WO2008152277A2 - Procédé de traitement d'envois incluant une classification graphique des signatures associées aux envois - Google Patents
Procédé de traitement d'envois incluant une classification graphique des signatures associées aux envois Download PDFInfo
- Publication number
- WO2008152277A2 WO2008152277A2 PCT/FR2008/050885 FR2008050885W WO2008152277A2 WO 2008152277 A2 WO2008152277 A2 WO 2008152277A2 FR 2008050885 W FR2008050885 W FR 2008050885W WO 2008152277 A2 WO2008152277 A2 WO 2008152277A2
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- signatures
- image
- signature
- postal
- sorting
- Prior art date
Links
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
- B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
- B07C3/00—Sorting according to destination
- B07C3/10—Apparatus characterised by the means used for detection ofthe destination
- B07C3/14—Apparatus characterised by the means used for detection ofthe destination using light-responsive detecting means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
- G06V30/42—Document-oriented image-based pattern recognition based on the type of document
- G06V30/424—Postal images, e.g. labels or addresses on parcels or postal envelopes
Definitions
- Shipment processing method including a graphic classification of signatures associated with shipments
- the invention relates to a method for processing postal items in which, during a first sorting pass of the items, a digital image is formed of the surface of each item comprising information blocks, and each image associated with the item is sending a digital signature to the image of a graphic imprint which is a logical identifier for sending, and storing in memory the signature associated with the sending in correspondence with sorting information and in which during a second sorting passes of the shipments, one again forms a digital image of said surface of each shipment to derive a current digital signature associated with the shipment and one of the signatures registered in the first sorting pass is searched for a match with this current signature in order to retrieve by association the sort information.
- each digital signature comprises a first component called “image” representative of physical characteristics of the digital image of the corresponding sending and a second component called “postal” indicative at least of a spatial position of the blocks of information present in the image of the shipment.
- the image component is formed by so-called “global” attributes that are representative of global physical characteristics taken on all the points of the digital image of the shipment.
- the image component is formed of second so-called “local” attributes that are representative of local physical characteristics taken on distinct parts of a grid of the image of the shipment (or several different grids).
- the object of the invention is therefore to propose a more robust method of processing mail which makes it possible to reduce the mismatches indicated above, more particularly in situations where the postal items to be sorted come from number senders and therefore have a very high graphical similarity.
- the subject of the invention is a method for processing postal items in which, during a first sorting pass of the items, a digital image of the surface of each item containing information blocks is formed. derives from each image associated with a sending a digital signature which is a logical identifier for sending and stores in memory the signature associated with the sending in correspondence with sorting information and in which during a second sorting pass sendings, it again forms a digital image of said surface of each shipment to derive a current digital signature associated with the sending and searched among the signatures recorded in the first sorting pass a concordance with this current signature to recover by association the sorting information, characterized in that it comprises the following steps: - we group the signatures into categories of signatures or graphic classes according to a c ertain graphical similarity criterion for calculating an average signature value for each graphical class;
- the idea underlying the invention is therefore to classify the successive signatures from consolidated real-time signature graphic models on the fly to reflect through each model the non-changing graphic appearance of the images of the items whose signatures relate to the model in question.
- the activity analysis (or low frequency change) that is done at each consolidation or update of the model when grouping a new signature in the graphical class corresponding to this signature model, it is possible to display the position of the destination address block in the images of the items because this destination postal address block is probably the one with the most significant low frequency activity, unlike other blocks of information which are normally stable when considering the consignments of the same sender in number.
- the process according to the invention may have the following particularities:
- the signatures are grouped into graphic classes according to a certain criterion of graphic similarity of their image component to calculate an average value of the image components of the signatures belonging to the graphic class considered.
- the activity of the average value of the image component is analyzed to detect a significant activity of an information block in the digital images; the results of said activity detection are used to compare the postal components of the signatures;
- the global attributes and the local attributes of the signatures are used to establish a similarity of signatures; a thresholding of the normed vector distance is performed on the global attributes to establish the said similarity
- the method according to the invention can be implemented with any type of postal items such as letters, flat objects of small or large format.
- the method according to the invention can also be applied to the sorting of packets and other articles identified by a virtual identification code.
- Figure 1 very schematically illustrates a mailing with several blocks of information.
- FIG. 2 schematically shows a sorting machine for implementing the method according to the invention.
- FIG. 3 illustrates in the form of a flow chart the sequence of the known sorting process during a first sorting pass.
- Figure 4 illustrates very schematically the extraction of the image component of the signature of a postal item.
- Figure 5 illustrates very schematically the extraction of the postal component of the signature of a postal item.
- FIG. 6 illustrates in the form of a flowchart the progress of the method according to a first embodiment of the invention during a second sorting pass.
- Figure 7 details the step of classification of signatures in the method according to the invention.
- FIG. 8 represents the comparison of the image components during the classification of signatures in the method according to the invention.
- FIG. 9 represents an example of updating the image components of the graphic classes in the method according to the invention.
- Figure 10 details the step of modifying the postal component of the current mail from the image component of the corresponding graphics class.
- Figure 11 details the step of modifying the postal component of the current mail from the postal component of the corresponding graphics class.
- FIG. 12 illustrates in the form of a flowchart the flow of the method according to a second embodiment of the invention during a second sorting pass.
- Figure 13 illustrates the principle of a reduction of exploration space by a prediction mechanism
- postal items such as letters, flat objects of small or large size with a paper or plastic envelope, and other article to be sorted automatically by digital signatures which are derived from the image are identified. from the surface of shipments that usually has a destination postal address block.
- This digital signature still called image signature is therefore used to uniquely identify a shipment in place of a barcode chronomarch in an automatic postal sorting machine.
- the term machine refers broadly to postal sorting equipment installed on one or more postal sorting sites including video encoders where appropriate.
- FIG. 1 illustrates a postal item referenced P whose surface comprises, for example, a first AD information block which corresponds to the postal delivery address or destination address, an AE information block which corresponds to a sender address and an information block L that can take the form of a graphic logo that corresponds to other additional textual information such as an advertising flame printed by the sender.
- FIG. 2 very schematically illustrates a postal sorting machine 1 which conventionally comprises a feed inlet 2 with a magazine for loading mailpieces P and a unstacker for serially putting on the edge of shipments a digital camera 3 to form an image of the surface of each shipment comprising the various information blocks recalled above, and a bucket carousel 4 which directs the shipments to sorting outlets 5.
- Each sorting outlet can be equipped with several bins (typically a front and a rear tray) sorting outlet or compartmentalized tray without departing from the scope of the invention.
- a referenced data processing system 6 for the OCR postal address recognition which is associated with a video coding system 7 as is well known.
- the system 6 is also arranged to derive from the digital image of a transmission formed by the camera 3 a digital signature.
- system 9 similar to the system 6 but which is part of another sorting machine 10, this system 9 being connected in communication for example by a telecommunication network 8 to system 6 for an application of the method according to the invention to a sorting process in several sorting passes on several sorting machines.
- FIG. 3 schematically illustrates the construction of the digital signatures during a first sorting pass in the machine 1.
- postal items P are injected into the sorting machine 1 through the feed inlet. 2.
- the postal items P are unstacked and conveyed serially on edge to the digital camera 3.
- a digital image is formed of the surface of the current mail item that can comprise the various information blocks illustrated in FIG. 1, for example AE, AD and L.
- step 13 the system 6 undertakes an automatic OCR distribution address recognition from the image and at the same time derives from this image a digital signature V-Id assigned to the transmission.
- step 14 if the postal address has been uniquely recognized in step 13, the address data resulting from OCR recognition in correspondence with the signature V is stored in machine memory in step 16. Id of the sending.
