CN105447489B - A kind of character of picture OCR identifying system and background adhesion noise cancellation method - Google Patents
A kind of character of picture OCR identifying system and background adhesion noise cancellation method Download PDFInfo
- Publication number
- CN105447489B CN105447489B CN201510776907.1A CN201510776907A CN105447489B CN 105447489 B CN105447489 B CN 105447489B CN 201510776907 A CN201510776907 A CN 201510776907A CN 105447489 B CN105447489 B CN 105447489B
- Authority
- CN
- China
- Prior art keywords
- boundary
- character
- point
- value
- line
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/273—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion removing elements interfering with the pattern to be recognised
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Character Input (AREA)
Abstract
The present invention discloses the character and background adhesion noise cancellation method of a kind of picture OCR identifying system, includes the following steps:Step 1)The effective range of the stroke length of character to be identified is set;Step 2)The image containing character to be identified under natural environment is chosen, and calculates image difference boundary graph in a manner of mean difference template;Step 3)The difference boundary graph of images to be recognized obtained above is subjected to thresholding, forms the three-valued boundary graph of images to be recognized;Step 4)In three-valued boundary graph, according in range of tilt angles, linear boundary scans for.The character and background adhesion noise cancellation method of a kind of picture OCR identifying system of the present invention, in natural environment complicated and changeable, this method solve existing OCR application systems to having the problem of adhesion noise character location difficulty, it realizes the accurate extraction of character position, and is able to satisfy the requirement of real-time of engineer application.
Description
Technical field
The present invention relates to natural scene picture OCR identifying system more particularly to a kind of character of picture OCR identifying system with
Background adhesion noise cancellation method.
Background technique
In artificial intelligence field, optical character identification OCR is a highly important technology, due to popularizing for smart phone
And the classification of mass picture and the needs searched in cloud storage, so that natural scene picture OCR identification becomes Recent study
A hot spot, OCR technique generally comprises character locating, Character segmentation, several processes such as character recognition, wherein character locating
Speed and precision directly affect the quality of OCR identification technology, are the key that entire OCR systems.
OCR character locating usually utilizes the structural information of character string, to character string location by way of global search
Domain is positioned, and the extraction of character string structural information, and the most frequently used and more effective method is mentioned by binaryzation technology
The edge of character string picture is taken, common binarization method has fixed threshold method, Adaptive Thresholding, Global thresholding and office
Portion's threshold method etc., no matter any algorithm, when facing environment complicated and changeable, such as Various Seasonal, different weather environment complexity
Situation, various noises of introducing that all can be more or less, these noises will easily lead to positioning failure, or generate a large amount of
False character information, so that the calculation amount of subsequent processing greatly increases.
The noise introduced to binaryzation, existing way usually go to filter out noise using some filtering methods, such as
Median filtering, mathematical morphology, two dimensional wavelet analysis etc., these different methods have different effects to different images, often
Although linear low-pass filters and the method for neighborhood averaging can remove partial noise, they have blurred image negative
Effect, the method for median filtering can eliminate isolated noise spot, and the Fuzzy comparisons generated are few, but it is to bianry image
The effect of removal noise is simultaneously bad, and mathematical morphology can erode part black patch to a certain extent, but past in this way
Toward will lead to the aggravation of original image distortion of image, it is unfavorable for subsequent calculating.
In addition, in natural scene picture OCR identification, due to the influence of the factors such as illumination, weather, sundries, so that two-value
Change noise is more, and in practical engineering application, independent, discrete noise is often easier to distinguish, and with character or wait know
The noise that adhesion has occurred in other object is often difficult to handle, adhesion noise bring result be exactly corresponding character position range by
Interference has been arrived, the accurate positionin of character is caused to fail, to affect the whole discrimination of system, how have been eliminated in natural scene
Characters Stuck noise at OCR identification in a critical issue, it is therefore proposed that one kind can satisfy practical application needs
, realize that the real-time Characters Stuck noise cancellation technique under various complex environments seems particularly significant, by eliminating adhesion noise,
The interference in non-character region can be reduced, to provide the accuracy of OCR character locating.
