integrated distribution this features in certain altitude range, isolated connected domain is deleted;
D) to be substantially consistent this characteristic according to the character height of characters on license plate own, connected domain height average is obtained by cluster analysis, if the difference of certain connected domain height and this mean value is greater than setting threshold value, then delete this connected domain, otherwise retain this connected domain;
The slope of the straight line e) formed according to connected domain central point between two, calculates the absolute value of difference between these slopes, according to absolute difference size, deletes the connected domain that two ends do not satisfy condition;
F) according to the spacing of connected domain, the connected domain being greater than 3 times of connected domain mean breadths is considered as the connected domain peeled off, and deletes the connected domain peeled off.
4. license plate image horizontal tilt bearing calibration as claimed in claim 3, it is characterized in that described step c) be specially: setting straight line L, scan car plate edge image from top to bottom, find straight line L ', make its number of passing connected domain maximum, delete coboundary L ' below or lower boundary in the connected domain of more than L '.
5. license plate image horizontal tilt bearing calibration as claimed in claim 3, it is characterized in that described step e) be specially: first straight slope formed with the center of penultimate connected domain is a, the straight slope of second and last connected domain central point formation is b, calculate abs (a-b) if be less than setting threshold value, be then considered as satisfying condition; If be greater than setting threshold value, just think that two ends have a connected domain to need to delete, then the straight slope c of first and last connected domain central point formation is calculated, the relatively size of abs (a-c) and abs (b-c), if abs (a-c) >abs (b-c), then delete last connected domain, otherwise, delete first connected domain; Step e is repeated to remaining connected domain) judgement, until satisfy condition.
6. license plate image horizontal tilt bearing calibration as claimed in claim 2, is characterized in that described search believable character connected domain method based on bianry image and being specially:
A) in bianry image, all connected domains are searched;
B) judge whether the height of two connected domains, horizontal level and spacing are less than setting threshold value, two connected domains being less than setting threshold value are merged;
C) judge the character area of connected domain, character depth-width ratio, delete character area, character depth-width ratio do not belong to the connected domain of setting range;
D) according to characters on license plate all integrated distribution this features in certain altitude range, setting straight line H, scans car plate bianry image from top to bottom, finds straight line H ', make its number of passing connected domain maximum, delete coboundary H ' below or lower boundary in the connected domain of more than H ';
E) to be substantially consistent this characteristic according to the character height of characters on license plate own, connected domain height average is obtained by cluster analysis, if the difference of connected domain height and this mean value is greater than setting threshold value, then delete this connected domain, otherwise retain this connected domain;
The slope of the straight line F) formed according to connected domain central point between two, calculates the absolute value of difference between these slopes, according to absolute difference size, deletes the connected domain that two ends do not satisfy condition;
G) according to the spacing of connected domain, the connected domain being greater than 3 times of connected domain mean breadths is considered as the connected domain peeled off, and deletes the connected domain peeled off;
H) if the credible character connected domain quantity found in bianry image satisfies condition, continue to perform subsequent step; Otherwise be judged as not containing license board information in license plate image, and terminate bearing calibration.
7. license plate image horizontal tilt bearing calibration as claimed in claim 6, it is characterized in that described step F) be specially: first straight slope a formed with the center of penultimate connected domain, the straight slope b of second and last connected domain central point formation, calculate abs (a-b) if be less than setting threshold value, be then considered as satisfying condition; If be greater than setting threshold value, just think that two ends have a connected domain to need to delete, then the straight slope c of first and last connected domain central point formation is calculated, the relatively size of abs (a-c) and abs (b-c), if abs (a-c) >abs (b-c), then delete last connected domain, otherwise, delete first connected domain; Step F is repeated to remaining connected domain) judgement, until satisfy condition.
8. license plate image horizontal tilt bearing calibration as claimed in claim 1, it is characterized in that described step 3) be specially: the highs and lows finding each connected domain, these points are considered as the up-and-down boundary point of connected domain, by least square method, linear fit is carried out respectively to all up-and-down boundary points, judge up-and-down boundary point middle distance separately fitting a straight line point farthest whether be less than or equal to threshold value from the distance of fitting a straight line, if be less than or equal to threshold value, then continue subsequent step, if be greater than threshold value, then delete this point farthest, and again least-squares algorithm linear fitting is carried out to the point of remainder, until point farthest after matching is less than or equal to threshold value apart from the distance of fitting a straight line, finally obtain up-and-down boundary straight line.
9. license plate image horizontal tilt bearing calibration as claimed in claim 1, it is characterized in that described step 4) in obtain the anglec of rotation and be specially: if the absolute value of the difference of the slope of up-and-down boundary straight line is less than setting threshold value, so the slope of up-and-down boundary straight line is added and averages, obtain the anglec of rotation of car plate; Otherwise select the slope of the maximum fitting a straight line of the number of reservation frontier point to calculate the anglec of rotation of car plate.
10. license plate image horizontal tilt bearing calibration as claimed in claim 1, it is characterized in that described calculate rotation according to the anglec of rotation after the position of up-and-down boundary of character zone be specially:
The computing method of position, coboundary choose to be positioned at straight line upper end, coboundary and the nearest point of distance fitting a straight line, according to the anglec of rotation obtained, calculates this and put position in image after rotation;
Lower boundary position calculating method chooses to be positioned at lower boundary straight-line lower end and the nearest point of distance fitting a straight line, according to the anglec of rotation obtained, calculates this and put position in image after rotation;
Its computing formula is as follows:
x＝Xcosα-Ysinα
y＝Xsinα+Ycosα
Wherein (X, Y) is the coordinate before the point nearest apart from fitting a straight line rotates, and (x, y) is the coordinate after the point nearest apart from fitting a straight line rotates, and α is the anglec of rotation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410561462.0A CN104318233B (en) | 2014-10-19 | 2014-10-19 | License plate image horizontal tilt bearing calibration |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410561462.0A CN104318233B (en) | 2014-10-19 | 2014-10-19 | License plate image horizontal tilt bearing calibration |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104318233A true CN104318233A (en) | 2015-01-28 |
| CN104318233B CN104318233B (en) | 2018-01-26 |
Family
ID=52373462
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410561462.0A Expired - Fee Related CN104318233B (en) | 2014-10-19 | 2014-10-19 | License plate image horizontal tilt bearing calibration |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104318233B (en) |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105488501A (en) * | 2015-11-26 | 2016-04-13 | 南京富士通南大软件技术有限公司 | Method for correcting license plate slant based on rotating projection |
| CN105608455A (en) * | 2015-12-18 | 2016-05-25 | 浙江宇视科技有限公司 | License plate tilt correction method and apparatus |
| CN105678296A (en) * | 2015-12-30 | 2016-06-15 | 小米科技有限责任公司 | Method and apparatus for determining angle of inclination of characters |
| CN106022333A (en) * | 2016-04-26 | 2016-10-12 | 安凯 | Vehicle license plate tilt image correcting method |
| CN106022334A (en) * | 2016-05-04 | 2016-10-12 | 安凯 | Stereoscopic-vision-system-based precise correction method for tilted license plate image |
| CN106127206A (en) * | 2016-06-28 | 2016-11-16 | 北京智芯原动科技有限公司 | The vertical angles detection method of a kind of car plate and device |
| CN106874904A (en) * | 2017-01-09 | 2017-06-20 | 北京大学深圳研究生院 | A kind of car plate picture antidote and device |
| CN107085834A (en) * | 2017-04-25 | 2017-08-22 | 西安工程大学 | A kind of image de-noising method based on image rotation and piecemeal singular value decomposition |
| CN107292305A (en) * | 2017-06-20 | 2017-10-24 | 北京小米移动软件有限公司 | Character rotation method and device |
| CN107506765A (en) * | 2017-10-13 | 2017-12-22 | 厦门大学 | A kind of method of the license plate sloped correction based on neutral net |
| CN107563330A (en) * | 2017-09-04 | 2018-01-09 | 南京邮电大学 | A kind of horizontal tilt car plate antidote in monitor video |
| CN107622245A (en) * | 2017-09-26 | 2018-01-23 | 武汉中旗生物医疗电子有限公司 | Papery Wave shape extracting method and device |
| CN108171229A (en) * | 2017-12-27 | 2018-06-15 | 广州多益网络股份有限公司 | A kind of recognition methods of hollow adhesion identifying code and system |
| CN108241859A (en) * | 2016-12-26 | 2018-07-03 | 浙江宇视科技有限公司 | The bearing calibration of car plate and device |
| WO2018219054A1 (en) * | 2017-06-02 | 2018-12-06 | 杭州海康威视数字技术股份有限公司 | Method, device, and system for license plate recognition |
| CN110020655A (en) * | 2019-04-19 | 2019-07-16 | 厦门商集网络科技有限责任公司 | A kind of character denoising method and terminal based on binaryzation |
| CN110390334A (en) * | 2018-04-19 | 2019-10-29 | 富士施乐株式会社 | Information processing unit and storage medium |
| CN110400259A (en) * | 2019-07-26 | 2019-11-01 | 公安部交通管理科学研究所 | The automotive number plate image inclination angle correction system and method rotated based on least square method and coordinate |
| CN110427937A (en) * | 2019-07-18 | 2019-11-08 | 浙江大学 | A kind of correction of inclination license plate and random length licence plate recognition method based on deep learning |
| WO2020133464A1 (en) * | 2018-12-29 | 2020-07-02 | Zhejiang Dahua Technology Co., Ltd. | Systems and methods for license plate recognition |
| CN111639642A (en) * | 2020-05-07 | 2020-09-08 | 浙江大华技术股份有限公司 | Image processing method, device and apparatus |
| CN111652230A (en) * | 2020-05-25 | 2020-09-11 | 浙江大华技术股份有限公司 | License plate recognition method, electronic device and storage medium |
| CN112183255A (en) * | 2020-09-15 | 2021-01-05 | 西北工业大学 | Underwater target visual identification and attitude estimation method based on deep learning |
| CN113436080A (en) * | 2021-06-30 | 2021-09-24 | 平安科技（深圳）有限公司 | Seal image processing method, device, equipment and storage medium |
| CN114719822A (en) * | 2022-02-24 | 2022-07-08 | 江苏省送变电有限公司 | GIS equipment fault detection method based on geometric method |
| CN114972824A (en) * | 2022-06-24 | 2022-08-30 | 小米汽车科技有限公司 | Rod detection method and device, vehicle and storage medium |
| CN116385315A (en) * | 2023-05-31 | 2023-07-04 | 日照天一生物医疗科技有限公司 | Image enhancement method and system for simulated ablation of tumor therapeutic instrument |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130163822A1 (en) * | 2006-04-04 | 2013-06-27 | Cyclops Technologies, Inc. | Airborne Image Capture and Recognition System |
-
2014
- 2014-10-19 CN CN201410561462.0A patent/CN104318233B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130163822A1 (en) * | 2006-04-04 | 2013-06-27 | Cyclops Technologies, Inc. | Airborne Image Capture and Recognition System |
Non-Patent Citations (2)
| Title |
|---|
| 王宗跃等: "基于字符上下边缘的车牌矫正方法", 《武汉理工大学学报》 * |
| 闫兵: "基于灰度信息的车牌识别方法的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (43)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105488501A (en) * | 2015-11-26 | 2016-04-13 | 南京富士通南大软件技术有限公司 | Method for correcting license plate slant based on rotating projection |
| CN105488501B (en) * | 2015-11-26 | 2018-11-16 | 南京富士通南大软件技术有限公司 | The method of license plate sloped correction based on rotation projection |
| CN105608455A (en) * | 2015-12-18 | 2016-05-25 | 浙江宇视科技有限公司 | License plate tilt correction method and apparatus |
| CN105608455B (en) * | 2015-12-18 | 2019-06-11 | 浙江宇视科技有限公司 | A kind of license plate sloped correcting method and device |
| CN105678296A (en) * | 2015-12-30 | 2016-06-15 | 小米科技有限责任公司 | Method and apparatus for determining angle of inclination of characters |
| CN105678296B (en) * | 2015-12-30 | 2019-12-06 | 小米科技有限责任公司 | Method and device for determining character inclination angle |
| CN106022333A (en) * | 2016-04-26 | 2016-10-12 | 安凯 | Vehicle license plate tilt image correcting method |
| CN106022334A (en) * | 2016-05-04 | 2016-10-12 | 安凯 | Stereoscopic-vision-system-based precise correction method for tilted license plate image |
| CN106127206A (en) * | 2016-06-28 | 2016-11-16 | 北京智芯原动科技有限公司 | The vertical angles detection method of a kind of car plate and device |
| CN106127206B (en) * | 2016-06-28 | 2019-04-05 | 北京智芯原动科技有限公司 | A kind of vertical angles detection method and device of license plate |
| CN108241859A (en) * | 2016-12-26 | 2018-07-03 | 浙江宇视科技有限公司 | The bearing calibration of car plate and device |
| CN106874904A (en) * | 2017-01-09 | 2017-06-20 | 北京大学深圳研究生院 | A kind of car plate picture antidote and device |
| CN107085834A (en) * | 2017-04-25 | 2017-08-22 | 西安工程大学 | A kind of image de-noising method based on image rotation and piecemeal singular value decomposition |
| WO2018219054A1 (en) * | 2017-06-02 | 2018-12-06 | 杭州海康威视数字技术股份有限公司 | Method, device, and system for license plate recognition |
| CN108985137A (en) * | 2017-06-02 | 2018-12-11 | 杭州海康威视数字技术股份有限公司 | A kind of licence plate recognition method, apparatus and system |
| CN107292305A (en) * | 2017-06-20 | 2017-10-24 | 北京小米移动软件有限公司 | Character rotation method and device |
| CN107563330A (en) * | 2017-09-04 | 2018-01-09 | 南京邮电大学 | A kind of horizontal tilt car plate antidote in monitor video |
| CN107563330B (en) * | 2017-09-04 | 2020-10-16 | 南京邮电大学 | Horizontal inclined license plate correction method in surveillance video |
| CN107622245A (en) * | 2017-09-26 | 2018-01-23 | 武汉中旗生物医疗电子有限公司 | Papery Wave shape extracting method and device |
| CN107622245B (en) * | 2017-09-26 | 2020-02-07 | 武汉中旗生物医疗电子有限公司 | Paper waveform extraction method and device |
| CN107506765A (en) * | 2017-10-13 | 2017-12-22 | 厦门大学 | A kind of method of the license plate sloped correction based on neutral net |
| CN107506765B (en) * | 2017-10-13 | 2020-09-01 | 厦门大学 | License plate inclination correction method based on neural network |
| CN108171229A (en) * | 2017-12-27 | 2018-06-15 | 广州多益网络股份有限公司 | A kind of recognition methods of hollow adhesion identifying code and system |
| CN108171229B (en) * | 2017-12-27 | 2021-11-16 | 广州多益网络股份有限公司 | Method and system for identifying hollow adhesion verification code |
| CN110390334A (en) * | 2018-04-19 | 2019-10-29 | 富士施乐株式会社 | Information processing unit and storage medium |
| CN113228033A (en) * | 2018-12-29 | 2021-08-06 | 浙江大华技术股份有限公司 | License plate recognition system and method |
| US11842535B2 (en) | 2018-12-29 | 2023-12-12 | Zhejiang Dahua Technology Co., Ltd. | Systems and methods for license plate recognition |
| WO2020133464A1 (en) * | 2018-12-29 | 2020-07-02 | Zhejiang Dahua Technology Co., Ltd. | Systems and methods for license plate recognition |
| CN110020655A (en) * | 2019-04-19 | 2019-07-16 | 厦门商集网络科技有限责任公司 | A kind of character denoising method and terminal based on binaryzation |
| CN110427937A (en) * | 2019-07-18 | 2019-11-08 | 浙江大学 | A kind of correction of inclination license plate and random length licence plate recognition method based on deep learning |
| CN110427937B (en) * | 2019-07-18 | 2022-03-22 | 浙江大学 | Inclined license plate correction and indefinite-length license plate identification method based on deep learning |
| CN110400259A (en) * | 2019-07-26 | 2019-11-01 | 公安部交通管理科学研究所 | The automotive number plate image inclination angle correction system and method rotated based on least square method and coordinate |
| CN110400259B (en) * | 2019-07-26 | 2022-07-05 | 公安部交通管理科学研究所 | Motor vehicle license plate image inclination angle correction system and method based on least square method and coordinate rotation |
| CN111639642A (en) * | 2020-05-07 | 2020-09-08 | 浙江大华技术股份有限公司 | Image processing method, device and apparatus |
| CN111652230A (en) * | 2020-05-25 | 2020-09-11 | 浙江大华技术股份有限公司 | License plate recognition method, electronic device and storage medium |
| CN111652230B (en) * | 2020-05-25 | 2023-05-12 | 浙江大华技术股份有限公司 | License plate recognition method, electronic device and storage medium |
| CN112183255A (en) * | 2020-09-15 | 2021-01-05 | 西北工业大学 | Underwater target visual identification and attitude estimation method based on deep learning |
| CN113436080A (en) * | 2021-06-30 | 2021-09-24 | 平安科技（深圳）有限公司 | Seal image processing method, device, equipment and storage medium |
| CN114719822A (en) * | 2022-02-24 | 2022-07-08 | 江苏省送变电有限公司 | GIS equipment fault detection method based on geometric method |
| CN114972824A (en) * | 2022-06-24 | 2022-08-30 | 小米汽车科技有限公司 | Rod detection method and device, vehicle and storage medium |
| CN114972824B (en) * | 2022-06-24 | 2023-07-14 | 小米汽车科技有限公司 | Rod detection method, device, vehicle and storage medium |
| CN116385315A (en) * | 2023-05-31 | 2023-07-04 | 日照天一生物医疗科技有限公司 | Image enhancement method and system for simulated ablation of tumor therapeutic instrument |
| CN116385315B (en) * | 2023-05-31 | 2023-09-08 | 日照天一生物医疗科技有限公司 | Image enhancement method and system for simulated ablation of tumor therapeutic instrument |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104318233B (en) | 2018-01-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104318233A (en) | Method for horizontal tilt correction of number plate image | |
| CN109886896B (en) | Blue license plate segmentation and correction method | |
| CN108280450B (en) | Expressway pavement detection method based on lane lines | |
| CN110298216B (en) | Vehicle deviation alarm method based on lane line gradient image self-adaptive threshold segmentation | |
| US8948455B2 (en) | Travel path estimation apparatus and program | |
| CN104102905B (en) | A kind of adaptive detection method of lane line | |
| CN102043959B (en) | License plate character segmentation method | |
| CN109583365B (en) | Method for detecting lane line fitting based on imaging model constrained non-uniform B-spline curve | |
| US20140063251A1 (en) | Lane correction system, lane correction apparatus and method of correcting lane | |
| CN105488501A (en) | Method for correcting license plate slant based on rotating projection | |
| CN102509383A (en) | Feature detection and template matching-based mixed number identification method | |
| JP5402828B2 (en) | Lane boundary detection device, lane boundary detection program | |
| CN107423735B (en) | License plate positioning method utilizing horizontal gradient and saturation | |
| CN109034019B (en) | Yellow double-row license plate character segmentation method based on row segmentation lines | |
| CN109308478B (en) | Character recognition method and device | |
| CN109740532B (en) | Path identification and center line optimization method based on circular road | |
| CN102509095B (en) | Number plate image preprocessing method | |
| CN111127498A (en) | Canny edge detection method based on edge self-growth | |
| CN115311277A (en) | Pit defect identification method for stainless steel product | |
| CN106326821B (en) | The method and device of License Plate | |
| CN109410235B (en) | Target tracking method fusing edge features | |
| CN112597846A (en) | Lane line detection method, lane line detection device, computer device, and storage medium | |
| CN104156727A (en) | Lamplight inverted image detection method based on monocular vision | |
| CN108256385A (en) | The front vehicles detection method of view-based access control model | |
| CN103745236A (en) | Texture image identification method and texture image identification device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information |
Inventor after: Ye Xiuzi Inventor after: Hong Zhenjie Inventor after: Sheng Jia Inventor after: Xiang Xubin Inventor after: Zhang Sanyuan Inventor before: Sheng Jia Inventor before: Xiang Xubin Inventor before: Ye Xiuzi Inventor before: Hong Zhenjie Inventor before: Zhang Sanyuan |
|
| CB03 | Change of inventor or designer information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20150128 Assignee: Big data and Information Technology Research Institute of Wenzhou University Assignor: Wenzhou University Contract record no.: X2020330000098 Denomination of invention: Horizontal tilt correction method of license plate image Granted publication date: 20180126 License type: Common License Record date: 20201115 |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180126 |