CN103824073A - Image information extraction method - Google Patents
Image information extraction method Download PDFInfo
- Publication number
- CN103824073A CN103824073A CN201210465474.4A CN201210465474A CN103824073A CN 103824073 A CN103824073 A CN 103824073A CN 201210465474 A CN201210465474 A CN 201210465474A CN 103824073 A CN103824073 A CN 103824073A
- Authority
- CN
- China
- Prior art keywords
- connected domain
- character
- image information
- template
- 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.)
- Pending
Links
- 238000000605 extraction Methods 0.000 title abstract description 3
- 238000000034 method Methods 0.000 claims description 37
- 230000008034 disappearance Effects 0.000 claims description 27
- 239000013589 supplement Substances 0.000 claims description 12
- 238000002372 labelling Methods 0.000 claims description 11
- 238000011002 quantification Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000009415 formwork Methods 0.000 claims description 3
- 230000001502 supplementing effect Effects 0.000 claims description 2
- 238000012216 screening Methods 0.000 abstract 1
- 238000004458 analytical method Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000009182 swimming Effects 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides an image information extraction method which comprises the steps of establishing an image information template, marking connected domains of a binary image, analyzing and screening the connected domains according to the image information template, determining a seed connected domain and the position thereof according to the image information template, adding similar connected domains according to the seed connected domain, and adding the remaining connected characters according to the detected connected domains and the template. By adopting the technical scheme provided by the invention, position information of characters can still be accurately extracted under the situation in which layout interference of a license plate image is complex.
Description
Technical field
The present invention relates to a kind of abstracting method of image information, be specifically related to the abstracting method of license board information in license plate image, belong to the technical fields such as image processing and pattern-recognition.
Background technology
In control of traffic and road work, car plate identification is one of them important step.The main method of at present car plate identification has the analytical approach based on textural characteristics, the method based on rim detection, method, the method based on image color information and the recognition methods based on wavelet analysis based on mathematical morphology location etc.Car plate identification comprises the several parts of car plate location, perspective correction, binaryzation, printed page analysis and OCR, in actual application environment, because car plate long term exposure is in the external world, may be subject to the pollution of external environment, produce the phenomenons such as car plate local color inequality, character breakage; Due to human factor, may there are some ornaments in car plate outer rim in addition, and these ornaments also can produce certain impact to the space of a whole page of license plate image; In addition, while utilizing collecting device to gather image, image can be subject to the impact of the factors such as extraneous objective condition (as illumination, shooting orientation etc.) and equipment quality, there are differences through the page image of overcorrect and the space of a whole page of standard.Therefore, accurately number-plate number character information is extracted by printed page analysis, and the connected domain of disappearance is supplemented is to obtain the steps necessary of license board information accurately.
But prior art generally adopts vertically and horizontal projection method carries out cutting and the extraction of character.The shortcoming of this technology is in the time that the interference region of characters on license plate and periphery produces adhesion or character generation fracture, can produce cutting mistake; Sciagraphy also cannot be processed the license plate image of inclination simultaneously, thereby Obtaining Accurate and identification to license board information have caused adverse effect.
Summary of the invention
The problem existing for prior art, the object of the invention is, for this step of printed page analysis, two-value license plate image to be processed, and the abstracting method of license board information in a kind of license plate image is provided.Target of the present invention is that the connected domain corresponding number-plate number is extracted from the license plate image space of a whole page, and the character connected domain of disappearance is supplemented.Can accurately identify by OCR with the character of guaranteeing to extract.
For achieving the above object, the present invention carries out printed page analysis to two-value license plate image, and the connected domain corresponding number-plate number is extracted from the license plate image space of a whole page, and it is as described below that the present invention solves the problems of the technologies described above adopted technical scheme:
An abstracting method for image information, comprises the following steps: (1) sets up image information template; (2) bianry image is carried out to connected component labeling; (3) according to image information template, connected domain is analyzed and screened; (4) determine kind of Sub Connected and position thereof according to image information template; (5) supplement similar connected domain according to kind of Sub Connected; (6) according to detecting connected domain and template, supplement residue and be communicated with character.
Preferably, described image is license plate image, and described information is number-plate number information.
Further, described similar connected domain be height with the absolute value difference of height of kind of Sub Connected within the specific limits and/or central point coordinate and the poor connected domain within the specific limits of centre coordinate of planting Sub Connected in y direction.
Alternatively, the license board information template that step (1) is set up comprises wide, the high information of standard car plate, wide, the high information of standard characters on license plate and the distributed intelligence of each character on car plate.The mode of in step (2), two-value license plate image being carried out to connected component labeling is for to carry out connected component labeling according to character color to two-value license plate image.
The step of more preferably, in step (3), connected domain being analyzed and being screened comprises:
(3-1) according to the following parameter of license board information formwork calculation: ratio, character duration that character height accounts for car plate height account for the ratio of car plate width, the ratio of width to height of character; Fall obviously not meet the connected domain of characters on license plate font style characteristic according to as above 3 choice of parameters;
(3-2) remove and disturb up and down connected domain according to position relationship, determine upper bound y0 and the lower bound y1 of number-plate number connected domain, described y0 and y1 are in such coordinate system: characters on license plate from left to right direction is x axle, are axially y axle down perpendicular to x, and true origin is in the car plate upper left corner.
Described step (3-2) further comprises the steps:
(3-2-1) projection vector of the width of calculating connected domain boundary rectangle in y direction; Computing method are: for any point curY on y axle, draw a horizontal line by curY, the width sum of all connected domains that its projection value passes through for this horizontal line;
(3-2-2) projection vector is quantized, carry out judging for the each element in projection vector: in the time of element >T in projection vector, be 1 by this Quantification of elements, otherwise be quantified as the maximal value in 0, T=k* projection vector;
(3-2-3) in the projection vector after quantification, find maximum run, the starting point of maximum run is y0, and terminal is y1, and the bound of number-plate number connected domain is identical with the starting point of maximum run;
(3-2-4) remove ineligible connected domain, if connected domain is at [y0, y1] outside scope or connected domain part at [y0, y1] in scope and within the connected domain of ratio <m, this connected domain is removed, described ratio is the ratio that the part of connected domain within [y0, y1] accounts for whole connected domain.
Further, the definite kind Sub Connected of step (4) is the connected domain that meets character glyphs feature in template most, the ratio of width to height of connected domain of planting Sub Connected is the most close with standard characters on license plate the ratio of width to height, i.e. the connected domain of the absolute value minimum of the difference of the ratio of width to height of connected domain and standard characters on license plate the ratio of width to height.
In step (5), judge that 2 similar conditions of connected domain are: for connected domain R1 and connected domain R2, absolute value <HeiDiff, the R1 of difference of R1 and R2 height and the absolute value <YcenterDiff of the difference of R2 central point y coordinate, wherein 0.02*H2<HeiDiff<0.2*H2,0.02*H2<YcenterDiff<0.2*H2, the height that H2 is R2.
Further, the step that in described step (6), supplementary residue is communicated with character comprises the step of supplementing residue simply connected character, and/or the supplementary residue characters that are communicated with more.
Above-mentioned supplementary residue simply connected character has further comprised following steps:
(6-1) first determine between two connected domains whether have disappearance connected domain according to template, if disappearance is further estimated the number of disappearance connected domain, then repeat (6-2) to (6-3), until all disappearance simply connected domains of completion;
(6-2) according to template, estimate the position of disappearance connected domain, in the center range of this position, find black pixel as Seed Points, this center range is preferably with center, central point position, highly tapers to original N1 doubly, and width reduction to N2 doubly, utilize region-growing method, between y0-y1, grow a connected domain, wherein 0.4<N1<0.8,0.2<N2<0.6;
If (6-3) font style characteristic of this connected domain is similar to two connected domains step (6-1) Suo Shu, set it as a number-plate number connected domain; Otherwise demarcate disappearance number connected domain according to template.
Described supplementary multiply connected domain character has further comprised following steps:
(7-1) judge whether to exist many connections character of disappearance according to information of vehicles template and the connected domain that detected, if existed, repeat (7-2) to (7-3), until all disappearances of completion are communicated with character more.
(7-2) demarcate according to information model the boundary rectangle iniRect that lacks connected domain;
(7-3) boundary rectangle is verified, the method that boundary rectangle is verified is: first find out all connected domains of central point in demarcation connected domain in initial connected domain array, then ask the boundary rectangle mergeRect of the union of these connected domains, then the upper bound of mergeRect is set to y0, lower bound is set to y1, if the font style characteristic of mergeRect is similar to kind Sub Connected, set it as a number-plate number connected domain, otherwise using iniRect as a number-plate number connected domain.
Effect of the present invention is: utilize technical scheme provided by the invention, by extracting connected domain corresponding to the number-plate number, and the character connected domain of disappearance is supplemented, can disturb comparatively and still the positional information of character can be extracted comparatively accurately under complicated situation at the space of a whole page.
Accompanying drawing explanation
Below in conjunction with embodiment and accompanying drawing thereof, the invention will be further described.
Fig. 1 is the invention process steps flow chart schematic diagram.
Fig. 2 is the view of in the present invention, two-value license plate image being carried out connected component labeling.
Fig. 3 is the step schematic diagram that connected domain is analyzed and screened.
Fig. 4 is the method for the projection vector of width in y direction that calculate connected domain boundary rectangle in the present invention;
Fig. 5 plants Sub Connected schematic diagram in the present invention.
Fig. 6 supplements similar connected domain schematic diagram.
Fig. 7 is supplementary schedule connected domain schematic diagram.
Fig. 8 supplements multiply connected domain schematic diagram.
Embodiment
Below in conjunction with Figure of description, the invention will be further described.
First the present invention provides a kind of image information abstracting method, Fig. 1 is the first embodiment of the present invention, this embodiment has described the method that extracts license board information from Chinese car license plate image, by two-value license plate image is carried out to printed page analysis, and the connected domain corresponding number-plate number is extracted from the license plate image space of a whole page.Particularly, this method that extract license board information from license plate image comprises the following steps:
(1) set up image information template.In the embodiment shown in fig. 1, set up license board information template according to standard car plate.
(2) bianry image is carried out to connected component labeling; In the embodiment shown in fig. 1, bianry image is two-value license plate image.
(3) connected domain is analyzed and screened;
(4) determine kind of Sub Connected and position thereof according to template;
(5) supplement similar connected domain according to kind of Sub Connected;
(6), according to detecting connected domain and template, supplement residue simply connected character;
(7), according to detecting connected domain and template, supplement the residue characters that are communicated with more.
In the present embodiment, the license board information template that step (1) is set up comprises wide, the high information of standard car plate, wide, the high information of standard characters on license plate and the distributed intelligence of each character on car plate.
The mode of in step (2), two-value license plate image being carried out to connected component labeling is for to carry out connected component labeling according to character color to two-value license plate image.As shown in Figure 2, this bianry image is white gravoply, with black engraved characters, black picture element is carried out to connected component labeling.
As shown in Figure 3, in a more detailed embodiment, the step of in step (3), connected domain being analyzed and being screened comprises:
(3-1) according to the following parameter of license board information formwork calculation: ratio, character duration that character height accounts for car plate height account for the ratio of car plate width, the ratio of width to height of character; Fall obviously not meet the connected domain of characters on license plate font style characteristic according to as above 3 choice of parameters.
(3-2) remove and disturb up and down connected domain according to position relationship, determine upper bound y0 and the lower bound y1 of number-plate number connected domain.Described y0 and y1 are in such coordinate system: characters on license plate from left to right direction is x axle, are axially y axle down perpendicular to x, and true origin is in the car plate upper left corner.
In the present embodiment, the method that step (3-2) adopts is that the width of connected domain boundary rectangle is in the method for y direction projection.Comprise the steps:
(3-2-1) projection vector of the width of calculating connected domain boundary rectangle in y direction; Computing method are: for any point curY on y axle, draw a horizontal line by curY, the width sum of all connected domains that its projection value passes through for this horizontal line.As shown in Figure 4, the width of 4 connected domains is respectively W1, W2, W3 and W4, and the horizontal line that curY is corresponding has passed through front 3 connected domains, and therefore, the projection value that curY is corresponding equals W1+W2+W3.
(3-2-2) projection vector is quantized, carry out judging for the each value in projection vector (i.e. an element): in the time of element >T in projection vector, be 1 by this Quantification of elements, otherwise be quantified as 0, maximal value in T=k* projection vector, wherein 0.4<k<0.9.
(3-2-3) in the projection vector after quantification, find maximum run (distance of swimming refers to continuous non-zero string), the starting point of maximum run is y0, and terminal is y1, and the bound of number-plate number connected domain is identical with the starting point of maximum run.
(3-2-4) remove ineligible connected domain, if connected domain is at [y0, y1] outside scope or connected domain part at [y0, y1] in scope and within the connected domain of ratio <m, this connected domain is removed, described ratio is the ratio that the part of connected domain within [y0, y1] accounts for whole connected domain, wherein, according to practical situations, the preferred span of m is 0.6<m<0.9.
The definite kind Sub Connected of step (4) of the present embodiment is the connected domain that meets character glyphs feature in template most.Be similar connected domain be height with the absolute value difference of height of kind of Sub Connected within the specific limits and/or central point coordinate and the poor connected domain within the specific limits of centre coordinate of planting Sub Connected in y direction.As shown in Figure 5.
As shown in Figure 6, in the step (5) of the present embodiment, supplement similar connected domain according to kind of Sub Connected, described similar connected domain is height and central point feature and plant the similar connected domain of Sub Connected, i.e. the less connected domain of absolute value of height, central point less with the absolute value of the difference of the height of kind of Sub Connected center difference of coordinate in y direction of coordinate and kind Sub Connected in y direction.Judging whether 2 connected domains are similar can change decision conditions according to the different car plates of processing, for example, the connected domain R1 decision condition similar to connected domain R2 is as follows: i.e. R1 and absolute value <HeiDiff, the R1 of difference of R2 height and the absolute value <YcenterDiff of the difference of R2 central point y coordinate.Wherein 0.02*H2<HeiDiff<0.2*H2,0.02*H2<YcenterDiff<0.2*H2, the height that H2 is R2.
In the present embodiment, step (6), according to detecting connected domain and template, is supplemented residue simply connected character and has further been comprised following steps:
(6-1) first determine between two connected domains whether have disappearance connected domain according to template, if disappearance is further estimated the number of disappearance connected domain, then repeat (6-2) to (6-3), until all disappearance simply connected domains of completion;
(6-2) according to template, estimate the position of disappearance connected domain, in the center range of this position, find black pixel as Seed Points, this center range is preferably with center, central point position, highly tapers to original N1 doubly, and width reduction to N2 doubly, utilize region-growing method, between y0-y1, grow a connected domain, wherein 0.4<N1<0.8,0.2<N2<0.6;
If (6-3) font style characteristic of this connected domain is similar to two connected domains step (6-1) Suo Shu, set it as a number-plate number connected domain; Otherwise demarcate disappearance number connected domain according to template.Supplement the state of residue simply connected character as shown in Figure 7.
In the present embodiment, step (7) is supplemented multiply connected domain character and has further been comprised following steps:
(7-1) judge whether to exist many connections character of disappearance according to information of vehicles template and the connected domain that detected.If existed, repeat (7-2) to (7-3), until all disappearances of completion are communicated with character more.
(7-2) demarcate according to information model the boundary rectangle iniRect that lacks connected domain;
(7-2) boundary rectangle is verified.The method that boundary rectangle is verified is: first find out all connected domains of central point in demarcation connected domain in initial connected domain array, then ask the boundary rectangle mergeRect of the union of these connected domains, then the upper bound of mergeRect is set to y0, lower bound is set to y1, if the font style characteristic of mergeRect is similar to kind Sub Connected, set it as a number-plate number connected domain, otherwise using iniRect as a number-plate number connected domain.
Supplement the state of the many connections of residue character as shown in Figure 8.
Can find out by the present embodiment, in the license plate image that the present invention proposes, the abstracting method of license board information, it is advantageous that: disturb comparatively and still the positional information of character can be extracted comparatively accurately under complicated situation at the space of a whole page.
The embodiment being not limited to described in embodiment of the present invention, as long as those skilled in the art draw other embodiment according to the present invention program, belongs to the scope of technological innovation of the present invention and protection equally.
Claims (12)
1. an abstracting method for image information, comprises the following steps:
(1) set up image information template;
(2) bianry image is carried out to connected component labeling;
(3) according to image information template, connected domain is analyzed and screened;
(4) determine kind of Sub Connected and position thereof according to image information template;
(5) supplement similar connected domain according to kind of Sub Connected;
(6) according to detecting connected domain and template, supplement residue and be communicated with character.
2. the abstracting method of image information as claimed in claim 1, is characterized in that: described image is license plate image, and described information is number-plate number information.
3. the abstracting method of image information as claimed in claim 1, is characterized in that: described similar connected domain be height with the absolute value difference of height of kind of Sub Connected within the specific limits and/or central point coordinate and the poor connected domain within the specific limits of centre coordinate of planting Sub Connected in y direction.
4. the abstracting method of image information as claimed in claim 2, is characterized in that: the license board information template that step (1) is set up comprises wide, the high information of standard car plate, wide, the high information of standard characters on license plate and the distributed intelligence of each character on car plate.
5. the abstracting method of image information as claimed in claim 2, is characterized in that: the mode of in step (2), two-value license plate image being carried out to connected component labeling is for to carry out connected component labeling according to character color to two-value license plate image.
6. the abstracting method of image information as claimed in claim 2, is characterized in that: the step of in step (3), connected domain being analyzed and being screened comprises:
(3-1) according to the following parameter of license board information formwork calculation: ratio, character duration that character height accounts for car plate height account for the ratio of car plate width, the ratio of width to height of character; Fall obviously not meet the connected domain of characters on license plate font style characteristic according to as above 3 choice of parameters;
(3-2) remove and disturb up and down connected domain according to position relationship, determine upper bound y0 and the lower bound y1 of number-plate number connected domain, described y0 and y1 are in such coordinate system: characters on license plate from left to right direction is x axle, are axially y axle down perpendicular to x, and true origin is in the car plate upper left corner.
7. the abstracting method of image information as claimed in claim 6, is characterized in that:
Described step (3-2) further comprises the steps:
(3-2-1) projection vector of the width of calculating connected domain boundary rectangle in y direction;
(3-2-2) projection vector is quantized, carry out judging for the each element in projection vector: in the time of element >T in projection vector, be 1 by this Quantification of elements, otherwise be quantified as the maximal value in 0, T=k* projection vector;
(3-2-3) in the projection vector after quantification, find maximum run, the starting point of maximum run is y0, and terminal is y1, and the bound of number-plate number connected domain is identical with the starting point of maximum run;
(3-2-4) remove ineligible connected domain, if connected domain is at [y0, y1] outside scope or connected domain part at [y0, y1] in scope and within the connected domain of ratio <m, this connected domain is removed, described ratio is the ratio that the part of connected domain within [y0, y1] accounts for whole connected domain.
8. the abstracting method of image information as claimed in claim 2, it is characterized in that: the definite kind Sub Connected of step (4) is the connected domain that meets character glyphs feature in template most, i.e. the connected domain of the absolute value minimum of the difference of the ratio of width to height of connected domain and standard characters on license plate the ratio of width to height.
9. the abstracting method of image information as claimed in claim 2, it is characterized in that: in step (5), judge that 2 similar conditions of connected domain are: for connected domain R1 and connected domain R2, absolute value <HeiDiff, the R1 of difference of R1 and R2 height and the absolute value <YcenterDiff of the difference of R2 central point y coordinate, wherein 0.02*H2<HeiDiff<0.2*H2,0.02*H2<YcenterDiff<0.2*H2, the height that H2 is R2.
10. the abstracting method of image information as claimed in claim 1, is characterized in that:
The step that in described step (6), supplementary residue is communicated with character comprises the step of supplementing residue simply connected character, and/or the supplementary residue characters that are communicated with more.
The abstracting method of 11. image informations as claimed in claim 10, is characterized in that: described supplementary residue simply connected character has further comprised following steps:
(6-1) first determine between two connected domains whether have disappearance connected domain according to template, if disappearance is further estimated the number of disappearance connected domain, then repeat (6-2) to (6-3), until all disappearance simply connected domains of completion;
(6-2) according to template, estimate the position of disappearance connected domain, in the center range of this position, find black pixel as Seed Points, this center range is preferably with center, central point position, highly taper to original N1 doubly, width reduction to N2 doubly, is utilized region-growing method, grows a connected domain between y0-y1;
If (6-3) font style characteristic of this connected domain is similar to two connected domains step (6-1) Suo Shu, set it as a number-plate number connected domain; Otherwise demarcate disappearance number connected domain according to template.
The abstracting method of 12. image informations as claimed in claim 10, is characterized in that: described supplementary multiply connected domain character has further comprised following steps:
(7-1) judge whether to exist many connections character of disappearance according to information of vehicles template and the connected domain that detected.If existed, repeat (7-2) to (7-3), until all disappearances of completion are communicated with character more.
(7-2) demarcate according to information model the boundary rectangle iniRect that lacks connected domain;
(7-3) boundary rectangle is verified, the method that boundary rectangle is verified is: first find out all connected domains of central point in demarcation connected domain in initial connected domain array, then ask the boundary rectangle mergeRect of the union of these connected domains, then the upper bound of mergeRect is set to y0, lower bound is set to y1, if the font style characteristic of mergeRect is similar to kind Sub Connected, set it as a number-plate number connected domain, otherwise using iniRect as a number-plate number connected domain.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210465474.4A CN103824073A (en) | 2012-11-16 | 2012-11-16 | Image information extraction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210465474.4A CN103824073A (en) | 2012-11-16 | 2012-11-16 | Image information extraction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103824073A true CN103824073A (en) | 2014-05-28 |
Family
ID=50759125
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210465474.4A Pending CN103824073A (en) | 2012-11-16 | 2012-11-16 | Image information extraction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103824073A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104200210A (en) * | 2014-08-12 | 2014-12-10 | 合肥工业大学 | License plate character segmentation method based on parts |
| CN104680531A (en) * | 2015-02-28 | 2015-06-03 | 西安交通大学 | Connection flux statistical information extraction method and VLSI structure |
| WO2018040342A1 (en) * | 2016-08-31 | 2018-03-08 | 百度在线网络技术（北京）有限公司 | Method and device for identifying text area in image |
| CN113128229A (en) * | 2021-04-14 | 2021-07-16 | 河海大学 | Chinese entity relation joint extraction method |
-
2012
- 2012-11-16 CN CN201210465474.4A patent/CN103824073A/en active Pending
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104200210A (en) * | 2014-08-12 | 2014-12-10 | 合肥工业大学 | License plate character segmentation method based on parts |
| CN104200210B (en) * | 2014-08-12 | 2018-11-06 | 合肥工业大学 | A kind of registration number character dividing method based on component |
| CN104680531A (en) * | 2015-02-28 | 2015-06-03 | 西安交通大学 | Connection flux statistical information extraction method and VLSI structure |
| WO2018040342A1 (en) * | 2016-08-31 | 2018-03-08 | 百度在线网络技术（北京）有限公司 | Method and device for identifying text area in image |
| US10803338B2 (en) | 2016-08-31 | 2020-10-13 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method and device for recognizing the character area in a image |
| CN113128229A (en) * | 2021-04-14 | 2021-07-16 | 河海大学 | Chinese entity relation joint extraction method |
| CN113128229B (en) * | 2021-04-14 | 2023-07-18 | 河海大学 | Chinese entity relation joint extraction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109284758B (en) | Invoice seal eliminating method and device and computer storage medium | |
| CN108364010B (en) | License plate recognition method, device, equipment and computer readable storage medium | |
| CN103198315B (en) | Based on the Character Segmentation of License Plate of character outline and template matches | |
| CN103310211B (en) | A kind ofly fill in mark recognition method based on image procossing | |
| CN104156704A (en) | Novel license plate identification method and system | |
| CN107301405A (en) | Method for traffic sign detection under natural scene | |
| CN103824091A (en) | Vehicle license plate recognition method for intelligent transportation system | |
| CN104050684B (en) | A kind of video frequency motion target sorting technique based on on-line training and system | |
| CN103679168A (en) | Detection method and detection device for character region | |
| CN102419820A (en) | Method for rapidly detecting car logo in videos and images | |
| CN102542244A (en) | Face detection method and system and computer program product | |
| CN104766344B (en) | Vehicle checking method based on movement edge extractor | |
| Xia et al. | The study of license plate character segmentation algorithm based on vetical projection | |
| CN104299009A (en) | Plate number character recognition method based on multi-feature fusion | |
| CN102184404B (en) | Method and device for acquiring palm region in palm image | |
| CN104200228A (en) | Recognizing method and system for safety belt | |
| CN104036244A (en) | Checkerboard pattern corner point detecting method and device applicable to low-quality images | |
| CN103824073A (en) | Image information extraction method | |
| CN109086772A (en) | A kind of recognition methods and system distorting adhesion character picture validation code | |
| CN106845482A (en) | A kind of license plate locating method | |
| CN106355180A (en) | Method for positioning license plates on basis of combination of color and edge features | |
| CN110633635A (en) | ROI-based traffic sign board real-time detection method and system | |
| CN106960196A (en) | Industrial video decimal fractions recognition methods based on template matches and SVM | |
| CN103699876A (en) | Method and device for identifying vehicle number based on linear array CCD (Charge Coupled Device) images | |
| CN105512662A (en) | Detection method and apparatus for unlicensed vehicle |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20140528 |