Summary of the invention
The purpose of the present invention is to provide a kind of character of picture OCR identifying system and background adhesion noise cancellation method,
For the OCR technique of existing natural scene picture, the existing deficiency in terms of noise processed proposes a kind of new adhesion noise
Removing method, in natural environment complicated and changeable, this method solve existing OCR application systems to adhesion noise character
It the problem of location difficulty, realizes the accurate extraction of character position, and is able to satisfy the requirement of real-time of engineer application.
In order to achieve the above object, technical solution provided by the invention is:The natural scene picture OCR identifying system
Character and background adhesion noise cancellation method include the following steps executed in order:
A kind of character of picture OCR identifying system and background adhesion noise cancellation method, the picture OCR identifying system
Character and background adhesion noise cancellation method include as follows:
1) the stroke length of character to be identified is arranged in the priori knowledge of the engineering-environment according to applied by OCR system
The effective range of StrokeLen;
2) image containing character to be identified under natural environment is chosen, and calculates image in a manner of mean difference template
Difference boundary graph, the difference boundary graph include character zone and background area;
3) the difference boundary graph of images to be recognized obtained above is subjected to thresholding, forms the three-valued of images to be recognized
Boundary graph, the three-valued boundary graph include region to be identified and background area;
4) in the three-valued boundary graph of images to be recognized obtained above, according in range of tilt angles, linear type side
Boundary scans for, and checks whether the boundary has exceeded the effective range of stroke length, when without departing from effective model of stroke length
It encloses, then it is assumed that be effective boundary point, and give and retain, when the effective range for exceeding stroke length, then the edge determination is to make an uproar
Sound removes it from three value boundary images.
In the step 3), the formula that the difference boundary graph of the images to be recognized is carried out thresholding is as follows:
Wherein, P (x, y) is the difference value of pixel (x, y) in difference diagram, and ParamA and ParamB are pre-set thresholds
Value, c1, c2 and c3 are signless integers, and value range is [0,255];Wherein c1 is indicated around the brightness ratio of current pixel point more
Bright boundary, c2 indicate boundary darker around the brightness relatively of current pixel point, and c3 indicates the value on non-boundary.
The c1 color is black, value c1=0;The c2 color is white, and value c2=255 is described
C3 color be grey, value c3=128, the value of difference boundary pixel point is respectively 0,255, rather than the value of boundary pixel point
It is 128, i.e., the pixel of the black and white is boundary pixel point, and the pixel of the grey is non-border pixel
Point.
In the step 4), by the three-valued image of the images to be recognized according in range of tilt angles, linear type side
The algorithm steps that boundary scans for are as follows:
4.1) it calculates straight incline offset of the tilt angle in [α, β] range and detects template;
4.2) to each boundary point in three-valued image, using straight-line detection template obtained in step a), check from
The boundary point sets out, boundary value is identical, and the continuum boundary line with stencil matching;
4.3) it to matched continuum boundary line, checks whether its length belongs to effective range, when belonging to effective range, then protects
It stays, otherwise the boundary line is noise margin, and sets c3 for the border pixel values that boundary points all on boundary line are included.
The step 4-1) in, the algorithm steps of straight-line detection template of the calculating tilt angle in [α, β] range
It is as follows:
4.1.1) the longest stroke length StrokeLen according to set by step 1) is calculated straight according to engineering practice
Length Len and the template maximum deviation value MaxOffY in y-direction of line template quantity Num, straight line template lines, meter
It is as follows to calculate formula difference;
4.1.2) to each straight incline offset template, the deviant in the direction y is calculated as follows.
The searching algorithm that the three-valued image of images to be recognized is carried out to linear boundary in the step 4.2), judge from
Certain boundary point sets out, boundary value is identical, and as follows with the method for the continuum boundary line of straight incline offset stencil matching:
4.2.1) when the generic linear that the line width of the straight line is 1, since starting point, from left to right check is this method
The no equivalent boundary for having direct neighbor, works as presence, then all boundary points belong to match point on the straight line, otherwise just belongs to not
Match point;
4.2.2) when the straight line line width is N, the general line based on eight connectivity, this method is since starting point, inspection
The equivalent boundary under eight connectivity meaning is surveyed, when there are such general lines, then all boundary points all belong on the general line
In match point, otherwise just belong to mismatch point.
The beneficial effects of the invention are as follows:A kind of character of picture OCR identifying system of the invention disappears with background adhesion noise
Except method, for the OCR technique of existing natural scene picture, the existing deficiency in terms of noise processed is proposed a kind of new
Adhesion noise cancellation method, in natural environment complicated and changeable, this method solve existing OCR application systems to adhesion
It the problem of noise character location difficulty, realizes the accurate extraction of character position, and is able to satisfy the requirement of real-time of engineer application.
Detailed description of the invention
Fig. 1 is the character and background adhesion noise cancellation method of a kind of picture OCR identifying system of the embodiment of the present invention
Flow chart;
Fig. 2 is the mean difference template of the embodiment of the present invention;
Fig. 3 is a kind of schematic diagram of scanning difference boundary general line of the embodiment of the present invention, the rectangular block of blank in figure
Indicate non-border pixel point, non-blank-white rectangular block indicates boundary pixel point in figure;
Fig. 4 is the process that the embodiment of the present invention carries out that linear type general line boundary is scanned for the three-valued image
Figure;
Fig. 5 is that the present invention calculates straight-line detection template flow chart of the tilt angle in [α, β] range;
Fig. 6 is the generic linear schematic diagram being made of in the embodiment of the present invention difference boundary pixel;
Fig. 7 be the present invention be general line judgement in use, based on the adjacent pixel schematic diagram under eight connectivity meaning;
Fig. 8 is the general line schematic diagram that width of the present invention is 2;
Fig. 9 is the general line schematic diagram that width of the present invention is 3;
Figure 10 is the picture that the embodiment of the present invention is applied to before the license plate processing that the 1st has adhesion;
Figure 11 is that the embodiment of the present invention is applied to the 1st license plate treated effect picture;
Figure 12 is the picture that the embodiment of the present invention is applied to before the license plate processing that the 2nd has adhesion;
Figure 13 is that the embodiment of the present invention is applied to the 2nd license plate treated effect picture;
Figure 14 is the picture that the embodiment of the present invention is applied to before the license plate processing that the 3rd has adhesion;
Figure 15 is that the embodiment of the present invention is applied to the 3rd license plate treated effect picture;
Figure 16 is the picture that the embodiment of the present invention is applied to before the license plate processing that the 4th has adhesion;
Figure 17 is that the embodiment of the present invention is applied to the 4th license plate treated effect picture;
Figure 18 is the picture that the embodiment of the present invention is applied to before the license plate processing that the 5th has adhesion;
Figure 19 is that the embodiment of the present invention is applied to the 5th license plate treated effect picture.
Specific embodiment
The present invention is described in further detail below in conjunction with the accompanying drawings.
As shown in Figure 1, for a kind of character of picture OCR identifying system of the embodiment of the present invention and background adhesion noise elimination side
The flow chart of method, the method includes:
1) the stroke length of character to be identified is arranged in the priori knowledge of the engineering-environment according to applied by OCR system
The effective range of StrokeLen.The parameter and actual engineering have close relationship, to be set as to match in actual application
The parameter set.When this method is applied to the OCR identification of bayonet license plate image, which is 35 pixels.In traffic electric
In the OCR identification process of sub- police's picture, which is 30 pixels.
2) image containing character to be identified under natural environment is chosen, and with mean difference template as shown in Figure 2
Mode calculates image difference boundary graph, and the difference boundary graph includes character zone and background area.
3) the difference boundary graph of the images to be recognized is subjected to thresholding, forms the three-valued boundary of images to be recognized
Figure, the images to be recognized includes region to be identified and background area, by the difference diagram of the images to be recognized according to following public affairs
Formula carries out three-valued:
Wherein, P (x, y) is the difference value of pixel (x, y) in difference diagram, and ParamA and ParamB are pre-set thresholds
The value range of value, c1, c2 and c3 are [0,255], and wherein c1 indicates boundary brighter around the brightness ratio of current pixel point, c2
Indicate boundary darker around the brightness relatively of current pixel point, c3 indicates the value on non-boundary.In engineering, adjust for convenience
Examination, the value of three constants should have bigger otherness as far as possible, and the c1 color is black, value c1=0;Institute
The c2 color stated is white, and value c2=255, the c3 color is grey, value c3=128, difference boundary pixel
The value of point is respectively 0,255, rather than the value of boundary pixel point is 128, i.e., the pixel of the black and white is boundary picture
Vegetarian refreshments, the pixel of the grey are non-boundary pixel points.
4) in the three-valued boundary graph of images to be recognized obtained above, according in range of tilt angles, linear type side
Boundary scans for, and checks whether the boundary has exceeded the effective range of stroke length, if without departing from the effective of stroke length
Range, then it is assumed that be that effective boundary point and giving retains, if exceeding the effective range of stroke length, which is
Noise margin removes it from three value boundary images, as shown in figure 3, wherein indicate detection process three times, be respectively from
Difference boundary pixel point A, C and E set out, and have detected the linear boundary in Num direction, in Fig. 3, there is two linear type difference
Boundary is AB and CD respectively, and E is then an isolated boundary point, if AB and CD exceeds the effective range of stroke length, just
Non- boundary value c2 is set by its color value, thus removes the linear boundary.
By above-mentioned steps treated difference diagram, it is just a cancellation the image of adhesion noise, can be answered at this time
In the follow-up process for using OCR character recognition.
In above process, the search of linear boundary is even more important, as shown in figure 4, being the present invention to three-valued image
The flow chart of linear boundary search is carried out, the process is as follows:
Step 4.1) calculates straight incline offset of the tilt angle in [α, β] range and detects template, involved here
Two parameters actually illustrate the range of tilt angles of OCR character, minimum cant is α, and maximum tilt angle is β,
Their selection has important influence to the elimination effect and performance of noise, to be set as configurable in actual application
Parameter, when this method is applied to the OCR identification of bayonet license plate image, parameter selection can be α=- 15 °, β=15 °, when answering
When using the OCR identification of traffic electronic police picture, it is proposed that parameter selection is α=- 25 °, β=25 °, is stopped when being applied to traffic and disobeying
Collect evidence picture OCR identification when, often bigger due to disobeying stop board tilt angle, parameter recommendation is selected as α=- 45 °, β=
45°;
Step 4.2) is to each boundary point in three-valued image, using the straight-line detection template described in step 4.1),
Inspection is identical from the boundary point, boundary value, and the continuum boundary line with stencil matching;
Step 4.3) checks whether its length belongs to effective range to matched continuum boundary line.If belonging to effective model
It encloses, then retains, otherwise the boundary line is noise margin, and the border pixel values for being included by boundary points all on boundary line are arranged
For c3, the length effective range of continuum boundary line described here refers to the effective range of stroke length.
Template is detected to aforesaid rectilinear, its purpose is to accelerate the speed of search, so that entire algorithm satisfaction is actually answered
To the requirement of real-time in, straight-line detection template has several, each template is directed to and answers some specific inclination angle,
Middle saved information is the offset in angled straight lines between each y-coordinate and horizontal line, and Fig. 5 is the process of line detection algorithm
Figure, concrete operations are as follows:
Step 4.1.1) according to longest stroke length StrokeLen set in the step 1), according to engineering reality
Situation calculate straight line template quantity Num, straight line template lines length Len and y maximum deviation value MaxOffY.It calculates public
Formula is as follows:
Step 4.1.2) to each straight incline offset template, the deviant in the direction y is calculated as follows.
The search of linear boundary is the key that this method in three-valued image, to each difference boundary point, from the point
It sets out, goes to match with straight-line detection template, check for the general line type boundary that continuous difference boundary pixel point is constituted,
General line is detected, as follows:
4.2.1) when the generic linear that the line width of the straight line is 1, since starting point, from left to right check is this method
Otherwise the no equivalent boundary for having direct neighbor just belongs to mismatch point, such as if it is present the boundary point belongs to match point
Shown in Fig. 6, under the definition of generic linear, there are two boundary straight lines in the figure, be AB and CD respectively.
4.2.2) when the line width of the straight line is N, the general line (as shown in Figure 7) based on eight connectivity, this method is from out
Hair point starts, and detects the equivalent difference boundary point within the scope of eight connectivity, if there is such boundary point, then the boundary point belongs to
Otherwise match point just belongs to mismatch point.The meaning of eight connectivity described here is as shown in fig. 7, if two difference boundary points
Meet positional relationship shown in it, is then that eight connectivity is adjacent at the two difference boundary points, according to this definition, such as Fig. 8,9
Shown, wherein just giving the example of two general lines, in fig. 8, general line AB has extended to B from starting point A, wide
Degree is 2, and horizontal length 13, in Fig. 9, general line CD has extended to D, width 3 from starting point C, and horizontal length is
16, defining the general line come in this manner has very high matching degree to the adhesion noise in engineering, will be such wide
Adopted straight line is disposed from difference boundary graph, can substantially reduce the influence of noise, particularly adhesion noise, and it is fixed to improve OCR character
The precision of position, in the application of Practical Project, parameter N is usually less than 4.
The embodiment of the present invention proposes the character and background adhesion noise cancellation method of a kind of picture OCR identifying system, according to
Engineering practice after setting corresponding parameter, is then calculated using mean difference template and obtains the three-valued difference of image
Component, finally, search meets the general line of correlated condition in all difference boundaries, which is exactly interference noise, is passed through
The relevant difference boundary point of the general line is eliminated, the noise information in picture to be identified is inhibited significantly, and finally to wait
Character zone is selected accurately to be positioned, to improve the final accuracy of OCR character recognition.
It is the effect picture that the present invention is applied in Car license recognition, there is shown the vehicles that 5 have adhesion shown in Figure 10-19
Board image, as shown in Figure 10, the reason of four character application adhesion noises among license plate are joined together, if according to biography
When connected component of uniting carries out License Plate, the connected component that this four characters are formed will be considered as an entirety, and apply
After the present invention, as shown in figure 11, this four characters are fully segmented out, to be conducive to the positioning of license plate, as Figure 12,14,
16, described in 18, these license plates are also similar effect, that is, after present invention processing, as shown in Figure 13,15,17,19,
Effectively adhesion noise can be eliminated, enable the specific dividing processing of character therein, to improve the knowledge of system
Not rate.
The character of picture OCR identifying system of the present embodiment a kind of and background adhesion noise cancellation method, for it is existing from
The OCR technique of right scene picture, the existing deficiency in terms of noise processed propose a kind of new adhesion noise cancellation method,
In natural environment complicated and changeable, this method solve existing OCR application systems to adhesion noise character location difficulty
The problem of, realize the accurate extraction of character position, and be able to satisfy the requirement of real-time of engineer application.
It is of the art it will be recognized by the skilled artisan that above embodiments of the present invention are only of the invention preferred
One of embodiment limits for length, cannot enumerate all embodiments one by one here, any to embody the claims in the present invention
The implementation of technical solution, all within the scope of the present invention.
It should be noted that the above content is combine specific embodiment made for the present invention further specifically
It is bright, and it cannot be said that a specific embodiment of the invention is only limitted to this, under above-mentioned guidance of the invention, those skilled in the art can
To carry out various improvement and deformations on the basis of the above embodiments, and these are improved or deformation falls in protection model of the invention
In enclosing.
Claims (6)
1. a kind of character of picture OCR identifying system and background adhesion noise cancellation method, it is characterised in that:The picture
The character and background adhesion noise cancellation method of OCR identifying system include as follows:
1) the stroke length StrokeLen of character to be identified is arranged in the priori knowledge of the engineering-environment according to applied by OCR system
Effective range;
2) image containing character to be identified under natural environment is chosen, and calculates image difference in a manner of mean difference template
Boundary graph, the difference boundary graph include character zone and background area;
3) the difference boundary graph of images to be recognized obtained above is subjected to thresholding, forms the three-valued boundary of images to be recognized
Figure, the three-valued boundary graph include region to be identified and background area;
4) in the three-valued boundary graph of images to be recognized obtained above, according in range of tilt angles, linear boundary into
Row search, checks whether the boundary has exceeded the effective range of stroke length, when without departing from the effective range of stroke length, then
It is considered effective boundary point, and gives and retain, when the effective range for exceeding stroke length, then the edge determination is noise, will
It is removed from three value boundary images.
2. a kind of character of picture OCR identifying system according to claim 1 and background adhesion noise cancellation method, special
Sign is:In the step 3), the formula that the difference boundary graph of the images to be recognized is carried out thresholding is as follows:
Wherein, P (x, y) is the difference value of pixel (x, y) in difference diagram, and ParamA and ParamB are pre-set threshold values,
C1, c2 and c3 are signless integers, and value range is [0,255], and wherein c1 indicates brighter around the brightness ratio of current pixel point
Boundary, c2 indicates darker boundary around the brightness relatively of current pixel point, and c3 indicates the value on non-boundary.
3. a kind of character of picture OCR identifying system according to claim 2 and background adhesion noise cancellation method, special
Sign is:The c1 color is black, value c1=0;The c2 color is white, and value c2=255 is described
C3 color is grey, and value c3=128, the value of difference boundary pixel point is respectively 0,255, rather than the value of boundary pixel point is
128, i.e. the pixel of the black and white is boundary pixel point, and the pixel of the grey is non-boundary pixel point.
4. a kind of character of picture OCR identifying system according to claim 1 and background adhesion noise cancellation method, special
Sign is:In the step 4), by the three-valued image of the images to be recognized according in range of tilt angles, linear boundary
The algorithm steps scanned for are as follows:
4.1) it calculates straight incline offset of the tilt angle in [α, β] range and detects template;
4.2) to each boundary point in three-valued image, mould is detected using straight incline offset obtained in step 4.1)
Version, inspection is identical from the boundary point, boundary value, and the continuum boundary line with stencil matching;
4.3) it to matched continuum boundary line, checks whether its length belongs to effective range, when belonging to effective range, then retains,
Otherwise the boundary line is noise margin, and sets c3 for the border pixel values that boundary points all on boundary line are included.
5. a kind of character of picture OCR identifying system according to claim 4 and background adhesion noise cancellation method, special
Sign is:The step 4-1) in, the algorithm steps of straight-line detection template of the calculating tilt angle in [α, β] range are such as
Under:
4.1.1 it) according to longest stroke length StrokeLen set in the step 1), is calculated according to engineering practice
Straight line template quantity Num, straight line template lines length Len and y maximum deviation value MaxOffY;Calculation formula is as follows:
4.1.2) to each straight incline offset template, the deviant in the direction y is calculated as follows:
6. a kind of character of picture OCR identifying system according to claim 4 and background adhesion noise cancellation method, special
Sign is:The searching algorithm that the three-valued image of images to be recognized is carried out to linear boundary in the step 4.2), judge from
Certain boundary point sets out, boundary value is identical, and as follows with the method for the continuum boundary line of straight incline offset stencil matching:
4.2.1) when the generic linear that the line width of the straight line is 1, this method is from left to right checked whether there is since starting point
The equivalent boundary of direct neighbor works as presence, then all boundary points belong to match point on the straight line, otherwise just belongs to mismatch
Point;
4.2.2) when the straight line line width is N, the general line based on eight connectivity, this method is since starting point, detection eight
The equivalent boundary being connected under meaning, when there are such general lines, then all boundary points belong on the general line
With point, otherwise just belong to mismatch point.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510776907.1A CN105447489B (en) | 2015-11-13 | 2015-11-13 | A kind of character of picture OCR identifying system and background adhesion noise cancellation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510776907.1A CN105447489B (en) | 2015-11-13 | 2015-11-13 | A kind of character of picture OCR identifying system and background adhesion noise cancellation method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105447489A CN105447489A (en) | 2016-03-30 |
CN105447489B true CN105447489B (en) | 2018-11-16 |
Family
ID=55557649
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510776907.1A Active CN105447489B (en) | 2015-11-13 | 2015-11-13 | A kind of character of picture OCR identifying system and background adhesion noise cancellation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105447489B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109389024B (en) * | 2018-01-30 | 2020-09-04 | 长城汽车股份有限公司 | Method and device for identifying road cone based on image, storage medium and vehicle |
CN109993761B (en) * | 2018-06-29 | 2021-04-09 | 长城汽车股份有限公司 | Ternary image acquisition method and device and vehicle |
CN111553317B (en) * | 2020-05-14 | 2023-08-08 | 北京惠朗时代科技有限公司 | Anti-fake code acquisition method and device, computer equipment and storage medium |
CN112560866B (en) * | 2021-02-25 | 2021-05-04 | 江苏东大集成电路系统工程技术有限公司 | OCR recognition method based on background suppression |
US11615634B2 (en) | 2021-07-01 | 2023-03-28 | International Business Machines Corporation | Character recognition of license plate under complex background |
CN113822815B (en) * | 2021-09-24 | 2024-02-06 | 广州光锥元信息科技有限公司 | Method and apparatus for high performance picture clutter removal using GPU rendering |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1928891A (en) * | 2005-09-05 | 2007-03-14 | 富士通株式会社 | Method and device for assigned text line extraction |
CN101599124A (en) * | 2008-06-03 | 2009-12-09 | 汉王科技股份有限公司 | A kind of from video image the method and apparatus of separating character |
CN104112130A (en) * | 2014-06-26 | 2014-10-22 | 小米科技有限责任公司 | Optical character recognition method and device |
CN104361336A (en) * | 2014-11-26 | 2015-02-18 | 河海大学 | Character recognition method for underwater video images |
EP2899671A1 (en) * | 2012-09-21 | 2015-07-29 | ZTE Corporation | Identification method and device for target object in image |
-
2015
- 2015-11-13 CN CN201510776907.1A patent/CN105447489B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1928891A (en) * | 2005-09-05 | 2007-03-14 | 富士通株式会社 | Method and device for assigned text line extraction |
CN101599124A (en) * | 2008-06-03 | 2009-12-09 | 汉王科技股份有限公司 | A kind of from video image the method and apparatus of separating character |
EP2899671A1 (en) * | 2012-09-21 | 2015-07-29 | ZTE Corporation | Identification method and device for target object in image |
CN104112130A (en) * | 2014-06-26 | 2014-10-22 | 小米科技有限责任公司 | Optical character recognition method and device |
CN104361336A (en) * | 2014-11-26 | 2015-02-18 | 河海大学 | Character recognition method for underwater video images |
Also Published As
Publication number | Publication date |
---|---|
CN105447489A (en) | 2016-03-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105447489B (en) | A kind of character of picture OCR identifying system and background adhesion noise cancellation method | |
CN108460757B (en) | Mobile phone TFT-LCD screen Mura defect online automatic detection method | |
CN110163219B (en) | Target detection method based on image edge recognition | |
CN109242853B (en) | PCB defect intelligent detection method based on image processing | |
US8712114B2 (en) | Elegant solutions for fingerprint image enhancement | |
CN109472788B (en) | Method for detecting flaw on surface of airplane rivet | |
CN108918093B (en) | Optical filter mirror surface defect detection method and device and terminal equipment | |
CN111476804A (en) | Method, device and equipment for efficiently segmenting carrier roller image and storage medium | |
CN110674812A (en) | Civil license plate positioning and character segmentation method facing complex background | |
CN114037691A (en) | Carbon fiber plate crack detection method based on image processing | |
CN114127784A (en) | Method, computer program product and computer readable medium for generating a mask for a camera stream | |
CN112528868A (en) | Illegal line pressing judgment method based on improved Canny edge detection algorithm | |
Ashoori-Lalimi et al. | An efficient method for vehicle license plate detection in complex scenes | |
CN110807763A (en) | Method and system for detecting ceramic tile surface bulge | |
CN117094975A (en) | Method and device for detecting surface defects of steel and electronic equipment | |
Lin et al. | License plate recognition based on mathematical morphology and template matching | |
CN109800641B (en) | Lane line detection method based on threshold value self-adaptive binarization and connected domain analysis | |
CN113971681A (en) | Edge detection method for belt conveyor in complex environment | |
CN113283439A (en) | Intelligent counting method, device and system based on image recognition | |
CN106530292B (en) | A kind of steel strip surface defect image Fast Identification Method based on line scan camera | |
Wang et al. | Adaptive binarization: A new approach to license plate characters segmentation | |
CN115359449B (en) | Automatic identification method and system for turnout notch image of point switch | |
Lin et al. | Locating license plate based on edge features of intensity and saturation subimages | |
Nain et al. | Dynamic thresholding based edge detection | |
CN113505811A (en) | Machine vision imaging method for hub production |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |