CN108806059A - The text filed localization method of the bill alignment and eight neighborhood connected component offset correction of feature based point - Google Patents
The text filed localization method of the bill alignment and eight neighborhood connected component offset correction of feature based point Download PDFInfo
- Publication number
- CN108806059A CN108806059A CN201810434069.3A CN201810434069A CN108806059A CN 108806059 A CN108806059 A CN 108806059A CN 201810434069 A CN201810434069 A CN 201810434069A CN 108806059 A CN108806059 A CN 108806059A
- Authority
- CN
- China
- Prior art keywords
- connected component
- bill
- offset correction
- neighborhood
- alignment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07D—HANDLING OF COINS OR VALUABLE PAPERS, e.g. TESTING, SORTING BY DENOMINATIONS, COUNTING, DISPENSING, CHANGING OR DEPOSITING
- G07D7/00—Testing specially adapted to determine the identity or genuineness of valuable papers or for segregating those which are unacceptable, e.g. banknotes that are alien to a currency
- G07D7/20—Testing patterns thereon
- G07D7/2008—Testing patterns thereon using pre-processing, e.g. de-blurring, averaging, normalisation or rotation
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07D—HANDLING OF COINS OR VALUABLE PAPERS, e.g. TESTING, SORTING BY DENOMINATIONS, COUNTING, DISPENSING, CHANGING OR DEPOSITING
- G07D7/00—Testing specially adapted to determine the identity or genuineness of valuable papers or for segregating those which are unacceptable, e.g. banknotes that are alien to a currency
- G07D7/20—Testing patterns thereon
- G07D7/2016—Testing patterns thereon using feature extraction, e.g. segmentation, edge detection or Hough-transformation
Abstract
The present invention relates to the technical fields of image, more particularly, to the text filed localization method of the bill alignment and eight neighborhood connected component offset correction of feature based point.The present invention proposes a kind of novel and efficient dual-stage character area localization method, i.e. the first stage is normalized into uniform sizes to bill images first, utilize the matched thinking of SIFT feature, the bill to be identified to every carries out feature extraction, and corresponding template reference map carries out Feature Points Matching, projection transformation is carried out after acquiring transformation matrix, completes bill alignment.Second stage searches for connected component using eight fields to one of information area using existing Template Information and queue is added with centainly sequence in the connected component for meeting agreement property in eight fields, and the offset by calculating the center and the Template Information class mark of queue head of the queue connected component, it is inferred to the predicted position of remaining information area, and by inferring whether the calculating of offset is accurate with the presence or absence of suitable connected component in the position to each region after progress offset correction.This method combines Feature Points Matching algorithm in computer vision and is based on location finding algorithm, realizes and is accurately positioned to Template Information region, and especially training off the serious bill of shifting to set has good effect.
Description
Technical field
The present invention relates to the technical fields of image, connect more particularly, to the bill alignment of feature based point and eight neighborhood
The text filed localization method of entire body offset correction.
Background technology
It is often taken directly according to letter good defined in every class template in the system of the bill automatic identification based on template
The position in breath area directly positions the position of bill to be identified, and the method positioning of this direct localization method in Template Information region is just
True rate is relatively low.This is because the characteristics of every bill has itself, because of the difference of shooting angle and the intact degree of bill, this leads
It has caused directly to apply mechanically Template Information and has come that localization of text success rate is not high, this leads to the failure of subsequent Text region.
China Patent Publication No. CN104916034A discloses a kind of based on the bank slip recognition system that can intervene template and knowledge
Other method, the text filed positioning interested in the system are the location informations for the text for directly using original template, not
Additional optimization operation is made to String localization.The method pair and template bill shooting size have differences or cover that beat effect bright
The positioning of aobvious bill is failed substantially.
China Patent Publication No. CN107622255A discloses a kind of bill field based on situation template and semantic template
Localization method and system, the system is same in the positioning of text interested simply to use situation template information, the method for
Some shooting angle have differences the bill for causing the size of the bill in image different from template bill size and are not suitable for.Simultaneously
The String localization that the system plays class invoice for set has used the method that location information is combined with bill attribute field into line position
Set amendment, but the method require be first defined per class invoice feature field and by detect the feature field of every invoice come
Opposite offset is calculated, the method has limitation.First this method needs manually for every class bill design feature field, time-consuming to take
Power, and do not ensure that and can design signature for every class bill;Second may be because in search characteristics field
Identification is wrong and feature field is caused to find failure and then String localization failure;Third, the premise using this method are needed to institute
There is line of text to be identified, efficiency is very low.
Invention content
The present invention is at least one defect overcome described in the above-mentioned prior art, provide feature based point bill alignment and
The text filed localization method of eight neighborhood connected component offset correction, the matching alignment of feature based point are connected to solid offsetting with eight neighborhood and repair
Positive character area localization method can utilize spy to most of bill in the system for carrying out the text location based on template
Sign point information completes bill alignment, and accurately calculates offset error by efficient connected component searching algorithm, and then successfully solves
Determine the problem that set is bought a ticket according to String localization hardly possible.A whole set of location algorithm can solve the orientation problem of most of bills, and this method has
Very strong universality.
The technical scheme is that:The text area of the bill alignment and eight neighborhood connected component offset correction of feature based point
Domain localization method, wherein including the alignment of feature based point matched bill and eight neighborhood connected component offset correction two large divisions,
The matched bill aligned portions of feature based point are the first stages of entire text filed location algorithm, first to bill
It is size normalised, and SIFT feature extraction is carried out to bill, then characteristic matching is carried out with template bill characteristic point, acquire transformation
Matrix, and projection transformation is carried out, complete the bill alignment of first stage;
Eight neighborhood connected component offset correction is the second stage of text filed location algorithm, on the basis of bill has been aligned,
One of information area is chosen as main operation region, according in original ticket templates character area information and connected component team
The offset error of row, and the position in other regions of template is modified using the offset error, and then acquire each information area
The exact position in domain.
Further, the similar bill alignment is to acquire transformation square by the extraction and matching primitives of characteristic point
Battle array, and perspective transform is carried out according to the transformation matrix, complete bill alignment.
Further, described some information area in ticket templates generates the eight neighborhood field of search, then to search
Area carries out binaryzation and Morphological scale-space, and the connected component in region of search is stored according to specified sequence enqueue, generates master
Connected component queue.
Further, during the connected component in the region of search is according to the sequence that specified sequence enqueue stores, it is right,
Left, upper and lower, upper right, bottom right, upper left, lower-left.
Further, the centre coordinate of the connected component of queue head is calculated, and calculates the coordinate and the region in template
The offset of centre coordinate, and with the position of information area in the offset correction other templates, and it is same to the information area after calculating
Sample eight neighborhood searches for connected component and is joined the team, if navigating to qualified connected component in the connected component queue of other positions,
It is correct then to show that the offset calculates, otherwise connected component goes out team, continues to carry out identical behaviour to next connected component in major queue
Make.
Bill is aligned to the first stage of a whole set of location algorithm, and bill align stage is as follows:
(1)It is size normalised to the progress of pending bill images first, i.e., the specified size of such bill is zoomed to it.
(2)Gray processing is carried out to pending bill images, then carries out SIFT feature extraction, and with corresponding mould
The characteristic point of plate image is matched, and then acquires transformation matrix H.
(3)Projection transformation is carried out to pending image according to transformation matrix H, the bill images after being converted complete
The alignment operation of pending image and benchmark image.
The text location for evidence of buying a ticket generally, for common, non-set, after the bill alignment for carrying out the first stage directly
Apply mechanically the information area location information in template can get accurately it is text filed.It buys a ticket to solution set inclined according to word
The problem of shifting, it is also necessary to continue the second stage of algorithm, is i.e. in the offset correction stage, be as follows:
(1)Some information area is randomly selected in template as main operation area, with the size in the region to its surrounding totally eight sides
To being extended, the eight neighborhood field of search is generated.
(2)Binaryzation and corresponding Morphological scale-space are carried out to the eight neighborhood field of search, to the connected component in nine grid
Storage of joining the team is carried out according to certain rule.
(3)The centre coordinate for calculating first connected component of queue is made with this coordinate and the region template information centre coordinate
Difference obtains offset K.
(4)The more specific location information that other information region is recalculated according to offset K obtains new information area position
Confidence ceases, and amplifies the band of position to each new location-appropriate, detects in the region in and arranges item with the presence or absence of meeting the region
The connected component of part, has, and shows that the offset is exactly true excursions amount, the offset correction stage completes, otherwise to next in queue
Connected component carries out identical offset correction operation.
Compared with prior art, advantageous effect is:Similarity is higher between the present invention takes full advantage of bill, can be by same
Characteristic point completes the alignment of bill between class ticket bill.It is opposite between our utilization set typewriting bodies on the basis of bill has been aligned
The constant feature in position carries out the accurate positionin of text using eight neighborhood connection solid offsetting correction.The applicable bill of this method is very
Extensively, and it is not necessarily to prior design feature mark auxiliary positioning, easy to operate and accuracy rate is high.
Description of the drawings
Fig. 1 indicates the algorithm frame of the present invention.
Fig. 2 indicates that eight neighborhood when algorithm second stage searches for schematic diagram.
Fig. 3 indicates the algorithmic procedure in offset correction stage.
Specific implementation mode
The attached figures are only used for illustrative purposes and cannot be understood as limitating the patent;It is attached in order to more preferably illustrate the present embodiment
Scheme certain components to have omission, zoom in or out, does not represent the size of actual product;To those skilled in the art,
The omitting of some known structures and their instructions in the attached drawings are understandable.Being given for example only property of position relationship described in attached drawing
Illustrate, should not be understood as the limitation to this patent.
As shown in Figure 1, our programs are divided into bill alignment and two big stage of offset correction.Bill align stage it is specific
Step includes:(1)Bill images to be identified are read, classified to it and obtain corresponding ticket templates information;(2)It is right
The processing of images to be recognized gray processing, and the uniform sizes of such bill are turned to its dimensional standard;(3)SIFT is carried out to the image
The characteristic point point set of the figure is matched with the characteristic point point set of template image, acquires transformation matrix H by feature point extraction;(4)
Perspective transform is carried out to the figure according to transformation matrix H, realization current bill is registrated with template bill, that is, realizes pair of bill
Together.
As shown in figure 3, the specific steps in bill offset correction stage include:(1)We, which randomly select in the class template, appoints
An information area anticipate as main operation area, and region of search identical with its size is generated to eight directions in the region, such as
Shown in Fig. 2, main operation area includes nine grid altogether;(2)We carry out binaryzation and shape to the field of search of this nine grid
State processing uses expansive working so that line of text adhesion uses etching operation so that little particle is small at one piece of formation connected component
Noise is eliminated;(3)Storage of joining the team is carried out according to certain sequence to the connected component in the field of search, sequence therein of joining the team is according to each
The center of connected component falls into which of nine grid to determine the priority of current connected component, and the priority of each grid is such as
Shown in Fig. 2,1 is highest priority, and 9 be lowest priority.It determines to join the team according to the height of the priority of connected component when joining the team
Sequentially, the high connected component of priority is first joined the team;(4)The centre coordinate for calculating first connected component of queue, with this coordinate and the area
Domain Template Information centre coordinate makees difference, obtains offset K;(5)The tool in other information region is recalculated according to offset K
Body position information, obtains new information area location information, and amplifies the band of position to each new location-appropriate, such as by mould
The height and width for the regional location that plate defines are multiplied by 1.2 times;(6)It detects to whether there is in the region and meets region agreement
The connected component of condition, has, and shows that the offset is exactly true excursions amount, the offset correction stage completes, otherwise to next in queue
A connected component carries out identical offset correction operation, until calculations of offset success, terminates the offset correction stage, or until queue
Suitable offset cannot be still found out for sky, then terminates algorithm, restoring to normal position failure information.
Obviously, the above embodiment of the present invention be only to clearly illustrate example of the present invention, and not be pair
The restriction of embodiments of the present invention.For those of ordinary skill in the art, may be used also on the basis of the above description
To make other variations or changes in different ways.There is no necessity and possibility to exhaust all the enbodiments.It is all this
All any modification, equivalent and improvement etc., should be included in the claims in the present invention made by within the spirit and principle of invention
Protection domain within.
Claims (5)
1. the text filed localization method of the bill alignment and eight neighborhood connected component offset correction of feature based point, feature exist
In, including the matched bill alignment of feature based point and eight neighborhood connected component offset correction two large divisions,
The matched bill aligned portions of feature based point are the first stages of entire text filed location algorithm, first to bill
It is size normalised, and SIFT feature extraction is carried out to bill, then characteristic matching is carried out with template bill characteristic point, acquire transformation
Matrix, and projection transformation is carried out, complete the bill alignment of first stage;
Eight neighborhood connected component offset correction is the second stage of text filed location algorithm, on the basis of bill has been aligned,
One of information area is chosen as main operation region, according in original ticket templates character area information and connected component team
The offset error of row, and the position in other regions of template is modified using the offset error, and then acquire each information area
The exact position in domain.
2. the bill alignment of feature based point according to claim 1 is text filed with eight neighborhood connected component offset correction
Localization method, it is characterised in that:The similar bill alignment is to acquire transformation by the extraction and matching primitives of characteristic point
Matrix, and perspective transform is carried out according to the transformation matrix, complete bill alignment.
3. the bill alignment of feature based point according to claim 1 is text filed with eight neighborhood connected component offset correction
Localization method, it is characterised in that:Described generates the eight neighborhood field of search to some information area in ticket templates, then to searching
Rope area carries out binaryzation and Morphological scale-space, and the connected component in region of search is stored according to specified sequence enqueue, generates
Main connected component queue.
4. the bill alignment of feature based point according to claim 1 is text filed with eight neighborhood connected component offset correction
Localization method, it is characterised in that:Connected component in the region of search be according to the sequence that specified sequence enqueue stores in,
Right, left, upper and lower, upper right, bottom right, upper left, lower-left.
5. the bill alignment of feature based point according to claim 1 is text filed with eight neighborhood connected component offset correction
Localization method, it is characterised in that:The centre coordinate of the connected component of queue head is calculated, and calculates the coordinate and the region in template
Centre coordinate offset, and with the position of information area in the offset correction other templates, and to the information area after calculating
Same eight neighborhood search connected component is simultaneously joined the team, if navigating to qualified connection in the connected component queue of other positions
It is correct then to show that the offset calculates for body, and otherwise connected component goes out team, continues to carry out next connected component in major queue identical
Operation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810434069.3A CN108806059B (en) | 2018-05-08 | 2018-05-08 | Text region positioning method based on note alignment and eight-neighborhood connector offset correction of feature points |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810434069.3A CN108806059B (en) | 2018-05-08 | 2018-05-08 | Text region positioning method based on note alignment and eight-neighborhood connector offset correction of feature points |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108806059A true CN108806059A (en) | 2018-11-13 |
CN108806059B CN108806059B (en) | 2020-05-22 |
Family
ID=64092020
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810434069.3A Active CN108806059B (en) | 2018-05-08 | 2018-05-08 | Text region positioning method based on note alignment and eight-neighborhood connector offset correction of feature points |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108806059B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109558844A (en) * | 2018-11-30 | 2019-04-02 | 厦门商集网络科技有限责任公司 | The method and apparatus of self-defined template discrimination is promoted based on image normalization |
CN109558846A (en) * | 2018-11-30 | 2019-04-02 | 厦门商集网络科技有限责任公司 | It is normalized based on OCR template and promotes self-defined template discrimination method and apparatus |
CN109800750A (en) * | 2019-01-25 | 2019-05-24 | 济南浪潮高新科技投资发展有限公司 | A kind of character area mask method of the text based on morphological image for rule composing |
CN109886257A (en) * | 2019-01-30 | 2019-06-14 | 四川长虹电器股份有限公司 | Using the method for deep learning correction invoice picture segmentation result in a kind of OCR system |
CN111275037A (en) * | 2020-01-09 | 2020-06-12 | 上海知达教育科技有限公司 | Bill identification method and device |
CN111325669A (en) * | 2020-03-05 | 2020-06-23 | 北京远心科技有限责任公司 | Correction scale for oblique photography and oblique photography image correction method |
CN111612967A (en) * | 2020-04-29 | 2020-09-01 | 武汉卓目科技有限公司 | Method and device for CIS image preprocessing of financial machine |
CN112132016A (en) * | 2020-09-22 | 2020-12-25 | 平安科技(深圳)有限公司 | Bill information extraction method and device and electronic equipment |
CN112434727A (en) * | 2020-05-02 | 2021-03-02 | 支付宝实验室(新加坡)有限公司 | Identity document authentication method and system |
CN117347367A (en) * | 2023-08-24 | 2024-01-05 | 广州鹰视信息科技有限公司 | Board card device positioning method, board card device detection method, device and medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101266654A (en) * | 2007-03-14 | 2008-09-17 | 中国科学院自动化研究所 | Image text location method and device based on connective component and support vector machine |
CN101377847A (en) * | 2007-08-29 | 2009-03-04 | 中国科学院自动化研究所 | Method for registration of document image and selection of characteristic points |
CN101872475A (en) * | 2009-04-22 | 2010-10-27 | 中国科学院自动化研究所 | Method for automatically registering scanned document images |
CN105469408A (en) * | 2015-11-30 | 2016-04-06 | 东南大学 | Building group segmentation method for SAR image |
CN107103323A (en) * | 2017-03-09 | 2017-08-29 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | A kind of target identification method based on image outline feature |
CN107832757A (en) * | 2017-11-03 | 2018-03-23 | 深圳航天信息有限公司 | A kind of recognition methods of invoice image |
-
2018
- 2018-05-08 CN CN201810434069.3A patent/CN108806059B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101266654A (en) * | 2007-03-14 | 2008-09-17 | 中国科学院自动化研究所 | Image text location method and device based on connective component and support vector machine |
CN101377847A (en) * | 2007-08-29 | 2009-03-04 | 中国科学院自动化研究所 | Method for registration of document image and selection of characteristic points |
CN101872475A (en) * | 2009-04-22 | 2010-10-27 | 中国科学院自动化研究所 | Method for automatically registering scanned document images |
CN105469408A (en) * | 2015-11-30 | 2016-04-06 | 东南大学 | Building group segmentation method for SAR image |
CN107103323A (en) * | 2017-03-09 | 2017-08-29 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | A kind of target identification method based on image outline feature |
CN107832757A (en) * | 2017-11-03 | 2018-03-23 | 深圳航天信息有限公司 | A kind of recognition methods of invoice image |
Non-Patent Citations (1)
Title |
---|
杜刚: "《银行票据识别系统的研究》", 《中国优秀硕士学位论文全文数据库(电子期刊)信息技术辑》 * |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109558844A (en) * | 2018-11-30 | 2019-04-02 | 厦门商集网络科技有限责任公司 | The method and apparatus of self-defined template discrimination is promoted based on image normalization |
CN109558846A (en) * | 2018-11-30 | 2019-04-02 | 厦门商集网络科技有限责任公司 | It is normalized based on OCR template and promotes self-defined template discrimination method and apparatus |
CN109800750A (en) * | 2019-01-25 | 2019-05-24 | 济南浪潮高新科技投资发展有限公司 | A kind of character area mask method of the text based on morphological image for rule composing |
CN109886257A (en) * | 2019-01-30 | 2019-06-14 | 四川长虹电器股份有限公司 | Using the method for deep learning correction invoice picture segmentation result in a kind of OCR system |
CN109886257B (en) * | 2019-01-30 | 2022-10-18 | 四川长虹电器股份有限公司 | Method for correcting invoice image segmentation result by adopting deep learning in OCR system |
CN111275037B (en) * | 2020-01-09 | 2021-06-08 | 上海知达教育科技有限公司 | Bill identification method and device |
CN111275037A (en) * | 2020-01-09 | 2020-06-12 | 上海知达教育科技有限公司 | Bill identification method and device |
CN111325669A (en) * | 2020-03-05 | 2020-06-23 | 北京远心科技有限责任公司 | Correction scale for oblique photography and oblique photography image correction method |
CN111612967A (en) * | 2020-04-29 | 2020-09-01 | 武汉卓目科技有限公司 | Method and device for CIS image preprocessing of financial machine |
CN111612967B (en) * | 2020-04-29 | 2023-03-14 | 武汉卓目科技有限公司 | Method and device for preprocessing CIS image of financial machine |
CN112434727A (en) * | 2020-05-02 | 2021-03-02 | 支付宝实验室(新加坡)有限公司 | Identity document authentication method and system |
CN112132016A (en) * | 2020-09-22 | 2020-12-25 | 平安科技(深圳)有限公司 | Bill information extraction method and device and electronic equipment |
CN112132016B (en) * | 2020-09-22 | 2023-09-15 | 平安科技(深圳)有限公司 | Bill information extraction method and device and electronic equipment |
CN117347367A (en) * | 2023-08-24 | 2024-01-05 | 广州鹰视信息科技有限公司 | Board card device positioning method, board card device detection method, device and medium |
Also Published As
Publication number | Publication date |
---|---|
CN108806059B (en) | 2020-05-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108806059A (en) | The text filed localization method of the bill alignment and eight neighborhood connected component offset correction of feature based point | |
Liu et al. | Autoshape: Real-time shape-aware monocular 3d object detection | |
US8798357B2 (en) | Image-based localization | |
WO2019042426A1 (en) | Augmented reality scene processing method and apparatus, and computer storage medium | |
CN111435438A (en) | Graphical fiducial mark recognition for augmented reality, virtual reality and robotics | |
US7519197B2 (en) | Object identification between non-overlapping cameras without direct feature matching | |
JP5385105B2 (en) | Image search method and system | |
US20140219569A1 (en) | Image recognition system and method for identifying similarities in different images | |
US20070183665A1 (en) | Face feature point detecting device and method | |
Hsu et al. | Regressive tree structured model for facial landmark localization | |
JP2011008507A (en) | Image retrieval method and system | |
CN113240716B (en) | Twin network target tracking method and system with multi-feature fusion | |
CN112132013B (en) | Vehicle key point detection method | |
CN116630608A (en) | Multi-mode target detection method for complex scene | |
Ma et al. | Location-aware box reasoning for anchor-based single-shot object detection | |
Liu et al. | MFPN-6D: Real-time One-stage Pose Estimation of Objects on RGB Images | |
CN113076891B (en) | Human body posture prediction method and system based on improved high-resolution network | |
WO2022016838A1 (en) | Positioning method and device based on visual tag | |
CN113505808A (en) | Detection and identification algorithm for power distribution facility switch based on deep learning | |
Yu et al. | Learning bipartite graph matching for robust visual localization | |
Yu et al. | Improving feature-based visual localization by geometry-aided matching | |
Liu et al. | Modelling and simulation: an improved RANSAC algorithm based on the relative angle information of samples | |
Li et al. | RaP-Net: A region-wise and point-wise weighting network to extract robust features for indoor localization | |
Lu et al. | Road vanishing-point detection: A multi-population genetic algorithm based approach | |
Fu et al. | Performance analysis of single-query 6-DoF camera pose estimation in self-driving setups |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |