CN101009747A

CN101009747A - The method for accurate digit extraction based on multiple OCR scheme combination verification

Info

Publication number: CN101009747A
Application number: CN 200710000525
Authority: CN
Inventors: 刘强
Original assignee: Individual
Current assignee: Individual
Priority date: 2007-01-10
Filing date: 2007-01-10
Publication date: 2007-08-01
Anticipated expiration: 2027-01-10
Also published as: CN100556062C

Abstract

The disclosed method to accurate extract data corrected based on multiple OCR schemes comprises: virtual printing, recognizing and configuring by OCR, scheduling the software recognization task, OCR recognizing, fault-tolerance correcting for single number digit result, correcting and matching result integrated; only full matching, starting fax transmission, or else manual operating to ensure the security. It also discloses the corresponding system.

Description

Based on the method for multiple OCR scheme combination verification with accurate extraction numeral

Technical field

The present invention relates to a kind of method of extracting numeral by OCR, relate in particular to a kind of based on the method for multiple OCR scheme combination verification with accurate extraction numeral.

Background technology

When more and more enterprises is chosen software, all pay much attention to integrated and degree of integration with the existing large scale system of enterprise, and enterprise itself does not wish that the existed system of oneself does change, the very strict system of encapsulation such as the R/3 of especially world-class large-scale management software system such as SAP change, and the change difficulty is very high, and enterprise itself does not often have this type of engineering development personnel.At present, with regard to fasystem, the user is starved of and carries out integrated with systems such as the existing ERP of enterprise, reach and need not manual intervention, need not to move the fasystem interface, directly in systems such as ERP, send fax, reach the integrated of ERP and electronic fax system, in the situation of not changing any existing systems such as ERP, at present, a kind of scheme is arranged, adopt the virtual fax printer, after various systems such as the ERP generation forms that need send are sent, to picture format (as TIF) file through generating after the virtual fax printer prints, carry out OCR identification, find out fax number numeral wherein, automatically on the backstage this part fax paper, send automatically according to the fax number that identifies, need not manual intervention, also need not systems such as ERP and do any change.

But with regard to present this OCR identifying schemes, can't really get up in the fasystem sector application always, reason just is 3 points: though the unusual standard of the fax picture file that one comes out through the virtual fax printer, but no matter any OCR identification software, all exist a spot of identification error, can be fax number digit recognition mistake; Two, system's forms such as ERP of enterprise itself, especially the purchase order of fax amount maximum, sales slip etc. all belong to secret sensitive information, never allow to occur sending out wrong situation, otherwise cause the loss that is difficult to retrieve; Three, fax has legal effect, send cautiously.

Based on above 3 points, the fax number identification error of the certain probability of OCR, may cause system's form, send out the misconnection debit as purchase order, but be subjected to legal recognition, can bring the loss that is difficult to retrieve to enterprise, therefore up to the present, also do not have good scheme, do not needing system of enterprise oneself own to do the situation of change, solve the integration problem of fasystem and its business system.

The OCR recognition methods of number of different types is disclosed in the prior art, comprise Chinese king OCR, the OCR of Qing Hua Ziguang, Microsoft Office2003 OCR, civilian pure OCR etc., these OCR softwares all have to provide specializes in the SDK interface that integrated exploitation is used, use the integrated exploitation of this type of OCR SDK, only need to load dynamic base, configure identification parameter, call corresponding discriminance analysis function interface and can realize OCR identification.

Yet, do not have a kind of method or system the organic combination of these OCR recognition methodss and the software systems that are applied to fax extracted numeral to form the higher recognition methods of discrimination.

Summary of the invention

This patent technical issues that need to address, i.e. the purpose of this patent is for a method is provided, make fasystem can with business system combination ideally, and need not to change business system itself.

For achieving the above object, the invention provides a kind of based on multiple OCR scheme combination verification with the accurate method of extracting numeral, it comprises,

The virtual printing step provides the virtual printer interface, in order to by the virtual printing to document, generates a file that can carry out OCR identification;

OCR discerns configuration step, in order to carry out relevant configuration, the fault tolerant mechanism configuration of the configuration of OCR number identidication key, each OCR software, verification scheme configuration, check results circulation configuration;

OCR software identification mission adaptation step, be responsible for receiving the fax picture task of submitting to from business system, carry out formation queuing distribution, first in first out, handle successively and distribute at least one OCR software identification number numeral, and the recognition rule that each OCR software is set by the configuration of OCR identification dispensing unit is set;

OCR software identification step, use at least two kinds of OCR softwares, call the SDK development interface recognition function of OCR software, fax picture file path and relevant identification configuration parameter are sent into recognition function, begin the fax picture that submission is come is discerned respectively, the final spanned file content that identifies is analyzed,, obtained an original recognition result at this software according to information such as the configuration cut-off point of each OCR identification dispensing unit, keywords;

The fault-tolerant error correction step of the single recognition result of number, original recognition result according to each software, at first do a fault-tolerant filtration treatment, all spcial characters in the number that identifies are removed, and at the feature of each OCR software algorithm and the characteristic of digital number, numeral is carried out once fault-tolerant, correction process again, obtain a filtered recognition result;

A step is joined in the whole recognition result verification of numeral, to the filtered recognition result at each OCR software, carries out the verification coupling, whether in full accordly see, draw a final number result, as each OCR identify digital in full accord, then generate a definite final recognition result; If find inconsistently, then according to prior configuration, determine one of them OCR recognition result priority principle, take out this OCR recognition result as final result, but be labeled as uncertain result.

The present invention adopts the OCR software identification fax picture of multiple different identification styles, obtain correlated digitals such as fax number wherein exactly, according to recognizer feature separately the numeral that identifies is carried out fault-tolerant, error correction then, all fax number numerals that identify are unified verification at last, have only after joining fully, just start facsimile transmission automatically, otherwise send, guarantee fail safe by manual intervention.

Adopt this kind method, can guarantee the system integration of height fully, guaranteed that also fax can not send out the recipient wrong, bring loss, fasystem and enterprise's existing system Integrated Solution can really should be used in enterprise to enterprise.Can be from following three aspects demonstration: at first, because the picture file that the picture of identification all be regular form passes through the virtual fax printer output forms such as () TIF, file format is very regular, and definition is also relatively good, for fasystem, only need identification correlated digital wherein, as fax number, telephone number etc., OCR itself is very high for the recognition accuracy of numeral, basically all reach more than 99%, guarantee the accurate integration of system of height; Since secondly OCR itself also may exist 1% error, but this patent passes through type selecting, adopt two or two above OCR softwares, OCR is algorithm deviation to some extent each other, all fax picture is once discerned, the fax number numeral that multiple OCR software is identified is carried out verification then, because the different identifying schemes OCR that when OCR software type selecting, select for use, therefore the overlapping probability of 1% erroneous judgement each other is zero substantially, therefore the fax number numeral that occurs is in case upcheck, join fully, then correct certainly, need not to worry; Even join by this kind verification once more, wilfully occur under the non-existent substantially situation of identification error probability, also it doesn't matter, this fax number will be sent out successfully, and the probability of fax still is impossible, particularity because of fax, need many conditions to satisfy, so fax number must exist, and the fax number recipient just is that facsimile machine just works.

One of these or other feature and advantage of the present invention or part will become apparent those skilled in the art by following specification, in specification, by the optimal mode that description carries out an invention, provide and set forth a kind of embodiment of the present invention.To be realized as it, the present invention can have different execution modes, and in the case of without departing from the present invention, its some details also can have multiple variant.In view of the above, accompanying drawing and specification only should be counted as descriptive and nonrestrictive.

Description of drawings

Fig. 1 is the flow chart of method of the present invention.

Fig. 2 is the system diagram of system of the present invention.

Embodiment

Show according to a specific embodiment of the present invention a kind of based on multiple OCR scheme combination verification with reference to Fig. 1 and may further comprise the steps with the accurate method of extracting numeral:

Virtual printing step 800 provides the virtual printer interface, in order to by the virtual printing to document, generates a file that can carry out OCR identification, as TIF fax picture file;

OCR discerns configuration step 801, in order to carrying out relevant configuration, the fault tolerant mechanism configuration of the configuration of OCR number identidication key, each OCR software, verification scheme configuration, check results circulation configuration, this step can all setting completed in advance, adopts acquiescence this moment;

Multiple OCR software identification mission adaptation step 802, be responsible for receiving the fax picture task of submitting to from business system, carry out formation queuing distribution, first in first out, handle successively and distribute to each OCR software identification number numeral, and the recognition rule that each OCR software is set by the configuration of OCR identification configuration step is set;

Multiple OCR software identification step 803, select multiple OCR software for use, here, preferably select dissimilar OCR software for use, at each software, call the SDK development interface recognition function of OCR software, fax picture file path and relevant identification configuration parameter are sent into recognition function, begin the fax picture that submission is come is discerned respectively, the final spanned file content that identifies is analyzed, according to information such as the configuration cut-off point of each OCR identification configuration step, keywords, obtain an original recognition result at this software;

The fault-tolerant error correction step 804 of the single recognition result of number, according to original recognition result, at first do a fault-tolerant filtration treatment, all spcial characters in the number that identifies are removed, and at the feature and the digital number (fax number of each OCR software algorithm, telephone number etc.) characteristic, carry out once fault-tolerant again to numeral, correction process, be identified as " one " easily as "-" in the number " 0757-22362099 ", own feature according to digital number, can not there be " one ", therefore can be converted into "-" to " one " that is clipped in the numeral automatically; Be identified as 5 to 8 easily under special circumstances as OFFICE2003 OCR SDK, need other OCR SDK recognition results of contrast this moment,, 5 of wrong identification is corrected as 8 etc. automatically, finally obtain filtered recognition result at software according to multiple authentication;

A step 805 is joined in the whole recognition result verification of numeral, the filtered recognition result that each OCR software is identified, carry out the verification coupling, whether in full accordly see, draw a final number result, as each OCR identify digital in full accord, then discern errorlessly, generate a definite recognition result; If find inconsistently, then according to prior configuration, determine one of them OCR recognition result priority principle, take out this OCR recognition result as final result, but be labeled as uncertain result;

At print application, can also comprise the check results execution in step, in order to the recognition result of determining according to generation, the configuration of discerning configuration step in conjunction with OCR decides the direct transmission of fax to result, still artificial transmission; Or, according to uncertain recognition result, then must come out by obvious mark, to handle by manual intervention, hand inspection contrast fax picture is determined to send after errorless again, and the facsimile transmission fail safe is guaranteed sending in wrong then artificial corrigendum back.

Fig. 2 has provided based on of the present invention a kind of based on the system of multiple OCR scheme combination verification with accurate extraction numeral, and described system comprises

Virtual printer device 100 provides the virtual printer interface, in order to by the virtual printing to document, generates a file that can carry out OCR identification, as TIF fax picture file;

OCR discerns inking device 101, in order to carrying out relevant configuration, the fault tolerant mechanism configuration of the configuration of OCR number identidication key, each OCR software, verification scheme configuration, check results circulation configuration, this step can all setting completed in advance, adopts acquiescence this moment;

Multiple OCR software identification mission deployment device 102, be responsible for receiving the fax picture task of submitting to from business system, carry out formation queuing distribution, first in first out, handle successively and distribute to each OCR software identification number numeral, and the recognition rule that each OCR software is set by the configuration of OCR identification inking device is set;

Multiple OCR software recognition device 103, select multiple OCR software for use, here, preferably select dissimilar OCR software for use, at each software, call the SDK development interface recognition function of OCR software, fax picture file path and relevant identification configuration parameter are sent into recognition function, begin the fax picture that submission is come is discerned respectively, the final spanned file content that identifies is analyzed, according to information such as the configuration cut-off point of each OCR identification inking device, keywords, produce an original recognition result at this software;

The fault-tolerant error correction device 104 of the single recognition result of number, according to original recognition result, at first do a fault-tolerant filtration treatment, all spcial characters in the number that identifies are removed, and at the feature and the digital number (fax number of each OCR software algorithm, telephone number etc.) characteristic, carry out once fault-tolerant again to numeral, correction process, be identified as " one " easily as "-" in the number " 0757-22362099 ", own feature according to digital number, can not there be " one ", therefore can be converted into "-" to " one " that is clipped in the numeral automatically; Be identified as 5 to 8 easily under special circumstances as OFFICE2003 OCR SDK, need other OCR SDK recognition results of contrast this moment,, 5 of wrong identification is corrected as 8 etc. automatically, finally obtain filtered recognition result at software according to multiple authentication;

A device 105 is joined in the whole recognition result verification of numeral, the filtered recognition result that each OCR software is identified, carry out the verification coupling, whether in full accordly see, draw a final number result, as each OCR identify digital in full accord, then discern errorlessly, generate a definite recognition result; If find inconsistently, then according to prior configuration, determine one of them OCR recognition result priority principle, take out this OCR recognition result as final result, but be labeled as uncertain result;

At print application, can also comprise the check results final controlling element, in order to the recognition result of determining according to generation, the configuration of discerning inking device in conjunction with OCR decides the direct transmission of fax to result, still artificial transmission; Or, according to uncertain recognition result, then must come out by obvious mark, to handle by manual intervention, hand inspection contrast fax picture is determined to send after errorless again, and the facsimile transmission fail safe is guaranteed sending in wrong then artificial corrigendum back.

Those skilled in the art should be understood that above the specific embodiment of the present invention in specification and accompanying drawing description only is illustrative rather than restrictive.

Aforementioned explanation to the specific embodiment of the invention is only with example be described as purpose.It is not intended to execution mode or a certain fixed pattern that each execution mode of limit maybe is confined to invention to have disclosed.In view of the above, foregoing description should be counted as exemplary but not circumscribed.Obviously be much to revise and variant visible concerning those skilled in the art.Having selected and described each embodiment is in order the most clearly to explain principle of the present invention and best practical application mode, to understand the various embodiment and the various variant of special purpose that is suitable for expecting of the present invention or realization to make those skilled in the art.Be intended to limit scope of the present invention by follow-up claim and equivalence variation thereof, in the claims, unless point out separately, all terms all should be interpreted as its widest implication.Should be appreciated that those skilled in the art can make variation under the situation that does not break away from the scope of the present invention that is defined by following claim in embodiment.In addition, no matter whether following claim has clearly described some key element and assembly, all is not intended to these key elements and assembly in having now are openly offered in known field.

Claims

One kind based on multiple OCR scheme combination verification with the accurate method of extracting numeral, it is characterized in that: may further comprise the steps

The virtual printing step provides the virtual printer interface, in order to by the virtual printing to document, generates a file that can carry out OCR identification;

OCR discerns configuration step, in order to carry out relevant configuration, the fault tolerant mechanism configuration of the configuration of OCR number identidication key, each OCR software, verification scheme configuration, check results circulation configuration;

OCR software identification mission adaptation step, be responsible for receiving the fax picture task of submitting to from business system, carry out formation queuing distribution, first in first out, handle successively and distribute at least one OCR software identification number numeral, and the recognition rule that each OCR software is set by the configuration of OCR identification dispensing unit is set;

OCR software identification step, use at least two kinds of OCR softwares, call the SDK development interface recognition function of OCR software, fax picture file path and relevant identification configuration parameter are sent into recognition function, begin the fax picture that submission is come is discerned respectively, the final spanned file content that identifies is analyzed,, obtained an original recognition result at this software according to information such as the configuration cut-off point of each OCR identification dispensing unit, keywords;

The fault-tolerant error correction step of the single recognition result of number, original recognition result according to each software, at first do a fault-tolerant filtration treatment, all spcial characters in the number that identifies are removed, and at the feature of each OCR software algorithm and the characteristic of digital number, numeral is carried out once fault-tolerant, correction process again, obtain a filtered recognition result;

A step is joined in the whole recognition result verification of numeral, to the filtered recognition result at each OCR software, carries out the verification coupling, whether in full accordly see, draw a final number result, as each OCR identify digital in full accord, then generate a definite final recognition result; If find inconsistently, then according to prior configuration, determine one of them OCR recognition result priority principle, take out this OCR recognition result as final result, but be labeled as uncertain result.
2. as claimed in claim 1 based on the method for multiple OCR scheme combination verification with accurate extraction numeral, it is characterized in that: described OCR software is dissimilar OCR software.
One kind based on multiple OCR scheme combination verification with the accurate system that extracts numeral, it is characterized in that: comprise

Virtual printer device provides the virtual printer interface, in order to by the virtual printing to document, generates a file that can carry out OCR identification;

OCR discerns inking device, in order to carry out relevant configuration, the fault tolerant mechanism configuration of the configuration of OCR number identidication key, each OCR software, verification scheme configuration, check results circulation configuration;

OCR software identification mission deployment device, be responsible for receiving the fax picture task of submitting to from business system, carry out formation queuing distribution, first in first out, handle successively and distribute at least one OCR software identification number numeral, and the recognition rule that each OCR software is set by the configuration of OCR identification inking device is set;

OCR software recognition device, use at least two kinds of OCR softwares, call the SDK development interface recognition function of OCR software, fax picture file path and relevant identification configuration parameter are sent into recognition function, begin the fax picture that submission is come is discerned respectively, the final spanned file content that identifies is analyzed,, obtained an original recognition result at this software according to information such as the configuration cut-off point of each OCR identification inking device, keywords;

The fault-tolerant error correction device of the single recognition result of number, original recognition result according to each software, at first do a fault-tolerant filtration treatment, all spcial characters in the number that identifies are removed, and at the feature of each OCR software algorithm and the characteristic of digital number, numeral is carried out once fault-tolerant, correction process again, obtain a filtered recognition result;

A device is joined in the whole recognition result verification of numeral, to the filtered recognition result at each OCR software, carries out the verification coupling, whether in full accordly see, draw a final number result, as each OCR identify digital in full accord, then generate a definite final recognition result; If find inconsistently, then according to prior configuration, determine one of them OCR recognition result priority principle, take out this OCR recognition result as final result, but be labeled as uncertain result.
4. as claimed in claim 3 based on the system of multiple OCR scheme combination verification with accurate extraction numeral, it is characterized in that: described OCR software is dissimilar OCR software.