CN104217203A - Complex background card face information identification method and system - Google Patents
Complex background card face information identification method and system Download PDFInfo
- Publication number
- CN104217203A CN104217203A CN201310216817.8A CN201310216817A CN104217203A CN 104217203 A CN104217203 A CN 104217203A CN 201310216817 A CN201310216817 A CN 201310216817A CN 104217203 A CN104217203 A CN 104217203A
- Authority
- CN
- China
- Prior art keywords
- character information
- image
- block
- pixels
- information image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Abstract
The application provides a complex background card face information identification method and system. The method comprises the following steps: positioning a region where character information in an image of a card face to be identified in order to acquire one or more character information images; extracting complex characteristic vectors from the acquired character information images; and identifying character information in the character information images according to the complex characteristic vectors. Through the application, a problem concerned with accurate information identification under a complex card face background (such as a bank card) is solved, the defect that complex background card face information cannot be identified effectively and accurately in a conventional method based on threshold segmentation is overcome on the basis of complex image characteristics, the card face information identification accuracy is increased, and efficient identification of complex background card face information is realized.
Description
Technical field
The application relates to image processing field, particularly relates to a kind of complex background card surface information recognition methods and system.
Background technology
Along with the development of internet, on-line payment, mobile-phone payment etc. fast pay more and more general.During this period, the situation of the direct input card information of bank card user is needed to get more and more.Use image-capturing apparatus to obtain bank card card surface information, supplementary Rapid input has also become trend trend.
At present, optical character recognition can be used for decoding the image of the character using the line of text including but not limited to the such as level that scanner and digital camera obtain at interior various types of image-capturing apparatus and so on, for reading the information that bank card card face is printed, as card number, the term of validity, holder name and bank card hair fastener tissue etc.Optical character recognition method, its basic calculating flow process is display foreground background segment, binaryzation, refinement, encoding and identification.Owing to having dull background hypothesis, in traditional Optical Character Recognition system, the ratio of division of word prospect and background is easier; Can carry out binaryzation in the word prospect extracted, refinement, coding then identify, calculate easier on the whole, calculated amount is few.But these class methods are mainly applicable to the text identification of dull background, None-identified goes out the character in complex background.Thus, recognition technology based on traditional optical character recognition principle has significant limitation, its limitation is mainly lower based on the recognition result degree of reliability, secondly traditional optical character identifying method is mainly for dull background text document design, effectively cannot be applied to the card surface information identification of the bank card usually with complex background pattern, user needs to spend more energy inspection, correcting identification result, with ensure the stored correctness of guarantor.
Therefore, a kind of technology that accurately can identify card release surface information from the image of complex background card face of demand.
Summary of the invention
The fundamental purpose of the application is to provide a kind of complex background card surface information recognition methods and system thereof, realizes to solve the problem that the information of (as bank card) accurately identifies under the background of complicated card face.
According to the embodiment of an aspect of the application, the recognition methods of a kind of complex background card surface information is provided, comprises: locate the character information region in the image in described card face to be identified, to obtain one or more character information image; Carry out complex characteristic vector to the described character information image obtained to extract; According to described complex characteristic vector, identify the character information in described character information image.
According to the embodiment of the another aspect of the application, a kind of complex background card surface information recognition system is provided, comprises: positioning unit, locate the character information region in the image in described card face to be identified, to obtain one or more character information image; Fisrt feature extraction unit, carries out complex characteristic vector to the described character information image obtained and extracts; Character information recognition unit, according to described complex characteristic vector, identifies the character information in described character information image.
Compared with prior art, the application provide a kind of based on complicated image feature as Harris, SIFT feature, the card surface information recognition methods being applicable to complex background and system thereof, overcome the defect cannot carrying out effectively accurately identification to complex background card surface information that traditional method based on Threshold segmentation faces, improve the recognition accuracy of card surface information, realize the identification of high efficiency complex background card surface information.
Accompanying drawing explanation
Accompanying drawing described herein is used to provide further understanding of the present application, and form a application's part, the schematic description and description of the application, for explaining the application, does not form the improper restriction to the application.In the accompanying drawings:
Fig. 1 is the main flow figure of the complex background card surface information recognition methods of the application's embodiment;
Fig. 2 is the process flow diagram of preferred preprocess method in the recognition methods of the application's embodiment;
Fig. 3 is feature extraction and calculation mode schematic diagram in the preprocess method of the application's embodiment;
Fig. 4 is the process flow diagram of preferred character locating method in the recognition methods of the application's embodiment;
Fig. 5 is images match glide direction precedence diagram in the character locating method of the application's embodiment;
Fig. 6 preferably judges recognition methods process flow diagram in the recognition methods of the application's embodiment;
Fig. 7 is the structured flowchart of the complex background card surface information recognition system of the application's embodiment.
Embodiment
The main thought of the application is based on complicated image feature, be applicable to the card surface information identification of complex background, its application complicated image feature is as Harris and SIFT feature, by gather a large amount of through flag data sample training sorter as support vector machine, artificial neural network etc. obtain training after there is the sorter (hereinafter referred to as recognizer) that can carry out the series of parameters data identified, this recognizer is used to fit through character picture region that above-mentioned recognizer is partitioned into obtain corresponding character information, thus solve the problem of the None-identified complex background information that tradition faces based on the method for Threshold segmentation.
For making the object of the application, technical scheme and advantage clearly, below in conjunction with drawings and the specific embodiments, the application is described in further detail.
The application does not do picture prospect, background segment, directly uses character candidates region to be identified entirety to calculate its proper vector as word prospect, and uses machine learning method training classifier to obtain recognizer; Meanwhile, for obtaining character candidates region accurately, need in the calculating of the application repeatedly to identify, and the characteristics of image used is comparatively large, calculates relative complex thus higher to the requirement of operation platform.
Fig. 1 shows main flow Figure 100 of the complex background card surface information recognition methods of the application's embodiment.
Step S110, inputs the card face image with complex background card to be identified.
Usually image capture device such as camera, camera, scanner etc. are utilized to obtain the card face image of object (as bank card) to be identified.These have on the card face image of complex background, and have the various information needing typing, mainly character information, comprising: numeral, letter, word, other various characters.
Step S120, on this card face image, the image-region of the character information (the various information as card: card number, numeral, date, name etc.) of location care information/to be identified.
Can through coarse localization to the process of accurately locating (or crying accurately location).
Such as, the region at character (comprising numeral, card number, word etc.) the information place on the card face image of bank card to be identified is located.
Here can according to card, as bank card production standard, to bank card number region, date of expiration region, holder name region etc. these comprise the region of character (numeral, card number), carry out Primary Location, as the coarse localization mentioned by preferred characters information locating method 400.
And then accurately locate, such as carry out accurately locating further fast at Primary Location result periphery based on card face texture information, thus obtain the accurate location in character information region, card face.Preferably, the accurate locator meams that character information localization method 400 can be adopted to mention realizes, and it even can obtain character information image corresponding to single character information area in accurate position fixing process.
Step S130, based on the location to the character information region on the image of card face, identifies the character information in character information image to be identified.
Card surface information mainly various character information comprises word (as holder name, date of expiration), numeral (card number, date number as bank card) etc.Main as follows:
The first, obtain single character information image (image of each independent character information).
A kind of mode is after accurate location character information area, for kinds of characters information area, can carry out the segmentation of character information.Partitioning scheme is such as: card number information region segmentation can be become 21 sub regions as card number alternative area, region date of expiration can be divided into month and time (month and year information back slash "/" are split), holder name can be separated its each English alphabet.Can determine single character information region like this, and obtain single character information image.
Preferred another kind of mode, just character information image can be obtained when locating, such as: by following character information positioning step 400, orient the accurate location of each character information, and in this accurate position fixing process, just obtain single character information image, and a series of single character information image is provided in order to identifying use.Thus, each independent character information image obtained can be received.
Second, to a series of character information images (image of a series of single character information) obtained, carry out feature extraction, preferably, the feature extraction mode in following preprocess method 200 can be adopted, as the 8*8 pixel node-by-node algorithm to 24*24 pixel criterion image, obtain as Harris isocenter feature, SIFT have the feature (namely complex characteristic is vectorial) of rotational invariance.
3rd, then according to these features (complex characteristic vector), utilize the recognizer that training in advance sorter obtains, carry out judging to identify the character information in correspondence image region.Can be trained character database in advance by machine learning algorithm (sorter such as neural network, support vector machines), obtain recognizer (sorter namely with supplemental characteristic or title character classifier), by extracting character notable feature as complex characteristic vectors such as above-mentioned Harris, SIFT, carrying out Classification and Identification with recognizer and judge thus identify character information.Preferably, as utilized the judgement recognition methods 600 of following preferred preprocess method 200 and character information.
Step S140, verifies the result identified, and adopts algorithm to verify and judges through recognizer the character information that identifies.Preferably, for bank card card, the Luhn mould ten checking algorithm verification that bank card number can be used to encode generally adopt such as judges that the card number identified in recognition methods 600 is to confirm final recognition result.Verification succeeds then exports net result.Error message can also be provided when verifying unsuccessfully.
As shown in Figure 2, be preferred pretreatment method flow Figure 200 adoptable in the recognition methods of complex background card surface information.
Acquisition step S210, gathers a large amount of view data samples in advance and splits these view data samples and mark.The independent character information image (sample image) that a large amount of classification is good can be obtained thus.
Before information on the image to be identified to input identifies, need to carry out pre-service, to obtain sample (character information sample) database and recognizer (being with the sorter of supplemental characteristic after training) that can identify information to be identified.
Concrete a kind of mode, can be gather a large amount of card face (as: credit card) view data samples in advance, the normally image of tape character information, then splits these view data samples, the single character frame (calling sample image in the following text) obtained is marked after segmentation.
By this segmentation, by the character information in view data sample separately, independent character information image (sample image) can be formed.And by this mark, can by each independent character information image in the sample image that gathers, determine in the character information classification belonging to it.
This segmentation, mark (i.e. view data sample analysis) by modes such as artificial or machine algorithms, can specify the classification of character information belonging to the sample image (independent character information image) after the segmentation gathered.
Here, acquisition mode, can provide bank card image to carry out view data sample collection by picture catching/capturing apparatus.
Such as: by the image of bank's card of image acquisition device sample, the card face image of these cards comprises the such as character information such as card number, date (view data sample).Next can analyze these view data samples (segmentation and mark), such as manual type analysis, concrete can be use image editing tools, as Photoshop etc., numerical character region can be found on these view data samples collected, choose the scope of a numerical character, as first 6, then the scope of this numeral of covering 6 is taken out, save as an image separately, namely divided image data sample is independent character information image (sample image), and be 6 by its flag settings, namely mark this as sample image independent character information image belonging to character information classification.
Normalization step S220, the sample image after being marked by each acquisition, is normalized.
Concrete as, a sample image is carried out the process amplifying, correct or be contracted to normal size.Preferably, whole sample image is all unified for 24*24 pixel size.
Normalized, can comprise the image processing operations such as rectification and convergent-divergent of sample image.Can by image size normalization by convergent-divergent.Here, by correcting, the situation that further image gathering (as shooting) can also be existed such as sense of rotation is rotated the situation (as the character information direction that people get used to) of getting back to reference direction.
Here, by the sample image after these normalization, can also be kept in suitable memory device, database and/or buffer memory further.
Characteristic extraction step S230, the sample image after each normalization carries out feature extraction process, carries out proper vector description to it.Obtain thus and carry out the complicated vector characteristics judging the sample image needed for identification character information, as feature samples.
Specifically, on a standard picture (i.e. unified image size criteria), here, can be the sample image carrying out normalized, as the image of 24*24 pixel size, to its characteristics of image of image node-by-node algorithm of each 8*8 pixel on this image.Extract characteristics of image as: (Harris Corner Detection Algorithm is namely based on the characteristics of image used in the Robust Algorithm of Image Corner Extraction of gray level image for Harris feature, proper vector as relevant with angle point), SIFT(Scale-invariant feature transform scale invariant feature changes) feature (be that SIFT extracts the characteristics of image used in local feature algorithm in image, as to the irrelevant proper vector of scaling, rotation, brightness change), etc.
One is extracted the mode of above-mentioned complex characteristic vector to the image pointwise of 8*8 pixel each on 24*24 sample image, such as: can be that a 8*8 block of pixels 8*8 block of pixels be adjacent forms the large large block of image block as 16*16 and carries out; This large block can be chosen such as 36 the strongest dimensional feature vectors and describe; Simultaneously, this block can as on the image of this 24*24 on standard picture, according to the step-length of 8 pixels, respectively in transverse direction, vertically move, the final sample image for this 24*24 standard after combination, can obtain the character pattern of the whole 24*24 of high dimensional feature vector description.
Node-by-node algorithm, such as: laterally, longitudinally move by 8 pixel step length respectively, each block of pixels (block) has 16 pixels, so transverse direction, longitudinal proper vector having this 36 dimension of 2 position calculation respectively.The total length of proper vector is 36*2*2=144.As shown in Figure 3.
Further, this high dimensional feature vector (as 144 dimensions in above-mentioned example) can also carry out dimensionality reduction, can promote below by the recognition speed in the identification processing procedure of description after reducing dimension.Dimension reduction method such as can adopt pivot analysis PCA etc. to reduce the algorithm of dimension.
Here, extract the characteristics of image in each sample image after normalization, preferably, extract more complicated characteristics of image, and as far as possible for the complex characteristic that direction is constant is vectorial, the feature samples of each sample image used can will be needed in the identification processing procedure of description below as the application, especially in identifying processing, optical identification is carried out by the mode of target detection identification, need the feature samples using this kind of complexity (as Harris feature, the complex characteristic vector that the directions such as SIFT feature are constant), to overcome the impact that complicated card face background causes in identification processing procedure.
Here, extract these features of image, will use as in the recognizer of feature samples when judging identification, these feature samples can be saved in various memory device, buffer memory, database further.
Based on aforesaid way, the application is without the need to carrying out the segmentation of prospect, background to card image, the entirety of the candidate region of the card face image of direct use character to be identified is as word prospect, calculate its proper vector, namely use more more complicated characteristics of image, it is in recognition accuracy, the recognition result degree of reliability, all by higher than traditional traditional binary segmentation method segmentation prospect, background and identify the mode of prospect.
Training step S240, utilize the feature of the sample image extracted in characteristic extraction step S205, as the feature samples in the identifying processing of image information, and with these feature samples training classifiers, to obtain sorter and the recognizer of corresponding identifying processing (as recognizer) containing parameter data used.
Sorter, as adopted support vector machines or artificial neural network.
And sorter training, such as, the sorters such as Training Support Vector Machines SVM or artificial neural network.
Here, based on the complex characteristic vector extracted, such as, the Harris feature of the multiple sample images extracted in step S205, can utilize Harris feature to carry out sorter training, can obtain sorter and recognizer that one comprises supplemental characteristic.This recognizer, based on the feature samples had, will be used for the judgement identifying processing to character information image to be identified.
As shown in Figure 4, be preferred character locating method flow diagram 400 adoptable in the recognition methods of complex background card surface information.
Coarse localization step S410, carries out coarse localization to the character information region on the card face image in the card face (as bank card) of the needs identification of input.
Here, the card face image of the needs identification of input, can provide the image in the card face (such as bank card) with complex background by picture catching/capturing apparatus.Further, needing the card face identified, is the card face with complex background, as various bank card (credit card) etc.Traditional Optical Character Recognition system, needs word (character) prospect and background segment, also just can only carry out character information identification to the simple card face of background.The card faces such as the bank card of card face background complexity, it cannot split prospect and background, thus, is to adopt traditional optical character recognition system to identify the information (word, character etc.) on it.
Specifically, first to the image in the card face of background complexity, carry out the location of the rough character information band of position.Can use experience information, international standard ISO7810, the ISO7816 etc. of such as bank's card, come the character position region on card face image that rough location needs to carry out identifying, such as: horizontal, longitudinal positional locations (or horizontal, the longitudinal positional locations of ultimate character/card number) of locating the first character (card number).On the basis of this band of position, then carry out following accurate location.
Accurate positioning step S420(or claim accurate positioning step), based on the character information region of above-mentioned coarse localization, carry out accurate character information zone location.
Character information on these card faces is a series of character to be identified (as numeral, card number, word, letter, other symbols etc.) information in the image in card face to be identified.
Accurate location can obtain the position, kinds of characters region on the image of card face to be identified, and then navigates to each character information position, can also obtain each character information region and obtain the image of corresponding a series of single character information.The character information image of these the single character informations detected can be utilized in identifying processing.
Particularly, near the initial character position in the character information region of above-mentioned coarse localization, slip recognizer detects, and peak response position when detecting utilizing this recognizer is as the position at final the first character (such as: card number, numeral) place.
Wherein, this sorter preferably, can be the sorter such as support vector machine and artificial neural network.
Wherein, detect preferably, can from the centre position of the character information region of coarse localization (the first character approximate region as coarse localization), then according to the order shown by the numeral in images match shown in Fig. 5 (recognizer detection) glide direction precedence diagram, utilize recognizer to detect to each position up and down (first to detect in the position of 1 instruction, the position sliding into 2 instructions is again detected, then the position of 3 instructions is slided into, until the position of 9 instructions), when recognizer detects certain position, output stops when being greater than certain threshold value detecting search, and export the position of initial character location.Such can minimizing detects number of times, raising locating speed.
Equally, by accurately locating the above-mentioned concrete mode of initial character, the position of last (end) character accurately can also be located.
Further, by between initial character and end character according to the uniform distribution of 24*24 pixel size, multiple intermediate character information region can be obtained.Like this, transversely slide this recognizer, just progressively can obtain the position of each character information between initial character and end character.The character information of these diverse locations also just can be located exactly.
Accurately located these character information positions, and then by the detection (coupling of image) of recognizer, also corresponding each character information image of each character information region can just be obtained, thus, be obtained by a series of character information images of initial character to end character.
Similar, on complex background card face as other character information regions on bank card card face, such as date of expiration, name, card type mark etc., aforesaid way also can be adopted accurately to locate, and can remove based on this character information image obtaining these diverse location places.
As shown in Figure 6, be the judgement recognition methods process flow diagram 600 of adoptable preferred character information in the recognition methods of complex background card surface information.
Receiving step S610, receives and obtains character information image according to location character information area.
The image in the card face to be identified of input, after localization process, can obtain a series of character information region, even get corresponding a series of character information image.Receive the character information image that these are to be identified.
Preferably, the mode of the accurate location character information region adopting the recognizer in character information localization method 400 to detect, obtains the character information image of a series of single character information.What receive is these character information images.
Characteristic extraction step S620, based on the character information image obtained, carries out the extraction of characteristics of image.
This image characteristics extraction step, can see the feature extraction mode in preprocess method 200, analyze each (image as single character information) in a series of character information images obtained in character information localization method 400, extract the constant complex characteristic vector in direction (Harris feature, SIFT feature etc.).
Judge identification step S630, based on the complex characteristic vector extracted, these characteristics of image (complex characteristic vector) are sent in recognizer and identifies.
Preferably, as used the recognizer that obtains after complex characteristic vector sample training in preprocess method 200, carrying out judgements and identifying, determining the character information of correspondence in single character information image to be identified.Judge as carried out identification to card number, then judge whether certain image containing a numerical character (having an image for the character information of picture numeral " 1 ") belongs to this numeral (whether representative digit " 1 ").
Here, due to the complex characteristic vector adopting direction constant, optical character information identification (i.e. recognizer) is carried out by object detection method, thus not by the impact of complicated card face background.
Here, the different sorter of utilization, it trains the recognizer classification obtained to judge that number of times is different.Use two classification device, such as support vector machines, when judging the carrying out classification of digital character information, numerical character frame to be identified needs to judge to determine which numeral (10 subseries could judge to belong to which numeral) by this recognizer of 10 numerical characters.And use multi classifier, as artificial neural network, then to digital character information carry out classification judge time, this numerical character frame to be identified does a subseries and just can judge, is namely undertaken judging by 1 recognizer.
According to the embodiment of the application, also provide the recognition system 700 of the information in a kind of complex background card face corresponding to described recognition methods.The recognition system structured flowchart of the information in a kind of complex background card face as shown in Figure 7.
Preferably, this system 700 comprises: input block 710, inputs the card face image with complex background card to be identified; Positioning unit 720, on this card face image that input block 710 inputs, the character information of location care information/to be identified; Character information judges recognition unit 730, locates the character information affiliated area on this card face image, identify the character information in character information image to be identified based on positioning unit 720; Authentication unit 740, to judging that the result that recognition unit 730 identifies is verified; Pretreatment unit 750, with the recognizer (sorters with series of parameters data) obtaining feature samples in advance and obtain after utilizing feature samples training classifier, judge the character information image characteristics extraction to be identified of recognition unit 730 and the judgement identification of character information to be applied to.
Input block 710, preferably, utilizes image capture device such as camera, camera, scanner etc. to obtain the card face image of object (as bank card) to be identified usually.These have on the card face image of complex background, and have the various information needing typing, mainly character information, comprising: numeral, letter, word, other various characters.
Positioning unit 720, realizes the process as step S120.Preferably, coarse localization unit 721 and accurate positioning unit 722 can be comprised, to realize the process of character information localization method 400.Preferably, coarse localization unit 721 and accurate positioning unit 722 realize the process of step S410 and step S420 respectively.
Character information judges recognition unit 730, realizes the process of step S130.Preferably, reception image information units 731, fisrt feature extraction unit 732 can be comprised, judge recognition unit 733, realize the process judging recognition methods 600.Receive image information units 731, obtain a series of word information symbol images that accurately location obtains, the process as Application way 400 obtains character information image (i.e. the process of the step S610 of method 600); Fisrt feature extraction unit 732, extracts characteristics of image (process of step S620, the process of the image characteristics extraction step S230 in reference method 200) to extract complex characteristic vector; Judge recognition unit 733, step S240 in its Application way 200 trains the feature samples of recognizer and the step S230 extraction obtained, the characteristics of image that feature based extraction unit 732 extracts carries out classification and judges, to identify character information belonging to character picture information (process of step S630).
Authentication unit 740, preferably, realizes the process of verification step S140.As, for bank card card, the Luhn mould ten checking algorithm verification that bank card number can be used to encode generally adopt such as judges that the card number identified in recognition unit 733 is to confirm final recognition result.Verification succeeds then exports net result.Error message can also be provided when verifying unsuccessfully.
Pretreatment unit 750, preferably, realizes the process of preprocess method 200.Preferably, can comprise: collecting unit 751 is to realize step S210 process, normalization unit 752 to realize the process of step S220, second feature extraction unit 753 to realize the process of step S230 and training unit 754 to realize the process of step S240.
The function realized due to the system of the present embodiment is substantially corresponding to the embodiment of the method shown in earlier figures 1 to Fig. 6, therefore not detailed part in the description of the present embodiment, see the related description in previous embodiment, can not repeat at this.
The application have employed the constant complex characteristic vector in direction, and carry out optical character identification by the method for target detection, overcome the impact that complicated card face background causes, and, the recognition result degree of reliability gets a promotion, apply in the card surface information identification of complex background, its correctness accuracy is high.
In one typically configuration, computing equipment comprises one or more processor (CPU), input/output interface, network interface and internal memory.Internal memory may comprise the volatile memory in computer-readable medium, and the forms such as random access memory (RAM) and/or Nonvolatile memory, as ROM (read-only memory) (ROM) or flash memory (flash RAM).Internal memory is the example of computer-readable medium.
Computer-readable medium comprises permanent and impermanency, removable and non-removable media can be stored to realize information by any method or technology.Information can be computer-readable instruction, data structure, the module of program or other data.The example of the storage medium of computing machine comprises, but be not limited to phase transition internal memory (PRAM), static RAM (SRAM), dynamic RAM (DRAM), the random access memory (RAM) of other types, ROM (read-only memory) (ROM), Electrically Erasable Read Only Memory (EEPROM), fast flash memory bank or other memory techniques, read-only optical disc ROM (read-only memory) (CD-ROM), digital versatile disc (DVD) or other optical memory, magnetic magnetic tape cassette, tape magnetic rigid disk stores or other magnetic storage apparatus or any other non-transmitting medium, can be used for storing the information can accessed by computing equipment.According to defining herein, computer-readable medium does not comprise temporary computer readable media (transitory media), as data-signal and the carrier wave of modulation.
Each embodiment in this instructions generally adopts the mode of going forward one by one to describe, and what each embodiment stressed is the difference with other embodiments, between each embodiment identical similar part mutually see.
The application can describe in the general context of computer executable instructions, such as program module or unit.Usually, program module or unit can comprise the routine, program, object, assembly, data structure etc. that perform particular task or realize particular abstract data type.In general, program module or unit can be realized by software, hardware or both combinations.Also can put into practice the application in a distributed computing environment, in these distributed computing environment, be executed the task by the remote processing devices be connected by communication network.In a distributed computing environment, program module or unit can be arranged in the local and remote computer-readable storage medium comprising memory device.
Finally, also it should be noted that, term " comprises ", " comprising " or its any other variant are intended to contain comprising of nonexcludability, thus make to comprise the process of a series of key element, method, commodity or equipment and not only comprise those key elements, but also comprise other key elements clearly do not listed, or also comprise by the intrinsic key element of this process, method, commodity or equipment.When not more restrictions, the key element limited by statement " comprising ... ", and be not precluded within process, method, commodity or the equipment comprising described key element and also there is other identical element.
Those skilled in the art should understand, the embodiment of the application can be provided as method, system or computer program.Therefore, the application can adopt the form of complete hardware embodiment, completely software implementation or the embodiment in conjunction with software and hardware aspect.And the application can adopt in one or more form wherein including the upper computer program implemented of computer-usable storage medium (including but not limited to magnetic disk memory, CD-ROM, optical memory etc.) of computer usable program code.
Apply specific case herein to set forth the principle of the application and embodiment, the explanation of above embodiment is just for helping method and the main thought thereof of understanding the application; Meanwhile, for one of ordinary skill in the art, according to the thought of the application, all will change in specific embodiments and applications, in sum, this description should not be construed as the restriction to the application.
Claims (20)
1. a complex background card surface information recognition methods, is characterized in that, comprising:
Locate the character information region in the image in described card face to be identified, to obtain one or more character information image;
Carry out complex characteristic vector to the described character information image obtained to extract;
According to described complex characteristic vector, identify the character information in described character information image.
2. method according to claim 1, is characterized in that, locates the character information region in card face image to be identified, to obtain one or more character information image, comprising:
Coarse localization step, based on the production standard in complex background card face, Primary Location character information region;
Accurate positioning step, according to the character information region of Primary Location, utilizes a recognizer sequentially to detect in certain direction, obtains each character information position accurately, and obtain the character information image of relevant position.
3. method according to claim 2, is characterized in that,
Coarse localization step also comprises:
During Primary Location character information region, the first character information of energy coarse localization and/or ultimate character information position;
Accurate positioning step also comprises:
Near described the first character information and/or ultimate character information position, utilize described recognizer by direction order shiding matching up and down to detect, obtain the position of the first character information and/or ultimate character information accurately;
By unified image size criteria, uniform distribution is by the determined character information region of accurate location of the first character information and/or ultimate character information, described recognizer is utilized by direction order shiding matching up and down to detect to each character information region distributed, obtains the position of each character information accurately;
Based on the position accurately of each character information, obtain each character information image in region, relevant position.
4. method according to claim 1, is characterized in that, carries out the extraction of complex characteristic vector comprise the described character information image obtained:
Based on locating the one or more character information images obtained, size criteria is unified to described character information image;
Preset a block of pixels, to the image node-by-node algorithm characteristics of image of each described block of pixels on described character information image, to obtain the complex characteristic vector of described character information image.
5. method according to claim 4, is characterized in that, described node-by-node algorithm characteristics of image comprises:
The block of pixels that described block of pixels is adjacent forms block, and the proper vector described block being chosen to 36 dimensions describes;
On described character information image by the step-length of 8 pixels respectively transversely, vertically move, combination obtain describing the proper vector of the higher-dimension of described character information image.
6. method according to claim 1, is characterized in that, also comprises: pre-treatment step, wherein,
Collection has the card face view data sample of complex background in a large number and carries out splitting and marking, to obtain the character information image that marked classification;
To the described character information image that marked classification, be normalized, to obtain the character information image of unified size criteria;
Preset a block of pixels, to the image node-by-node algorithm characteristics of image of each described block of pixels on described character information image, to obtain the complex characteristic vector of described character information image as feature samples;
Utilize described feature samples, carry out sorter training, to obtain the described recognizer of identifying processing.
7. method according to claim 6, is characterized in that, according to described complex characteristic vector, identifies that the character information in described character information image comprises:
Feature based sample, sends described character information image into described recognizer, judges and identifies corresponding character information.
8. method according to claim 6, it is characterized in that, preset a block of pixels, to the image node-by-node algorithm characteristics of image of each described block of pixels on described character information image, comprise as feature samples using the complex characteristic vector obtaining described character information image:
The block of pixels that described block of pixels is adjacent forms block, and the proper vector described block being chosen to 36 dimensions describes;
On described character information image by the step-length of 8 pixels respectively transversely, vertically move, combination obtain describing the proper vector of the higher-dimension of described character information image.
9. the method according to claim 4 or 6, is characterized in that:
The unified size criteria of character information image unifies for 24*24 pixel criterion image by described character picture; The block of pixels preset is 8*8 block of pixels; Described complex characteristic vector comprises the constant proper vector in direction.
10. method according to claim 1, is characterized in that, also comprises:
Input step, utilizes picture catching mode to input the image in described card face to be identified;
Verification step, verifies to confirm final recognition result to the described character information identified.
11. 1 kinds of complex background card surface information recognition systems, is characterized in that, comprising:
Positioning unit, locates the character information region in the image in described card face to be identified, to obtain one or more character information image;
Fisrt feature extraction unit, carries out complex characteristic vector to the described character information image obtained and extracts;
Character information recognition unit, according to described complex characteristic vector, identifies the character information in described character information image.
12. systems according to claim 11, it is characterized in that, positioning unit comprises:
Coarse localization unit, based on the production standard in complex background card face, Primary Location character information region;
Accurate positioning unit, according to the character information region of Primary Location, utilizes a recognizer sequentially to detect in certain direction, obtains each character information position accurately, and obtain the character information image of relevant position.
13. systems according to claim 12, is characterized in that,
Coarse localization unit also comprises:
During Primary Location character region, the first character information of energy coarse localization and/or ultimate character information position;
Accurate positioning unit also comprises:
Near described the first character information and/or ultimate character information position, utilize described recognizer by direction order shiding matching up and down to detect, obtain the position of the first character information and/or ultimate character information accurately;
By unified image size criteria, uniform distribution is by the determined character information region of accurate location of the first character information and/or ultimate character information, described recognizer is utilized by direction order shiding matching up and down to detect to each character information region distributed, obtains the position of each character information accurately;
Based on the position accurately of each character information, obtain each character information image in region, relevant position.
14. systems according to claim 11, is characterized in that, fisrt feature extraction unit comprises:
Based on locating the one or more character information images obtained, size criteria is unified to described character information image;
Preset a block of pixels, to the image node-by-node algorithm characteristics of image of each described block of pixels on described character information image, to obtain the complex characteristic vector of described character information image.
15. systems according to claim 14, is characterized in that, in fisrt feature extraction unit, described node-by-node algorithm characteristics of image comprises:
The block of pixels that described block of pixels is adjacent forms block, and the proper vector described block being chosen to 36 dimensions describes;
On described character information image by the step-length of 8 pixels respectively transversely, vertically move, combination obtain describing the proper vector of the higher-dimension of described character information image.
16. systems according to claim 11, is characterized in that, also comprise: pretreatment unit, and described pretreatment unit comprises:
Collecting unit, collection has the card face view data sample of complex background in a large number and carries out splitting and marking, to obtain the character information image that marked classification;
Normalization unit, to the described character information image that marked classification, is normalized, to obtain the character information image of unified size criteria;
Second feature extraction unit, presets a block of pixels, to the image node-by-node algorithm characteristics of image of each described block of pixels on described character information image, to obtain the complex characteristic vector of described character information image as feature samples;
Training unit, utilizes described feature samples, carries out sorter training, to obtain the described recognizer of identifying processing.
17. systems according to claim 16, is characterized in that, character information recognition unit comprises: judge recognition unit, and described character information image is sent into described recognizer by its feature based sample, judges and identifies corresponding character information.
18. systems according to claim 16, is characterized in that, second feature extraction unit comprises:
The block of pixels that described block of pixels is adjacent forms block, and the proper vector described block being chosen to 36 dimensions describes;
On described character information image by the step-length of 8 pixels respectively transversely, vertically move, combination obtain describing the proper vector of the higher-dimension of described character information image.
19. systems according to claim 14 or 16, is characterized in that:
The unified size criteria of character information image unifies for 24*24 pixel criterion image by described character picture; The block of pixels preset is 8*8 block of pixels; Described complex characteristic vector comprises the constant proper vector in direction.
20. systems according to claim 11, is characterized in that, also comprise:
Input block, utilizes picture catching mode to input the image in described card face to be identified;
Authentication unit, verifies to confirm final recognition result to the described character information identified.
Priority Applications (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310216817.8A CN104217203B (en) | 2013-06-03 | 2013-06-03 | Complex background card face information identifying method and system |
| CN201910468824.4A CN110222687B (en) | 2013-06-03 | 2013-06-03 | Complex background card surface information identification method and system |
| HK15103317.3A HK1202954A1 (en) | 2013-06-03 | 2015-04-01 | Method for recognizing information located on surface of card in complex background and system thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310216817.8A CN104217203B (en) | 2013-06-03 | 2013-06-03 | Complex background card face information identifying method and system |
Related Child Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910468824.4A Division CN110222687B (en) | 2013-06-03 | 2013-06-03 | Complex background card surface information identification method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104217203A true CN104217203A (en) | 2014-12-17 |
| CN104217203B CN104217203B (en) | 2019-08-23 |
Family
ID=52098668
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910468824.4A Active CN110222687B (en) | 2013-06-03 | 2013-06-03 | Complex background card surface information identification method and system |
| CN201310216817.8A Active CN104217203B (en) | 2013-06-03 | 2013-06-03 | Complex background card face information identifying method and system |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910468824.4A Active CN110222687B (en) | 2013-06-03 | 2013-06-03 | Complex background card surface information identification method and system |
Country Status (2)
| Country | Link |
|---|---|
| CN (2) | CN110222687B (en) |
| HK (1) | HK1202954A1 (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106469293A (en) * | 2015-08-21 | 2017-03-01 | 上海羽视澄蓝信息科技有限公司 | The method and system of quick detection target |
| CN106503703A (en) * | 2015-09-08 | 2017-03-15 | Sk普兰尼特有限公司 | System and method of the using terminal equipment to recognize credit card number and due date |
| CN106527674A (en) * | 2015-09-14 | 2017-03-22 | 上海羽视澄蓝信息科技有限公司 | Human-computer interaction method, equipment and system for vehicle-mounted monocular camera |
| CN107256384A (en) * | 2017-05-22 | 2017-10-17 | 汕头大学 | A kind of card recognition and method of counting based on image and signal transacting |
| CN107977665A (en) * | 2017-12-15 | 2018-05-01 | 北京科摩仕捷科技有限公司 | The recognition methods of key message and computing device in a kind of invoice |
| CN104809157B (en) * | 2015-03-25 | 2018-05-18 | 小米科技有限责任公司 | Number identification method and device |
| CN108596180A (en) * | 2018-04-09 | 2018-09-28 | 深圳市腾讯网络信息技术有限公司 | Parameter identification, the training method of parameter identification model and device in image |
| CN108764230A (en) * | 2018-05-30 | 2018-11-06 | 上海建桥学院 | A kind of bank's card number automatic identifying method based on convolutional neural networks |
| CN108876716A (en) * | 2017-05-11 | 2018-11-23 | Tcl集团股份有限公司 | Super resolution ratio reconstruction method and device |
| CN110088770A (en) * | 2016-12-28 | 2019-08-02 | 欧姆龙健康医疗事业株式会社 | Terminal installation |
| CN110619325A (en) * | 2018-06-20 | 2019-12-27 | 北京搜狗科技发展有限公司 | Text recognition method and device |
| CN111242932A (en) * | 2020-01-15 | 2020-06-05 | 中国银联股份有限公司 | Card style checking method and device |
| CN111814789A (en) * | 2020-07-15 | 2020-10-23 | 中国建设银行股份有限公司 | Card number detection method, device, equipment and storage medium |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110796146A (en) * | 2019-10-11 | 2020-02-14 | 上海上湖信息技术有限公司 | Bank card number identification method, model training method and device |
| CN111860487B (en) * | 2020-07-28 | 2022-08-19 | 天津恒达文博科技股份有限公司 | Inscription marking detection and recognition system based on deep neural network |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060222239A1 (en) * | 2005-03-31 | 2006-10-05 | Bargeron David M | Systems and methods for detecting text |
| CN201054150Y (en) * | 2006-12-15 | 2008-04-30 | 华南理工大学 | An automatic recognition system for post envelope |
| CN101373519A (en) * | 2007-08-20 | 2009-02-25 | 富士通株式会社 | Device and method for recognizing character |
| CN102147858A (en) * | 2011-03-31 | 2011-08-10 | 重庆大学 | License plate character identification method |
| CN102163287A (en) * | 2011-03-28 | 2011-08-24 | 北京邮电大学 | Method for recognizing characters of licence plate based on Haar-like feature and support vector machine |
| CN103020647A (en) * | 2013-01-08 | 2013-04-03 | 西安电子科技大学 | Image classification method based on hierarchical SIFT (scale-invariant feature transform) features and sparse coding |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100601957B1 (en) * | 2004-07-07 | 2006-07-14 | 삼성전자주식회사 | Apparatus for and method for determining image correspondence, apparatus and method for image correction therefor |
| CN101038686B (en) * | 2007-01-10 | 2010-05-19 | 北京航空航天大学 | Method for recognizing machine-readable travel certificate |
| CN102324044A (en) * | 2011-09-09 | 2012-01-18 | 上海合合信息科技发展有限公司 | Card information acquisition method and system |
-
2013
- 2013-06-03 CN CN201910468824.4A patent/CN110222687B/en active Active
- 2013-06-03 CN CN201310216817.8A patent/CN104217203B/en active Active
-
2015
- 2015-04-01 HK HK15103317.3A patent/HK1202954A1/en unknown
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060222239A1 (en) * | 2005-03-31 | 2006-10-05 | Bargeron David M | Systems and methods for detecting text |
| CN201054150Y (en) * | 2006-12-15 | 2008-04-30 | 华南理工大学 | An automatic recognition system for post envelope |
| CN101373519A (en) * | 2007-08-20 | 2009-02-25 | 富士通株式会社 | Device and method for recognizing character |
| CN102163287A (en) * | 2011-03-28 | 2011-08-24 | 北京邮电大学 | Method for recognizing characters of licence plate based on Haar-like feature and support vector machine |
| CN102147858A (en) * | 2011-03-31 | 2011-08-10 | 重庆大学 | License plate character identification method |
| CN103020647A (en) * | 2013-01-08 | 2013-04-03 | 西安电子科技大学 | Image classification method based on hierarchical SIFT (scale-invariant feature transform) features and sparse coding |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104809157B (en) * | 2015-03-25 | 2018-05-18 | 小米科技有限责任公司 | Number identification method and device |
| CN106469293A (en) * | 2015-08-21 | 2017-03-01 | 上海羽视澄蓝信息科技有限公司 | The method and system of quick detection target |
| CN106503703A (en) * | 2015-09-08 | 2017-03-15 | Sk普兰尼特有限公司 | System and method of the using terminal equipment to recognize credit card number and due date |
| CN106503703B (en) * | 2015-09-08 | 2019-12-20 | 11街股份有限公司 | System and method for recognizing credit card number and expiration date using terminal device |
| CN106527674A (en) * | 2015-09-14 | 2017-03-22 | 上海羽视澄蓝信息科技有限公司 | Human-computer interaction method, equipment and system for vehicle-mounted monocular camera |
| CN110088770A (en) * | 2016-12-28 | 2019-08-02 | 欧姆龙健康医疗事业株式会社 | Terminal installation |
| CN110088770B (en) * | 2016-12-28 | 2023-07-07 | 欧姆龙健康医疗事业株式会社 | Terminal device |
| CN108876716A (en) * | 2017-05-11 | 2018-11-23 | Tcl集团股份有限公司 | Super resolution ratio reconstruction method and device |
| CN107256384A (en) * | 2017-05-22 | 2017-10-17 | 汕头大学 | A kind of card recognition and method of counting based on image and signal transacting |
| CN107977665A (en) * | 2017-12-15 | 2018-05-01 | 北京科摩仕捷科技有限公司 | The recognition methods of key message and computing device in a kind of invoice |
| CN108596180A (en) * | 2018-04-09 | 2018-09-28 | 深圳市腾讯网络信息技术有限公司 | Parameter identification, the training method of parameter identification model and device in image |
| CN108764230A (en) * | 2018-05-30 | 2018-11-06 | 上海建桥学院 | A kind of bank's card number automatic identifying method based on convolutional neural networks |
| CN110619325A (en) * | 2018-06-20 | 2019-12-27 | 北京搜狗科技发展有限公司 | Text recognition method and device |
| CN110619325B (en) * | 2018-06-20 | 2024-03-08 | 北京搜狗科技发展有限公司 | Text recognition method and device |
| CN111242932A (en) * | 2020-01-15 | 2020-06-05 | 中国银联股份有限公司 | Card style checking method and device |
| CN111814789A (en) * | 2020-07-15 | 2020-10-23 | 中国建设银行股份有限公司 | Card number detection method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104217203B (en) | 2019-08-23 |
| CN110222687B (en) | 2021-02-26 |
| CN110222687A (en) | 2019-09-10 |
| HK1202954A1 (en) | 2015-10-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104217203A (en) | Complex background card face information identification method and system | |
| CN105528604B (en) | A kind of bill automatic identification and processing system based on OCR | |
| CN109308476B (en) | Billing information processing method, system and computer readable storage medium | |
| CN104217202B (en) | Information identifying method, equipment and system | |
| US8744196B2 (en) | Automatic recognition of images | |
| CN108717543B (en) | Invoice identification method and device and computer storage medium | |
| CN104809452A (en) | Fingerprint identification method | |
| Xie et al. | A robust license plate detection and character recognition algorithm based on a combined feature extraction model and BPNN | |
| US20090169072A1 (en) | Method and system for comparing prints using a reconstructed direction image | |
| CN104809453A (en) | Authentication method based on fingerprints | |
| CN106874901B (en) | Driving license identification method and device | |
| CN109766891B (en) | Method for acquiring equipment facility information and computer readable storage medium | |
| US10354134B1 (en) | Feature classification with spatial analysis | |
| CN110969166A (en) | Small target identification method and system in inspection scene | |
| CN111340035A (en) | Train ticket identification method, system, equipment and medium | |
| CN103383732A (en) | Image processing method and device | |
| Li et al. | Fast and effective text detection | |
| CN111079571A (en) | Identification card information identification and edge detection model training method and device | |
| CN110046085B (en) | Method and device for identifying application program control displayed on terminal equipment | |
| CN113158895A (en) | Bill identification method and device, electronic equipment and storage medium | |
| CN105468732A (en) | Image keyword inspecting method and device | |
| CN103488966A (en) | Intelligent mobile phone capable of identifying real-name ticket information | |
| CN112001200A (en) | Identification code identification method, device, equipment, storage medium and system | |
| CN112052702A (en) | Method and device for identifying two-dimensional code | |
| CN111008635A (en) | OCR-based multi-bill automatic identification method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1202954 Country of ref document: HK |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |