Method for processing postal items with optical character recognition ocr

The invention relates to a method of processing postal items in a sorting machine having sorting outlets with sorting outlet bins into which the items are directed during successive sorting passes, in which during a first pass for sorting postal items, a digital image is formed of each postal item comprising address information to automatically recognize, in particular by OCR (Optical character Recognition), the address of the item, a digital imprint which is a logical identifier for the shipment is derived from this image and address data representative of the address of the shipment recognized automatically in correspondence with the digital imprint of the shipment is recorded in memory and a logical identifier of the sorting output bin in which the shipment is directed, and in which during a second sorting pass of the postal items, a digital image of a current shipment comprising the address information of the current shipment is again formed, a digital fingerprint for the current shipment is derived from this image and a search is made among the digital fingerprints recorded in the first pass of sorting a match with the digital fingerprint of the current shipment in order to retrieve in particular the address data representative of the address information of the current shipment

More particularly, the invention relates to a method for automatically sorting postal items in several sorting passes on one or more sorting machines in which it is not necessary to affix (by printing with or without a label) on the items. postal items, a shipment identifier (called "chronomarque or Tag-ld in current literature), in particular an identifier in the form of a bar code. In the method according to the invention, the items are identified by a digital imprint or image signature as described in European patent document EP1222037. The use of digital fingerprints or image signatures to identify shipments during successive sorting passes avoids the use of barcode printing and reading devices, the use of verification devices. the print quality of bar codes (called BCV in current literature), the use of labeling machines for sticking labels on postal items having a plastic surface (in particular for items of the flat object type such as magazines and newspapers) and this thus results in a substantial saving in the cost of the printing machine. sorting and the cost of its maintenance but also on the consumables during the operation of the machine.

In patent document EP1222037, a bar code serving as an identifier for the bin is affixed to one wall of each sorting output bin. During the second sorting pass, the code of the bins loaded at the input of the machine is read and an association is made between these bins and the digital imprints of the items contained in these bins so that, when searching for a match in the second pass between the digital fingerprint of a current shipment and the digital fingerprints recorded in the first pass, an exploration space is exploited which is limited to the footprints of the items of a single first-pass sorting output bin so as to meet the requirements of real-time processing. But in this known method, the principle used to reduce the space for exploring digital fingerprints does not allow items contained in several sorting output bins from a first sorting pass in a machine, being transhipped into a single bin to be recycled at the input of another machine to continue the second sorting pass. With this known method, it is also not possible to have breaks in the sequence of the bins, for example the fact of introducing in the second sorting pass items contained in a bin which is not an output bin. of sorting from the first sorting pass. With this known process, it is imperative to keep in the second pass, the relative order of the items in each recycled bin entering the sorting machine. This known method is therefore not resistant to the disturbances which may occur in the supply store of the sorting machine, for example following a jam of the machine.

The aim of the invention is to provide a method for sorting postal items in several sorting passes which uses digital fingerprints to identify the postal items and which is not sensitive to the operating constraints indicated above.

To this end, the invention relates to a method for processing postal items in a sorting machine having sorting outlets with sorting outlet bins into which the items are directed during successive sorting passes, in which during a first sorting pass of the postal items, a digital image is formed of each postal item comprising address information to automatically recognize the address of the item, a digital imprint which is a logical identifier for the shipment is derived from this image and address data representative of the address of the shipment recognized automatically in correspondence with the digital imprint of the shipment is recorded in memory and a logical identifier of the sorting output bin in which the shipment is directed, and in which during a second sorting pass of the postal items, a digital image of a current shipment comprising the address information of the current shipment is again formed, a digital fingerprint for the current shipment is derived from this image and a search is made among the digital fingerprints recorded in the first pass of sorting a match with the digital fingerprint of the current shipment, characterized in that during the second sorting pass, an automatic address recognition is also carried out on the basis of the image of the current shipment to obtain address data representative of the address of the current shipment, a search is made among the address data stored in memory for a match with the address data of the current shipment to identify, from the logical identifiers of the sorting output bin, a current sorting output bin of the first sorting pass to which the current shipment corresponds so that in order to search for a match between two digital fingerprints, this search for a match of digital fingerprints is carried out in an exploration space formed by the digital fingerprints which correspond to said current sorting output bin.

The idea at the base of the method according to the invention is therefore to automatically identify mailing bins thanks to the recognition by OCR of a few items from the bin in order to reduce the footprint exploration space. The method according to the invention applies to a treatment in which the successive sorting passes can be made on several sorting machines (including with a segmentation of the sorting, routing / sorting distribution type) equipped with single sorting output bins or compartmentalized. The method according to the invention applies to a processing of items in which the search for a match of fingerprints is used to retrieve data indicative of the address information but can in the same way be applied to a processing of 'consignments in which the search for matching fingerprints is used to carry out a follow-up (traceability) of the consignments in the machines.

The method according to the invention can have the following features:

- To identify a current sorting output bin from the first sorting pass, the occurrences for which a match between the address data is obtained is counted for successive shipments;

- if no match is detected between the digital fingerprint of the current shipment and the digital fingerprints forming the exploration space, the current shipment is put on hold and the search for a match of digital fingerprints is started again for this dispatch running after an update of the exploration space with digital imprints which correspond to another sorting output bin of the first sorting pass;

the number of times that a current item is put on hold is counted and the current item is rejected if it is detected that the number of times that this current item has been put on hold is greater than a certain threshold.

An example of an embodiment of the method according to the invention is described in more detail below and illustrated by the drawings. This description is given only by way of indicative and in no way limiting example of the invention.

FIG. 1 very schematically illustrates a postal item.

FIG. 2 schematically shows a sorting machine for implementing the method according to the invention.

FIG. 3 illustrates in the form of a flowchart the progress of the method according to the invention during a first sorting pass.

FIG. 4 illustrates the organization of the data in the memory of the sorting machine.

FIG. 5 illustrates in the form of a flowchart the progress of the method according to the invention during a second sorting pass.

In Figure 1, there is illustrated a mail item P comprising delivery address information A. This mail item may be a single letter or a small or large format flat object such as a magazine, a newspaper, a catalog with plastic or paper envelope.

In Figure 2, there is schematically shown a sorting machine 1 implementing the method according to the invention.

The sorting machine 1 conventionally comprises a feed inlet 2 with a magazine and an unstacker for the serialization on edge of postal items such as P, a digital camera 3 for the acquisition of the image of the surface of each postal item P comprising the distribution address information A and a bucket carousel 4 which directs the items to sorting outlets each provided with a sorting outlet bin 5. Each output can be equipped with several sorting output bins (typically a front bin and a rear bin) or with a compartmentalized bin without departing from the scope of the invention.

In FIG. 2, there is also represented by 6 a data processing system associated with a video-coding system 7. The systems 6 and 7 constitute an address recognition system by OCR which, on the basis of an image d 'a shipment P formed by the camera 3, extracts automatically or via the intervention of a video-coding operator, the address information A on the basis of which is determined a sorting destination for the shipment which corresponds at a sorting outlet of the machine. In addition, the system 6 is arranged according to the invention to derive from each digital image of a shipment formed by the camera 3 a digital imprint or image signature constituting a logical identifier for the shipment.

In FIG. 2, there is also represented by 9 a system similar to system 6 but which forms part of another sorting machine 10, this system 9 being connected in communication, for example by a telecommunications network 8 to the system 6, in order to an application of the method according to the invention to a sorting process in several sorting passes on several sorting machines.

In the method according to the invention, the items are sorted in several sorting passes, for example in two sorting passes.

FIGS. 3 and 5 respectively represent the steps of the method according to the invention for a first sorting pass and for a second sorting pass on a single machine such as 1.

Referring to FIG. 3, in the initial step 11, mail is prepared to be sorted at the inlet of the machine 1 and empty bins are loaded into the sorting outlets of the machine. Logical identifiers Bac-ld are assigned in memory by machine 1 to each sorting output bin. As an example (Logical identifier for a bin, one can use for example a concatenation of the following information: Sorting site identifier / Sorting machine number / Sorting output number in the machine / Date / Time.

In step 12, the items are unstacked and serialized to pass in front of the camera 3. An image of the surface of each item is formed and stored in memory in the system 6.

At 13, on the basis of the image of a shipment, an automatic recognition by OCR of the address information A is carried out at the same time that it derives in the computer 6, from the image of the 'sending, a digital fingerprint V-ld serving as a logical identifier for the sending. In system 6, another logical identifier for the shipment, for example in the form of a number such as a time tag Tag-ld, can also be assigned to the shipment.

In step 14, if unambiguous address information A is obtained by automatic address recognition by OCR, the AD data representative of this address information for the item in question with the temporary item number Tag- id are matched to memory in step 16. In FIG. 3, 17 is represented by a database indicated by BD-OCR / VCS in which are recorded in correspondence for each shipment, the temporary Tag-id number assigned to the shipment and the data representative of the address. of the shipment, the Tag-id number serving as the access key in the database for the subsequent recovery of the address data during a second sorting pass. In step 14, if the result of the automatic address recognition by OCR gives a partial result, that is to say equivocal distribution address information, or even no result, the digital image of the the sending is sent (step 15) to the video-coding system 7 for extraction of the address information by a video coding operator, after which the address data retrieved through system 7 is registered in correspondence with a temporary send number Tag-ld in base 17 in step 16.

According to the method of the invention, in step 18, for the item in question, the digital fingerprint V-1 of the item (issue from step 13) in correspondence with the address data obtained automatically by OCR (at step 13), the temporary number of the shipment

Tag-ld and the logical identifier Bac-ld of the sorting output bin to which the shipment is directed. The address data recorded in the database 20 are data obtained exclusively by automatic address recognition by OCR, ie without the intervention of a video coding operator.

In step 19, the first sorting pass continues by conveying the item to the corresponding sorting outlet of the machine and unloading the item into the bin of this sorting outlet. This process loops back (not shown) in step 13 for each postal item unstacked at the entry of the machine with, where appropriate, management of the full bins during the first sorting pass.

FIG. 4 very schematically illustrates the structure of the records of the databases 17 and 20.

In particular, for the database 17, several records have been shown in each of which a temporary dispatch number TagId1, Tag-ld2, Tag-ldn, which serves as an access key, is recorded in correspondence with data of address AD1, AD2, Adn obtained by address recognition by OCR supplemented where appropriate by VCS video coding.

FIG. 4 shows the database 20 in a structured form which here consists of a set of tables each identified by a logical bac identifier Bac-ld1, Bac-ld2, Bac-ldi. Each table identified by Bac-ld contains a set of records corresponding respectively to items contained in a sorting output bin. Each record of a table contains here the temporary number assigned to a shipment Tag-ld1.1, ..., Tag-ldi.n, the digital fingerprint V-ld1.1 of the shipment, ...., V-ldi.n, and the address data AD-OCR1.1, ...., AD-OCRi.n of the shipment obtained exclusively by automatic recognition. An additional field indicated by M has been included in each record of a table, which serves as a marker in the course of the method according to the invention as explained below. It should be understood that the division of the tables 17 and 20 illustrated in FIG. 4 is given only by way of non-limiting example.

FIG. 5 represents the steps of the method according to the invention during a second shipment sorting pass.

In step 41, the sorted items contained in the sorting output bins of the first sorting pass are returned to the input of the sorting machine 1 and empty bins are introduced into the sorting outlets of the machine 1.

During the second sorting pass, the method according to the invention uses the database 20 and another temporary database indicated by in FIGS. 4 and 5 which contains digital fingerprints V-1d forming the exploration space. indicated by BR-RV-1d for finding a match between two digital fingerprints as described below. This temporary database 45 is initialized in step 41 and does not contain any record at this stage of the start of the second sorting pass.

In step 42, the items in the warehouse of the sorting machine are unstacked to pass in series in front of the camera 3. An image is again formed of the surface of each item comprising the address information A and in the. 'step 43, this image is derived from a current dispatch of a digital imprint V-1d.

In the method according to the invention, in step 43, an automatic address recognition is also carried out by OCR from the image of the current item in order to obtain again by OCR address data representative of the. address of the current shipment.

In step 44, a match is made between the digital imprint of the current shipment and the digital imprints recorded in the base 45. As indicated above, at the start of the second sorting pass, the base 45 is empty. which means that for the first current sending, in step 44 no match is found between two digital fingerprints. If no match is found in the step following step 44, the process continues at step 47 where it is determined whether the automatic address recognition by OCR performed in step 43 was successful. In practice, it will be possible for example to consider that the automatic address recognition by OCR carried out in step 43 has been successful if a routing code and a distribution code have been obtained which constitute the address data recorded in the base. 20.

In the case where in step 47, it is detected that the automatic address recognition by OCR has not succeeded in step 43, it is checked in step 48 whether the current postal item can be put on hold. . It may be considered initially that a postal item can be placed on standby for processing only over a defined period of time, for example over a period of 10 seconds and that beyond this period of time, the mail item put on hold must be rejected (step indicated by REJECT in Figure 5).

If in step 48, the current shipment can be put on hold, in step 49 a logical identifier of the shipment, for example the TagId, is kept in memory in a waiting list 50.

Now if in step 47, it has been detected that the automatic address recognition by OCR has succeeded in step 43, then in step a match is sought between the AD-OCR address data of the sending current and the AD-OCR address data recorded in database 20 to identify in base 20 one or more records for which or for which a match is detected, these records are identified by modifying the state of their marker M and the occurrences for which a match between address data is obtained is counted for each sorting output bin in the base 20.

In step 53, if a sufficient number of marked occurrences for a sorting output bin is not detected in base 20, the current mail item can be put on hold in step 49. If in step 53, on the other hand, it is detected in the base that for a sorting output bin, there is a number of marked occurrences greater than a certain threshold, for example three occurrences for the same bin, then in the step 54, we load into the base 45 (database BD-RV-ld) all the records of this sorting output bin in the database 20, these records then defining the exploration space for the search for a match between two digital fingerprints.

FIG. 4 shows the structure of the footprint exploration space formed by the database 45 which is formed by the prints of postal items from the bin indicated by Bac-ldj. There is illustrated in Figure 4, for the Bac-ldj tray for example, three records each comprising the digital fingerprint V-ld.j.1, V-Id.j.2, V-ld.jn of a shipment and the identifier Tag-ld.j.1, Tag-ld.j.2, Tg-ld.jn for shipments 1 to n of the bin. In practice, each sorting bin can contain a few dozen items, for example around fifty items, which represents the horizon (depth) ίο of the exploration space. In the method according to the invention, the exploration space can be formed by the digital imprints V-1d of the items of more than one bin, for example two bins as illustrated in FIG. 4 with the records of the bin indicated by Bac-ldk. In this case, we increase the size of the exploration space but we increase the probability of obtaining a match of footprints for situations where several sorting output bins in the first sorting pass have been emptied in the store d. feed the machine without respecting the fold sequence before starting the second sorting pass.

In this case, in step 54, provision can be made for a partial renewal of the database 45 by replacing, for example, the imprints of the items of the oldest bin in the base 45 by the imprints of the items of a bin for which one has just detected a crossing of the occurrence count threshold. Each time the database 45 is updated, the imprints for the shipments of one out of two containers are therefore replaced.

In step 55, if the list of pending items 50 is not empty, this list is scanned as a priority by proceeding, for each pending item indicated in the list, to steps 44 and 46 as for a current item. Normally, in step 46, a match is detected between the digital fingerprint of a shipment (retrieved from the waiting list 50) and one of the digital fingerprints recorded in the exploration space 45. From the association between the imprint found in the exploration space 45 and the identifier Tag-ld, the address data AD for the item in question is recovered in the database 17 (step 56). and sending the item to a corresponding sorting output (step 57). In step 58, the horizon of the exploration space is reduced by updating the database 45 by marking the record which corresponds to the digital fingerprint found in step 46 for possible deletion. The database 20 is also updated by initializing the markers 20 in the records corresponding to the bins which are not loaded into the database 45 in step 54. If necessary, the waiting list 50 is updated by deleting the identifier Tag-ld which corresponds to the digital fingerprint found in step 46.

As indicated above, step 44 is given priority to processing the items put on hold in the waiting list 50. If at step

46, no match is detected for this current shipment, this current shipment can be directly put on hold again in step 49. In step 49, if it is detected that a shipment is put on hold a number of times greater than a certain threshold, the item can be rejected (manual processing of step 51).

Following step 58, the process loops back to step 42 for processing a subsequent shipment from the waitlist 50 or another shipment.

If the second sorting pass is performed on another sorting machine such as 10, in step 41, the database 20 formed in machine 1 will be transferred to system 9.

As indicated above, the method according to the invention is resistant to disturbances in the order of shipments in the second sorting pass following a transshipment of bins or a break in the sequence, for example.