CN103810474A - Car plate detection method based on multiple feature and low rank matrix representation - Google Patents

Car plate detection method based on multiple feature and low rank matrix representation Download PDF

Info

Publication number
CN103810474A
CN103810474A CN201410051645.8A CN201410051645A CN103810474A CN 103810474 A CN103810474 A CN 103810474A CN 201410051645 A CN201410051645 A CN 201410051645A CN 103810474 A CN103810474 A CN 103810474A
Authority
CN
China
Prior art keywords
car plate
image
license plate
matrix representation
many
Prior art date
Application number
CN201410051645.8A
Other languages
Chinese (zh)
Inventor
王卫卫
冯象初
王佳琳
丁亚男
Original Assignee
西安电子科技大学
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 西安电子科技大学 filed Critical 西安电子科技大学
Priority to CN201410051645.8A priority Critical patent/CN103810474A/en
Publication of CN103810474A publication Critical patent/CN103810474A/en

Links

Abstract

The invention discloses a car plate detection method based on a multiple feature and low rank matrix representation. The car plate detection method based on the multiple feature and low rank matrix representation includes: (1) inputting an image; (2) extracting color features and self-adaption LBP operator features; (3) obtaining a multiple feature and low rank matrix representation model; (4) dividing and solving the model so as to obtain a sub model; (5) outputting a fake car plate area and obtaining a final accurate car plate area; (6) correcting a car plate and outputting an image after being positioned. The car plate detection method based on the multiple feature and low rank matrix representation combines a plurality of features of the car plate, can effectively improve robustness and accuracy of car plate detection, reduces false detection rate, and can be used in an intelligent transportation system.

Description

A kind of car plate detection method based on many features low-rank matrix representation

Technical field:

The present invention relates to image technique process field, be specifically related to a kind of car plate detection method based on many features low-rank matrix representation, be applied to Vehicle License Plate Recognition System in intelligent transportation field.

Background technology:

The concept of intelligent transportation system (Intelligent Transportation System) is to be proposed by intelligent transportation association of the U.S. nineties in 20th century, is the preface problem of current field of traffic.Vehicle License Plate Recognition System, for intelligent traffic administration system provides practical, efficient solution, has attracted the broad research of various countries' researchers.At present existing numerous detection algorithm, is widely used in the places such as road management violating the regulations, managing system of car parking, vehicle recording system.Domestic most city has all been equipped with ripe Vehicle License Plate Recognition System, for traffic administration, fighting crime plays a good role.

In Vehicle License Plate Recognition System, exporting recognition result from being input to of image is the process of a complicated image processing, is generally made up of three parts: car plate location (car plate detection), Character segmentation and character recognition.License plate locating method is varied at present, generally speaking can be divided into two large classes: the Location of Vehicle License Plate algorithm based on color image processing and the Location of Vehicle License Plate algorithm based on gray level image treatment technology.

Mainly utilize the intrinsic colour match of car plate to position analysis based on color image processing: the localization method to analysis based on fuzzy training and color edges; Utilize hsv color model to analyze, the image of input is carried out to gray shade scale classification, then judge car plate position in conjunction with mathematical morphology, word frequency statistic method; License plate locating method based on car plate geometric properties and textural characteristics etc.

The time that the location algorithm of gray level image treatment technology produces is wanted early compared with the location algorithm of color image processing, the calculated amount of this type of algorithm is less, arithmetic speed is very fast, and most algorithms are the bases using the bianry image after Binary Sketch of Grey Scale Image as location algorithm.Such as: the method for maximum between-cluster variance method (otsu method), local binarization, binarization method, Niblack Binarization methods and the Binarization methods based on various features based on texture information (wavelet decomposition, unity and coherence in writing feature, projective clustering etc.).

Summary of the invention:

The object of the invention is to exist verification and measurement ratio not high for car plate detection method in prior art, be subject to extraneous illumination effect large, robustness does not wait by force factors, the invention provides a kind of car plate detection method based on many features low-rank matrix representation.

Based on a car plate detection method for many features low-rank matrix representation, there is target image, said method comprising the steps of:

(1) input picture: calculate X i=[x 1, x 2x n], i gets 1 and 2; X 1for color characteristic matrix, rgb space is converted into HSV space and carries out unequal interval quantification, three color components are expressed as a n dimensional vector n, then calculate its histogram as color characteristic, X 2for a kind of LBP operator eigenmatrix of adaptive threshold.

(2) extract color characteristic and adaptive LBP operator feature: the concrete steps of extracting adaptive LBP operator characteristics algorithm are as follows:

A, convert the image of input system to gray level image, to the summation of image { grayv (i, j) } grey scale pixel value, then obtain mean value:

sum = Σ i = 1 m Σ j = 1 n grayv ( i , j )

avg = sum m × n

B, utilize total textural characteristics to remove background, the absolute value sum of the difference of the grey scale pixel value of computed image and mean pixel gray-scale value, ask its mean value:

Diffsum = Σ i = 1 m Σ j = 1 n | grayv ( i , j ) - avg |

Davg = diffsum m × n

Utilize Local textural feature to remove background, with the moving window of 3 × 3 sizes, traversing graph picture, ask for the poor of center pixel gray-scale value and neighboring pixel gray-scale value, averaged in each video in window:

Areavg = Σ i = 0 7 | g i - g c | 8

C, according to experimental data, the method for the Fitting Calculation adaptive threshold:

θ = 4 × Davg + Areavg

(3), on the basis of low-rank matrix representation model (LRR), following many features low-rank matrix representation model is proposed:

min Z 1 , . . . , Z K E 1 , . . . , E K Σ i = 1 K ( | | A | | * + λ | | E | | 2,1 ) + α | | A | | 2,1

s.t.X i=X iA i+E i,i=1,…,K

Wherein α is greater than 0 coefficient, be used for measuring noise and open country

min J 1 , . . . , J K S 1 , . . . , S K Z 1 , . . . , Z K E 1 , . . . , E K Σ i = 1 K ( | | J i | | * + λ | | E i | | 2,1 ) +α | | A | | 2,1

s.t.???X i=X iS i+E i,

A i=J i,

A i=S i,i=1,…,K.

Can utilize augmentation method of Lagrange multipliers to solve, make A=J, be obtained by augmentation method of Lagrange multipliers

&alpha; | | A | | 2,1 + &Sigma; i = 1 K ( | | J i | | * + &lambda; | | E i | | 2,1 ) + &Sigma; i = 1 K ( < W i , A i - J i > + < Y i , X i - X i S i - E i > + < V i , A i - S i > + &mu; 2 | | X i - X i S i - E i | | F 2 + &mu; 2 | | A i - J i | | F 2 + &mu; 2 | | A i - S i | | F 2 ) ,

Utilize alternating direction method to solve, draw and cut apart rear subspace piece;

(4) model decomposed and solved, obtaining submodel;

(5) export pseudo-license plate area and obtain finally license plate area accurately;

A, the external matrix that leaves every sub spaces according to car plate size, ratio etc. are doubtful license plate area;

B, a saltus step function f (i, j) is set, doubtful license plate area is accurately located, determine the up-and-down boundary of license plate area:

Wherein c (i, j) is

c(i,j)=LBP 8,1(i,j)-LBP 8,1(i,j-1)

I=1 in upper two formulas, 2,3,4 ... N, j=2,3,4 ... M, therefore transition times and the S (i) of a line i are arbitrarily:

S ( i ) = &Sigma; i = 2 N | f ( i , j ) |

Vehicle license one has seven characters, and any a line of each character laterally at least there will be 2 saltus steps, (minority Chinese character there will be 0 time, as rather, cloud), therefore laterally arbitrarily one-row pixels can minimum saltus step 12 times (supposing that character is peaceful Clllll) for vehicle license character, and have the most at most (suppose character for hide MMMMMM) 60 times, but due to car plate often noisy existence around, can cause transversal scanning time, the number of times of license plate area saltus step can increase, therefore in the time of scanning, select 12 values of closing as transition times, if arbitrarily a line transition times and S (i >=12), this a line just likely belongs to license plate area.From top to bottom entire image is scanned, find out the row of all S of meeting (i >=12), and record the line number i of this line.If there is the capable S of meeting of continuous h (i >=12) (capable by the i~i+h), can obtain a width is M, be highly the rectangular area of h, this region is likely just license plate area, and the region therefore in vehicle image without this feature has obtained eliminating.

(6) car plate is proofreaied and correct, the image behind output location: proposed a kind of simple and effective license plate sloped correcting method, the angle in certain limit is tilted to have corrective action well, concrete steps are as follows:

Method by Hough transformation and projection combination is the regional correction of car plate fine positioning, output license plate image.Consider practical situations, degree of tilt generally, between-10 ° ± 10 °, can reduce calculated amount so carry out car plate rectification in this interval.Using the center of image as initial point, level is to the right X-axis forward, is Y-axis forward vertically upward.Centered by initial point, image turn clockwise θ (θ ∈ 10 ° ,-9 ° ... 9 °, 10 ° }).As image rotation θ, on license plate image, the projection of arbitrfary point Point (x, y) on x axle can be calculated according to following formula:

P xcosθ+(y-xtgθ)sinθ

P x∈(-P max,P max)

P max = ( P w / 2 ) 2 + ( P h / 2 ) 2

In formula, P ω, P hfor width and the height of image.

If P θ(x) be projection on x axle after image rotation θ.In the time that image is horizontal, because character arranging direction is perpendicular to x axle, x axle Projective Curve P θ(x) upper maximal value distribution is the most concentrated, can determine fast angle of inclination according to this feature.

Beneficial effect of the present invention is:

1, the present invention, according to car plate color and LBP feature operator, obtains license plate area in conjunction with improved LRR model and morphological operation.Multiple features of car plate are combined, can effectively improve robustness and accuracy that car plate detects, reduce flase drop.

2, propose a kind of LBP operator of adaptive threshold, can carry out texture analysis to the license plate image under complex background, characteristic information more accurately can be provided.

Accompanying drawing explanation:

Fig. 1 is car plate overhaul flow chart of the present invention;

Fig. 2 is that image of the present invention is respectively original test sample figure, binary picture and adaptive LBP threshold values figure;

Fig. 3 is image up-and-down boundary of the present invention and vertical projection diagram;

Fig. 4 is image car plate testing result exemplary plot of the present invention.

Embodiment:

Below in conjunction with accompanying drawing, the invention will be further described.

As shown in Figure 1, the present invention is a kind of car plate detection method based on many features low-rank matrix representation, has target image, said method comprising the steps of:

Step 1, inputs image to be detected, extracts color characteristic and adaptive LBP operator.Suppose that image is g (i, j), calculating size is (2 k+ 1) 2active window in pixel average intensity value:

A i ( x , y ) = &Sigma; i = x - 2 k - 1 x + 2 k - 1 &Sigma; j = y - 2 k - 1 y + 2 k - 1 g ( i , j ) ( 2 k + 1 ) 2

For each pixel, the pixel mean intensity between the window of non-overlapping copies is poor in the horizontal and vertical directions to calculate respectively it:

E k , h ( x , y ) = | A k ( x + 2 k - 1 , y ) - A k ( x - 2 k - 1 , y ) | E k , v ( x , y ) = | A k ( x , y + 2 k - 1 ) - A k ( x , y - 2 k - 1 ) |

For each pixel, can make E k,h(x, y) or E k, νthe k value that (x, y) value reaches is maximum (no matter direction) is used for arranging optimum dimension:

S best(x,y)=2 k+1

Known above, S best(x, y) is the approximate size of the texture primitive of the pixel take (x, y) as coordinate.The combination of this size and LBP algorithm, has reduced the error that LBP occurs in the extraction of texture primitive feature.

The present invention proposes a kind of LBP algorithm of adaptive threshold, can remove to greatest extent background, obtains comparatively accurate texture information.The concrete steps of algorithm are as follows:

Convert the image of input system to gray level image, to the summation of image { grayv (i, j) } grey scale pixel value, then obtain mean value:

sum = &Sigma; i = 1 m &Sigma; j = 1 n grayv ( i , j )

avg = sum m &times; n

Utilize total textural characteristics to remove background.The absolute value sum of the difference of the grey scale pixel value of computed image and mean pixel gray-scale value, ask its mean value:

Diffsum = &Sigma; i = 1 m &Sigma; j = 1 n | grayv ( i , j ) - avg |

Davg = diffsum m &times; n

Utilize Local textural feature to remove background.With the moving window of 3 × 3 sizes, traversing graph picture, ask for the poor of center pixel gray-scale value and neighboring pixel gray-scale value, averaged in each video in window:

Areavg = &Sigma; i = 0 7 | g i - g c | 8

According to experimental data, the method for the Fitting Calculation adaptive threshold:

&theta; = 4 &times; Davg + Areavg

From above computing method, the threshold value in each window is not identical, because the difference of local grain.After θ is determined, converse the gray-scale value of center pixel.The image that these pixels with new gray-scale value form, is the transfer image acquisition obtaining after being processed by adaptive threshold LBP operator.

Under normal circumstances, license plate shared area ratio in whole image is all less than background area occupied ratio.Therefore, this algorithm utilization overall situation and local pixel grey scale inequality decide the size of threshold value jointly, have not only considered the grain distribution of entire image, have also considered local texture variations.If only consider single grain distribution situation, can affect the result of Texture classification.

Step 2, when iterative optimization problem, can be divided into 3 subproblems:

Subproblem 1. is fixed its dependent variable and solves J i:

J i = arg min J 1 &mu; | | J i | | * + 1 2 | | J i - ( A i + W 2 / &mu; ) | | F 2

This problem can utilize singular value threshold value to solve;

Subproblem 2. is fixed its dependent variable and solves A and obtain:

A = arg min A &alpha; 2 &mu; | | A | | 2,1 + 1 2 &Sigma; i = 1 K | | A - M | | F 2

M is a K × N 2matrix

F i=(J i+S i-(W i+V i)μ)/2,i=1,…,K.

S i = ( I + X i T X i ) - 1 ( X i T ( X i - E i ) + A i + X i T Y i + V i - W i &mu; ) .

Subproblem 3. is fixed its dependent variable and solves E i:

E i = arg min E i - &lambda; &mu; | | E i | | 2,1 + | | E i - ( X i - X i A i + Y i / &mu; ) | | F 2

This problem passing threshold computing solves.

Step 3, selected doubtful license plate area, accurately locates with saltus step function.Determine the border, left and right of car plate:

After above-mentioned processing, still have the existence of pseudo-license plate area, so will pseudo-license plate area be got rid of in border, definite car plate left and right.Shown in Fig. 3, can find that license plate area has obvious difference compared with other region: to car plate, to character, then change to background by car plate by change of background by change of background, very regular.And it is more outstanding to find after use sciagraphy that this rule changes: after vertical projection vehicle license character zone the variation of corresponding drop shadow curve slope very fierce, change to background by car plate by change of background again to car plate, at this moment drop shadow curve's slope has extremely significantly sudden change, but not license plate area the slope of corresponding drop shadow curve will be relatively mild many; Determine that because why pseudo-license plate area meets the condition of car plate up-and-down boundary is only because it exists many little interference regions, needed condition while making it meet transversal scanning and degree of contrast, but pseudo-license plate area does not have by background to car plate, background to character, car plate to the such variation of background, thereby the vertical projection curve that makes it will link up relatively and gently, do not have to change very fierce region, more do not meet the slope sudden change of rule.Therefore, can change the fiercest corresponding image-region in region by finding vertical projection diagram rate of curve, determine the border, left and right of vehicle license and remove pseudo-license plate area with this, recording row (being listed as by the j~j+1) and the width w at place, searched out region simultaneously.

License plate area checking:

Through determining of car plate border, all can orient comparatively accurately in most cases license plate area, meet above condition but sometimes may still have pseudo-license plate area, this just need to verify the car plate of locating out again, prevents mistake location.Because the ratio of width to height of car plate that China adopts is between 60:3l~22:7, therefore can verify by this condition again license plate area after positioning,

60 31 &le; w h &le; 22 7

But the difference of shooting angle when gathering image in reality, often makes the ratio of width to height slightly to exceed this scope, through experiment, this restriction is carried out to suitable relaxing and can obtain good locating effect,

1 &le; w h &le; 6

Therefore the ratio of width to height meets the license plate area for orienting of region of above formula.

More than show and described ultimate principle of the present invention and principal character and advantage of the present invention.The technician of the industry should understand; the present invention is not restricted to the described embodiments; that in above-described embodiment and instructions, describes just illustrates principle of the present invention; without departing from the spirit and scope of the present invention; the present invention also has various changes and modifications, and these changes and improvements all fall in the claimed scope of the invention.The claimed scope of the present invention is defined by appending claims and equivalent thereof.

Claims (7)

1. the car plate detection method based on many features low-rank matrix representation, has target image, it is characterized in that, said method comprising the steps of:
(1) input picture;
(2) extract color characteristic and adaptive LBP operator feature;
(3) many features end order matrix representation model;
(4) model decomposed and solved, obtaining submodel;
(5) export pseudo-license plate area and obtain finally license plate area accurately;
(6) car plate is proofreaied and correct, the image behind output location.
2. a kind of car plate detection method based on many features low-rank matrix representation according to claim 1, is characterized in that, the concrete steps of extracting adaptive LBP operator characteristics algorithm are as follows:
(1) convert the image of input system to gray level image, to the summation of image { grayv (i, j) } grey scale pixel value, then obtain mean value:
(2) utilize total textural characteristics to remove background, the absolute value sum of the difference of the grey scale pixel value of computed image and mean pixel gray-scale value, ask its mean value:
Utilize Local textural feature to remove background, with the moving window of 3 × 3 sizes, traversing graph picture, ask for the poor of center pixel gray-scale value and neighboring pixel gray-scale value, averaged in each video in window:
(3) according to experimental data, the method for the Fitting Calculation adaptive threshold:
3. a kind of car plate detection method based on many features low-rank matrix representation according to claim 1, is characterized in that, described many features end order matrix representation model:
s.t.???X i=X iA i+E i,i=1,…,K
Wherein α is greater than 0 coefficient, be used for measuring noise and the wild error of bringing of putting.
4. a kind of car plate detection method based on many features low-rank matrix representation according to claim 1, is characterized in that, described new model is equivalent to drag
s.t.???X i=X iS i+E i,
A i=J i,
A i=S i,i=1,…,K。
5. a kind of car plate detection method based on many features low-rank matrix representation according to claim 4, is characterized in that, when described iterative optimization problem, can be divided into 3 subproblems:
Subproblem 1. is fixed its dependent variable and solves J i:
This problem can utilize singular value threshold value to solve;
Subproblem 2. is fixed other variable solves A and obtains:
M is a K × N 2matrix
F i=(J i+S i-(W i+V i)μ)/2,i=1,…,K.
Subproblem 3. is fixed its dependent variable and solves E i:
This problem passing threshold computing solves.
6. a kind of car plate detection method based on many features low-rank matrix representation according to claim 1, is characterized in that, the pseudo-license plate area of described output also obtains finally license plate area accurately, is specially:
(1) external matrix that leaves every sub spaces according to car plate size, ratio etc. is doubtful license plate area;
(2) a saltus step function f (i, j) is set, doubtful license plate area is accurately located, determine the up-and-down boundary of license plate area:
Wherein c (i, j) is
c(i,j)=LBP 8,1(i,j)-LBP 8,1(i,j-1)
I=1 in upper two formulas, 2,3,4 ... N, j=2,3,4 ... M, therefore transition times and the S (i) of a line i are arbitrarily:
If arbitrarily a line transition times and S (i >=12), this line just likely belongs to license plate area.From top to bottom entire image is scanned, find out the row of all S of meeting (i >=12), and record the line number i of this line.If there is the capable S of meeting of continuous h (i >=12), can obtain a width is M, is highly the rectangular area of h, and this region is likely just license plate area, and the region therefore in vehicle image without this feature has obtained eliminating.
7. a kind of car plate detection method based on many features low-rank matrix representation according to claim 1, is characterized in that, the method by Hough transformation and projection combination is the regional correction of car plate fine positioning, output license plate image.
CN201410051645.8A 2014-02-14 2014-02-14 Car plate detection method based on multiple feature and low rank matrix representation CN103810474A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410051645.8A CN103810474A (en) 2014-02-14 2014-02-14 Car plate detection method based on multiple feature and low rank matrix representation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410051645.8A CN103810474A (en) 2014-02-14 2014-02-14 Car plate detection method based on multiple feature and low rank matrix representation

Publications (1)

Publication Number Publication Date
CN103810474A true CN103810474A (en) 2014-05-21

Family

ID=50707216

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410051645.8A CN103810474A (en) 2014-02-14 2014-02-14 Car plate detection method based on multiple feature and low rank matrix representation

Country Status (1)

Country Link
CN (1) CN103810474A (en)

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106326898A (en) * 2016-08-04 2017-01-11 安凯 License plate image extraction method
CN106548172A (en) * 2016-11-21 2017-03-29 深圳怡化电脑股份有限公司 A kind of angle of inclination of image determines method and device
CN106650731A (en) * 2016-12-23 2017-05-10 中山大学 Robust license plate and logo recognition method
CN107337281A (en) * 2017-07-17 2017-11-10 四川理工学院 A kind of strengthened artificial wet land sewage water treatment method
CN107410027A (en) * 2017-08-08 2017-12-01 王玉珍 A kind of bioreactor system for pilot scale culture medicinal plant
CN107596224A (en) * 2017-10-14 2018-01-19 续生堂科技发展(吉林)有限公司 The healthy medicine of one kind bow body and preparation method
CN107610028A (en) * 2017-09-08 2018-01-19 广西民族师范学院 A kind of atmosphere pollution on-line monitoring system based on wireless cloud Sensor Network
CN107632018A (en) * 2017-08-22 2018-01-26 甘肃农业大学 A kind of food microorganisms rapid detection system and method
CN107679764A (en) * 2017-10-23 2018-02-09 钦州学院 A kind of dynamic dispatching method of container hargour truck
CN107894252A (en) * 2017-11-14 2018-04-10 江苏科沃纺织有限公司 It is a kind of to monitor the buried telescopic monitoring system for being sprayed filling device running status in real time
CN107909708A (en) * 2017-10-31 2018-04-13 张珂 A kind of finance device well straightening device
CN107993501A (en) * 2017-12-04 2018-05-04 菏泽学院 A kind of human anatomy teaching system
CN107984756A (en) * 2017-12-06 2018-05-04 付文军 A kind of 3D printing device for being used to prepare angiocarpy bracket
CN108010141A (en) * 2017-12-27 2018-05-08 江西中升智能停车设备有限公司 A kind of intelligent-induction formula parking stall system
CN108257127A (en) * 2018-01-29 2018-07-06 佳木斯大学附属第医院 A kind of system applied to acute poisoning Safety Pre-Evaluation
CN108275530A (en) * 2018-01-18 2018-07-13 柯钢 A kind of elevator safety method for early warning based on machine learning
CN108567413A (en) * 2018-03-02 2018-09-25 黑龙江中医药大学 A kind of multi-functional disease examination equipment of gynaecology of hospital and inspection system
CN108734158A (en) * 2017-04-14 2018-11-02 成都唐源电气股份有限公司 A kind of real-time train number identification method and device
CN109460092A (en) * 2018-08-11 2019-03-12 西华大学 Device parameter control system and method in a kind of streptonigrin compound extraction process

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN203276567U (en) * 2013-05-14 2013-11-06 广州杰赛科技股份有限公司 License plate monitoring device and license plate identification system
KR101328533B1 (en) * 2012-04-10 2013-11-13 한국과학기술원 System and method for compressing vehicle image for detection of license plate

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101328533B1 (en) * 2012-04-10 2013-11-13 한국과학기술원 System and method for compressing vehicle image for detection of license plate
CN203276567U (en) * 2013-05-14 2013-11-06 广州杰赛科技股份有限公司 License plate monitoring device and license plate identification system

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
BIN CHENG,ETC: ""Multi-task Low_rank Affinity Pursuit for Image Segmentation"", 《INTERNATIONAL CONFERENCE ON COMPUTER VISION》 *
朱宁波,等: ""基于自适应LBP的视频文本检测算法"", 《计算机工程》 *
沈勇武,等: ""基于特征颜色边缘检测的车牌定位方法"", 《仪器仪表学报》 *
王钰: ""车辆识别系统中的基础性算法研究"", 《中国博士学位论文全文数据库 信息科技辑》 *
甘超,等: ""多特征稳健主成分分析的视频运动目标分割"", 《中国图像图形学报》 *

Cited By (22)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106326898A (en) * 2016-08-04 2017-01-11 安凯 License plate image extraction method
CN106326898B (en) * 2016-08-04 2019-09-20 安凯 A kind of license plate image extracting method
CN106548172A (en) * 2016-11-21 2017-03-29 深圳怡化电脑股份有限公司 A kind of angle of inclination of image determines method and device
CN106650731A (en) * 2016-12-23 2017-05-10 中山大学 Robust license plate and logo recognition method
CN106650731B (en) * 2016-12-23 2020-02-21 中山大学 Robust license plate and vehicle logo recognition method
CN108734158A (en) * 2017-04-14 2018-11-02 成都唐源电气股份有限公司 A kind of real-time train number identification method and device
CN108734158B (en) * 2017-04-14 2020-05-19 成都唐源电气股份有限公司 Real-time train number identification method and device
CN107337281A (en) * 2017-07-17 2017-11-10 四川理工学院 A kind of strengthened artificial wet land sewage water treatment method
CN107410027A (en) * 2017-08-08 2017-12-01 王玉珍 A kind of bioreactor system for pilot scale culture medicinal plant
CN107632018A (en) * 2017-08-22 2018-01-26 甘肃农业大学 A kind of food microorganisms rapid detection system and method
CN107610028A (en) * 2017-09-08 2018-01-19 广西民族师范学院 A kind of atmosphere pollution on-line monitoring system based on wireless cloud Sensor Network
CN107596224A (en) * 2017-10-14 2018-01-19 续生堂科技发展(吉林)有限公司 The healthy medicine of one kind bow body and preparation method
CN107679764A (en) * 2017-10-23 2018-02-09 钦州学院 A kind of dynamic dispatching method of container hargour truck
CN107909708A (en) * 2017-10-31 2018-04-13 张珂 A kind of finance device well straightening device
CN107894252A (en) * 2017-11-14 2018-04-10 江苏科沃纺织有限公司 It is a kind of to monitor the buried telescopic monitoring system for being sprayed filling device running status in real time
CN107993501A (en) * 2017-12-04 2018-05-04 菏泽学院 A kind of human anatomy teaching system
CN107984756A (en) * 2017-12-06 2018-05-04 付文军 A kind of 3D printing device for being used to prepare angiocarpy bracket
CN108010141A (en) * 2017-12-27 2018-05-08 江西中升智能停车设备有限公司 A kind of intelligent-induction formula parking stall system
CN108275530A (en) * 2018-01-18 2018-07-13 柯钢 A kind of elevator safety method for early warning based on machine learning
CN108257127A (en) * 2018-01-29 2018-07-06 佳木斯大学附属第医院 A kind of system applied to acute poisoning Safety Pre-Evaluation
CN108567413A (en) * 2018-03-02 2018-09-25 黑龙江中医药大学 A kind of multi-functional disease examination equipment of gynaecology of hospital and inspection system
CN109460092A (en) * 2018-08-11 2019-03-12 西华大学 Device parameter control system and method in a kind of streptonigrin compound extraction process

Similar Documents

Publication Publication Date Title
CN106874894B (en) Human body target detection method based on regional full convolution neural network
RU2708422C1 (en) Atm management system and method
CN102609686B (en) Pedestrian detection method
Chen et al. A double-threshold image binarization method based on edge detector
CN101872416B (en) Vehicle license plate recognition method and system of road image
CN104778721B (en) The distance measurement method of conspicuousness target in a kind of binocular image
CN101334836B (en) License plate positioning method incorporating color, size and texture characteristic
CN104751142B (en) A kind of natural scene Method for text detection based on stroke feature
Moghaddam et al. Application of multi-level classifiers and clustering for automatic word spotting in historical document images
Gerónimo et al. 2D–3D-based on-board pedestrian detection system
CN104866616B (en) Monitor video Target Searching Method
CN104077577A (en) Trademark detection method based on convolutional neural network
CN104463795B (en) A kind of dot matrix DM image in 2 D code processing method and processing device
US8340420B2 (en) Method for recognizing objects in images
Liu et al. Iterative infrared ship target segmentation based on multiple features
Lu et al. Perspective rectification of document images using fuzzy set and morphological operations
CN106650640B (en) Negative obstacle detection method based on laser radar point cloud local structure characteristics
CN104008553B (en) Crack detection method with image gradient information and watershed method conflated
CN104778444B (en) The appearance features analysis method of vehicle image under road scene
CN106909941A (en) Multilist character recognition system and method based on machine vision
CN102043950B (en) Vehicle outline recognition method based on canny operator and marginal point statistic
CN103020605B (en) Based on the bridge recognition of Decision-level fusion
CN102722712B (en) Multiple-scale high-resolution image object detection method based on continuity
CN103218816A (en) Crowd density estimation method and pedestrian volume statistical method based on video analysis
Wei et al. Multi-vehicle detection algorithm through combining Harr and HOG features

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20140521