- step 15 the image of the sending is transmitted to the video coding system 7 for an extraction of the address information by a video coding operator, whereupon the address information obtained by video coding in 15 are recorded at 16 in correspondence with the signature obtained at step 13.
- the block indicated by 17 represents a database called BDref in which the digital signature V-Id and sorting data (including the address information) of the sending are recorded in correspondence for each sending. recognized by OCR or video-coding during the first sort pass.
- Step 18 corresponds to the process of conveying the shipment from the camera 3 to the sorting outlets of the machine.
- FIG. 4 again shows the digital image of a postal item P, generally a digital image with several gray levels, with the information blocks AD, AE and L.
- a first component of the signature according to the invention is called “image component” Ci which is representative of the physical characteristics of the image.
- This image component is for example extracted by statistical analysis of the luminance of the points of the digital image which will have previously undergone a succession of filtering lowering the resolution level of the image in order to reduce the processing times for the statistical analysis and to have a low-frequency type of content which is insensitive to luminance fluctuations during multiple acquisitions.
- the luminance of a point in the image corresponds to the gray level value of the image point.
- the system 6 extracts by calculation, global attributes representative of global physical characteristics of the image such as the height and the width of the mailpiece, the average luminance value of the images. points of the digital image, standard deviation, entropy of luminance values.
- FIG. 4 shows the digital image of the transmission P subdivided into several distinct parts B11, B'45 coming from different grids or meshes M1, M2, M3, M4, M5 in the digital image.
- Grid M1 here defines 3x3 distinct parts.
- the M5 grid defines 8x10 distinct parts.
- the number of distinct parts in a grid and the number of grids may be a parameter of the statistical analysis applied to the digital image to extract the image component Ci from the signature.
- each part such as B11 and B'45 of the digital image resulting from a grid such as M1 or M3 one can extract local attributes such as the average luminance value of the points of this part of the digital image, standard deviation, entropy of the luminance values in this part of the digital image.
- These local attributes contain information all the more discriminating that postal items are heterogeneous.
- the set of global and local attributes extracted for a digital image therefore constitute the image component Ci of the signature.
- FIG. 5 there is illustrated a second component called "postal component" Cp of the signature which is indicative at least of the spatial position of the information blocks such as AD, AE and L in an image of a shipment.
- An OCR optical character recognition system conventionally used in a postal sorting machine is capable of providing data indicative of the spatial position of the textual information blocks detected in the digital image. These indicative spatial position data may be the spatial and orientation coordinates of the rectangular area forming each information block.
- An OCR system is also able to provide a textual description of each block of information detected in the digital image.
- a textual description of a block of information such as AD may consist in indicating the number of lines of characters detected in the information block, the number of words detected in each line of characters, the number of characters detected in each word of each line of characters.
- FIG. 5 illustrates an example of a textual description of the information block AD constituting the postal component Cp of the signature of a postal item:
- BLOCK # 0/3 denoted by 33, references the information block O among the 3 blocks of information detected in the digital image;
- HN denoted by 33 ', is a datum indicative of the orientation of the information block 0 in the digital image
- NbLines 4" denoted by 33 ' indicates that the information block 0 has 4 lines of characters;
- Line # 0 denoted by 34, references the first line of characters detected in information block 0;
- NbMots 03 denoted by 35, is a datum indicating that 3 words have been detected in the first line of characters;
- NbCarParMot 01 06 04 denoted by 36, are data indicating that the 3 words of the first line of characters comprise respectively 1, 6 and 4 characters;
- Car # 0 (1,007 I 009 i 019) denoted by 37, are data indicating that for the first character of the first line of characters, the OCR has identified 3 candidate characters respectively 1, I and i with distances resemblance respectively 007, 009 and 019;
- Car # 1 (L 008 E 009 D 057)" denoted by 38, are data indicating that for the second character of the first line of characters, the OCR has identified 3 candidate characters respectively L, E and D with distances resemblance respectively 008, 009 and 057;
- a digital image of a postal item is an interpretable two-dimensional signal whose content can be apprehended both physically and symbolically.
- the signature of a postal item is formed of two complementary components Ci and Cp which are not correlated with one another (independent of one another).
- FIG. 6 shows the signature management process during a second sorting pass following the first sorting pass illustrated in FIG. 3.
- the sorted items in the first pass are therefore entered again in the machine of FIG. sort 1 and pass serially on edge in front of the camera 3.
- the current signature V-IdC includes an image component CiC and a postal component CpC.
- step 46 comparing the image and postal components CiC and CpC of the current signature V-IdC with the image and postal components Ci, Cp of the signatures recorded in the BDRef database 17 to detect a match.
- This comparison can begin with the comparison of the respective global attributes of the image components which includes a thresholding of the absolute values of the variations on each global attribute to make a first filtering among the signatures recorded in the base 17.
- This filtering makes it possible to eliminate the signatures strongly. dissociating from the current signature and retaining only a small number of candidate signatures for further comparison.
- a comparison of the local attributes of the image component is performed on these candidate signatures, which can further reduce the number of candidate signatures in the base 17.
- This comparison can be based firstly on the calculation of a normalized sliding correlation coefficient. between the corresponding histograms in the current signature and the candidate signatures respectively, which makes it possible to overcome the luminance variations between the two compared digital images and then on the calculation of a normalized correlation coefficient, by attribute type, between the other local attributes in the current signature and in the candidate signatures respectively which makes it possible to overcome the problems of normalization due to the difference of variability of each local attribute.
- the candidate signatures are then sorted in descending order of similarity based on the correlation coefficients and a fixed number of most similar candidate signatures are retained.
- candidate signatures can then be sorted in descending order from a measure of similarity of the textual descriptions of the information blocks.
- step 46 If in step 46, it is not possible to detect a match, then in step 47, the current send is sent to a reject output so that it can be sorted manually, for example. If in step 46, a match is detected, the sort and address data for the current mail is retrieved from the base 17 and the current mail is automatically directed to a corresponding sorting output in step 48. .
- a reliability process is carried out on the one hand of the current signature and on the other hand of the signatures of the BDRef base using a classification of these signatures. This reliability makes it possible, during the concordance search, to adjust the postal component of the signatures to be compared so as to avoid mismatches.
- the principle of this classification is to first create a kind of dictionary in which are listed categories or "graphic classes" CG signatures.
- Each graphical class is modeled by a CiCG image component and a CpCG postal component similar to that of a signature as described above.
- this dictionary is scalable in real time in the sense that the image and postal components of the graphic classes are updated on the fly, that is to say that as and when we assign a signature to a graphical class and this graphical class is therefore consolidated.
- This update on the fly does not require treatment deferred, it is executed in real time.
- it is not necessary to transmit the dictionary for subsequent sorting in a remote sorting center.
- an average value of the image component (consolidation phase) is calculated at each updating of a graphic category and the low frequency activity of the image component is analyzed. This average value for evaluating the position of the destination address block, which makes it possible, during the concordance search at step 46, to use the appropriate information block when comparing the postal component of the signatures.
- Low frequency analysis is an analysis of changes among low resolution images.
- step 43 a classification process of the current signature VidC is thus performed, that is to say that the image component CiC of the current signature is compared with the image component CiCG of the graphic classes present in the dictionary 19 to determine the CGx graphical class of the dictionary that is most similar to the current signature. More particularly according to the method of the invention, the local and global attributes of the image components CiC and CiCG are compared to identify a graphical class of CGx membership at the output of step 43. If no graphical class of the dictionary corresponds to the current signature, there is creation of a new CGx graphical class.
- FIGS. 7 and 8 illustrate more particularly the process of comparing the image components to achieve this classification. These are discussed below.
- step 44 an analysis of the low frequency activity of the image component of the CGx graphic class is performed to detect a significant low frequency activity of an information block in the digital images. associated with the signatures in this graphical class.
- This step 44 is described more particularly with reference to FIG. 9.
- step 45 for the current signature, a reliability process is carried out on the basis of the results of the activity analysis in step 44 which is more detailed with reference to FIGS. 10 and 11.
- the postal components of the current signature and the graphic class CGx can be adjusted. These adjustments make it possible to avoid mismatches in step 46 of matching with the signatures recorded in the base 17.
- step 46 of concordance search one carries out in 50 a consolidation of the postal component of the graphical class CGx identified for the current sending that if in the preceding step 49 one is in a situation or a certain "postal criterion" loud "is checked.
- CpC of the current signature and a matched signature identified in step 46 are similar (for example: same identification of the destination address block, same number of lines of characters in that block, same number of words per line in this block, etc.).
- steps 41 to 50 in FIG. 6 is thus repeated for the successive current items processed in the second sorting pass.
- step 43 of the classification of signatures has thus been illustrated in greater detail.
- the dictionary or database 19 CG graphical classes is normally empty.
- NbMaxCG may for example be conditioned by a "priori" information given by the machine operator indicative of the heterogeneity of the postal items to be processed. In general for a two-pass sorting of mail from several shippers in number, we can set the NbMaxCG value to about 100.
- a value NbMaxCG equal to 100 corresponds to a batch of mail items weakly homogeneous, that is to say Say with a number of successive shipments belonging to the same sender which is relatively low.
- the current image component is compared to step 51.
- the comparison of the image components is carried out in the same manner as previously described for step 46. In particular, the comparison is first performed on the global attributes and then on the local attributes of the image components CiC and CiCG.
- the global and local attributes are obtained from a low resolution digital image of the transmission, for example 0.25 pixels per mm.
- FIG. 8 illustrates more particularly this step 51 for comparing the global (51a) and local (51b) attributes of the image components CiC and CiCG.
- global attributes we have illustrated: the height H of the postal item, the width L, average luminance value I of the points of the digital image, the standard deviation E and the entropy S of the luminance values.
- frame 52 On the left in FIG. 8, there is illustrated in frame 52 a first vector (one-column table) the values H, L, I, E, T for CiC and by some vectors the similar values for CiCGs associated with classes CG1 , CG2, ...., CGN
- each global attribute of the image component CiC is first compared with the corresponding attribute of one of the image components CG1, CG2, ...., CGN.
- the difference between the heights H of CiC and of CG1 is compared with a threshold ⁇ 1
- the difference between the widths L of CiC and of CGI is compared with a threshold ⁇ 2.
- the thresholds ⁇ 1, x2 ... can be different, and if for all the attributes, the difference is lower than the threshold then CG1 is candidate.
- a distance calculation is then carried out (standard vector distance designated in FIG. 8 by DVN) between the HLIET vector for CiC and the HLIET vector of each candidate graphical class.
- These normalized vector distances DVN are also compared to a threshold (which can be adjustable) in step 53 to determine the graphical class (es) CGi that are most similar to the current signature.
- a threshold which can be adjustable
- the local attributes of the image component CiCG of these graphic classes are compared with the local attributes of the image component CiC of the current signature.
- the local attributes in the frame 54 are illustrated by triplets of 4x4 matrices (thus resulting from a 4x4 grid for the construction of the signature) assigned respectively to the average luminance value I of points image, the standard deviation E and the entropy S of the luminance values.
- the local CiC attributes are correlated with the CiCG local attributes of each graphical class selected at 53 and the highest calculated DC correlation coefficient (-1 ⁇ CC ⁇ 1) is retained at the end of the step 51.
- the highest calculated DC correlation coefficient (-1 ⁇ CC ⁇ 1) is retained at the end of the step 51.
- a correlation coefficient CC is compared with a threshold (adjustable if necessary) to undertake an updating of the dictionary 19 of the graphic classes.
- a threshold adjustable if necessary
- a new graphic class CGx is added in step 56 in dictionary 19 and the number of graphic classes maintained in the dictionary is updated (if applicable deleting a graphical class if the NbMaxCG value is exceeded, the graphical class deleted from the dictionary being the one that is least used during a reference period, for example).
- the CiCG and CpCG components of the new CGx graphical class are initialized with the CiC and CpC components of the current signature used in step 46.
- step 55 the correlation coefficient CC is greater than or equal to the threshold, the utilization counter of the graphic class CGx is updated to 57 and this graphic class CGx is updated.
- step 55 makes it possible to adjust the level of similarity required for categorization. In practice, it is necessary to find a threshold which makes it possible to gather the signatures of important graphic resemblance by avoiding that the sending of the same sender with the same graphic appearance can lead to the creation of several graphic classes. At the end of step 43, a graphic class CGx is therefore identified as similar to the current signature VidC.
- FIG. 9 illustrates a set of local attributes of the image component of the graphic class CGx which are used for the analysis of low frequency activity in step 44 in FIG. 6.
- These local attributes are values standard deviation on: average luminance I, standard deviation E and entropy T. These values relate to distinct zones of a certain grid of the image of a transmission as explained above. As an example, a grid of the 5 ⁇ 5 image is shown in the figure, which gives 3 matrices 60 with each 25 standard deviation values. To calculate each standard deviation value in a matrix I55, E55 and T55, a consolidated average is made with the corresponding value of the local attribute of the current signature. It can therefore be considered that by calculating this standard deviation an average signature value is calculated for the graphical class CGx.
- the 3 matrices I55, E55 and T55 can be fused into a matrix G55, for example according to a standardized principal component analysis (PCA) well known to those skilled in the art.
- This matrix G is a kind of "grid of changes" that exhibits the low-frequency changes, that is to say the changes that occur at each consolidation of the matrix G.
- the "K-means” method makes it possible to isolate matrix elements in the matrix G according to 3 levels: Non-changing level - Little changing level - Very changing level, respectively illustrated by squares in white, gray and black.
- Non-changing level Little changing level
- Very changing level respectively illustrated by squares in white, gray and black.
- the spatial position of an information block having the most significant low frequency activity has normally been detected.
- FIG. 10 illustrates the taking into account of the result of the low frequency activity analysis in step 45.
- a low frequency activity of a matrix element of G or matrix elements of several matrices G
- it is checked at 64 if the position of this square in the matrix G coincides with the position of the information block in the postal component CpC the current signature that has been retained as the recipient address block.
- the coincidence can be established by a spatial projection of the coordinates of the square and the information block to measure a spatial distance and then by a comparison of this distance measured at a threshold.
- the process proceeds to step 70 in FIG.
- step 70 is continued in FIG.
- step 64 If at step 64, no spatial coincidence is detected between the information block identified in the postal component CpC of the current signature and the "block" of the matrix exhibiting the most significant low frequency activity, then at step 65, a spatial coincidence measurement is made between this "block" of higher low frequency activity and all the other information blocks identified in the postal component CpC of the current signature. If no spatial coincidence is detected with these other information blocks, the process is continued in step 70 of FIG.
- step 65 If at step 65, a spatial coincidence is detected with one of these other information blocks, the postal component CpC of the current signature is modified in step 66 to identify this information block as probably the address block address, this block will be used in step 46 for the search for matching signatures.
- step 70 a consolidated measurement is made on the dispersion of the spatial position of the information block identified as the address address block in the postal component of all the successive signatures forming the graphic class CGx. This dispersion measure is in the form of a standard deviation value. If it is detected that the distance measured is greater than a certain threshold, the process continues in step 46 of FIG. 6. We are in a situation where the postal component CpC of the graphic class CGx is not reliable enough fact that the postal component of the signatures used to build this graphical class is unstable.
- step 70 the process continues in step 71 where the spatial concordance between the postal component of the CGx graphic class and the postal component CpC of the current signature is measured.
- This measure consists, for example, in calculating the positional difference between the center of the information block identified as the destination address block in the postal component of the current signature and the average position of the center of the address address block of the class. corresponding graph. This measurement is compared with a low threshold and if the measurement is below this certain low threshold, the process is continued in step 46 of FIG. 6 for a signature matching search.
- step 71 If at step 71, a spatial match is not detected, then the process continues at step 72 where it is searched whether there is an information block identified in the postal component CpC of the current signature for which the distance with the information block identified as the destination address block of the postal component of the CGx class is below the low threshold.
- step 72 If yes in step 72, proceed to step 73 where this information block is now identified in the CpCG postal component of the CGx graphics class as likely to be the recipient address block.
- step 74 the process is continued in step 74 where a measurement of the spatial distance between all the blocks of information identified in the postal component CpC of the current signature and the block is carried out. of information identified as a destination address block in the postal component of the CGx graphics class. If this distance is greater than a high threshold, the postal component of the CGx class is re-initialized (RAZ) in step 75 in the dictionary 19.
- RZ re-initialized
- FIG. 12 shows another embodiment of the method according to the invention which differs from that of FIG. 6 in step 42.
- step 80 which follows step 42 of FIG. generating a current signature VidC
- the current signature is classified using a dictionary 19 as explained above.
- the postal component CpC of the current signature is modified from the image component CiCG of the graphic class CGx and, if appropriate, the postal component CpCG of the graphic class CGx is modified.
- step 81 signatures from the signature database 17 are recovered from the candidate signatures according to a mechanism for predicting sequences known from the French patent FR-2883943.
- Step 81 serves to limit the exploration space in the base 17 by taking advantage of the fact of taking into account a certain sequencing (order of passage) of the items in the first sorting pass which is repeated in the second sorting pass. .
- This limited exploration space consists of candidate signatures.
- step 82 a classification of each candidate signature in the graphics class dictionary 19 is performed according to the classification mechanism described above.
- a graphical class CGy has been identified for each candidate signature.
- step 83 If in step 83, it is detected that the graphical class CGx of the current signature is identical to the graphical class CGy of the candidate signature considered, then at step 84 the postal component Cp of the candidate signature is modified according to the mechanism described in connection with FIGS. 10 and 11 to correctly reposition the postal address block recipient in the postal component of the candidate signature. Then, in step 85, the postal component of the candidate signature is modified, as the case may be, according to the postal component of the graphic class CGx as described with reference to FIG. 11. The method then proceeds to step 46 of FIG. FIG. 6. It should be understood that the process is repeated between steps 82 to 85 for each candidate signature at the output of step 81.
- step 83 if in step 83 a common graphical class has not been detected between the current signature and the candidate signatures, the process proceeds directly to step 46 of FIG.
- the process of limiting the exploration space of step 81 starts from the idea of assigning in first sorting pass to each send a chronological order number NO which is stored in memory in correspondence with the signature of 17.
- Each chronological order number NO can be constituted for example by the juxtaposition of a sorting center number assigned to the sorting center in which the sorting machine 1 is located, a number sorting machine assigned to the sorting machine in which the mailing is sorted, a sorting outlet tray number assigned to the sorting output bin in which the mailing is directed and a chronological index assigned to sending.
- this index is for example the value of a counter associated with a sorting output bin and which is initialized when a first shipment is sent to the bin and which is incremented by one unit each time that a new shipment is sent to the ferry. In this way, each shipment is assigned a unique order number NO.
- the signatures are grouped in sequence in the database 17. For example, the signatures are grouped and ordered sequentially by sorting center, by machine and by tray according to the order in which shipments are stored in each sort output bin. As detailed below, the contiguous signature sequences in database 17 constitute identifiable segments.
- the V-IdC signature of the current item is assigned an IP crossing index.
- the mode of implementation of the method according to the invention illustrated in FIG. 12 advantageously exploits the prediction mechanism described above during the consolidation step 50.
- the postal component of the graphical class CGx identified for the current mail is consolidated if in step 49 one is in a situation of "strong postal criterion" as explained above and if in addition the chronological order number of the matching signature determined in step 46 corresponds at the estimated NO serial number.
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Sorting Of Articles (AREA)
- Packaging For Recording Disks (AREA)
- Pharmaceuticals Containing Other Organic And Inorganic Compounds (AREA)
- Medicines Containing Antibodies Or Antigens For Use As Internal Diagnostic Agents (AREA)
- Analysing Materials By The Use Of Radiation (AREA)
- Steroid Compounds (AREA)
- Character Discrimination (AREA)
- Image Analysis (AREA)
Abstract
Description
Claims
Priority Applications (8)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DK08805830.0T DK2155406T3 (da) | 2007-05-30 | 2008-05-22 | Fremgangsmåde til behandling af forsendelser inklusive en grafisk klassificering af signaturerne, der er associerede med forsendelserne |
AT08805830T ATE475492T1 (de) | 2007-05-30 | 2008-05-22 | Verfahren zur handhabung von übertragungen einschliesslich einer grafischen klassifizierung der zu den übertragungen gehörenden signaturen |
US12/282,438 US8126204B2 (en) | 2007-05-30 | 2008-05-22 | Method of processing mailpieces, the method including graphically classifying signatures associated with the mailpieces |
DE602008001985T DE602008001985D1 (de) | 2007-05-30 | 2008-05-22 | Verfahren zur handhabung von übertragungen einschliesslich einer grafischen klassifizierung der zu den übertragungen gehörenden signaturen |
JP2010509870A JP4785986B2 (ja) | 2007-05-30 | 2008-05-22 | 搬送処理に関連したシグネチャのグラフィック分類を含む搬送処理の取扱方法 |
CN2008800181090A CN101678404B (zh) | 2007-05-30 | 2008-05-22 | 包括对与邮寄件相关的签名进行图形分类的邮寄件处理方法 |
CA2688163A CA2688163C (fr) | 2007-05-30 | 2008-05-22 | Procede de traitement d'envois incluant une classification graphique des signatures associees aux envois |
EP08805830A EP2155406B1 (fr) | 2007-05-30 | 2008-05-22 | Procédé de traitement d'envois incluant une classification graphique des signatures associées aux envois |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
FR0755342A FR2916661B1 (fr) | 2007-05-30 | 2007-05-30 | Procede de traitement d'envois incluant une classification graphique des signatures associees aux envois |
FR0755342 | 2007-05-30 |
Publications (2)
Publication Number | Publication Date |
---|---|
WO2008152277A2 true WO2008152277A2 (fr) | 2008-12-18 |
WO2008152277A3 WO2008152277A3 (fr) | 2009-02-12 |
Family
ID=38951453
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/FR2008/050885 WO2008152277A2 (fr) | 2007-05-30 | 2008-05-22 | Procédé de traitement d'envois incluant une classification graphique des signatures associées aux envois |
Country Status (12)
Country | Link |
---|---|
US (1) | US8126204B2 (fr) |
EP (1) | EP2155406B1 (fr) |
JP (1) | JP4785986B2 (fr) |
CN (1) | CN101678404B (fr) |
AT (1) | ATE475492T1 (fr) |
CA (1) | CA2688163C (fr) |
DE (1) | DE602008001985D1 (fr) |
DK (1) | DK2155406T3 (fr) |
ES (1) | ES2348009T3 (fr) |
FR (1) | FR2916661B1 (fr) |
PT (1) | PT2155406E (fr) |
WO (1) | WO2008152277A2 (fr) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102010013220A1 (de) * | 2010-03-29 | 2011-09-29 | Siemens Aktiengesellschaft | Verfahren und Vorrichtung zum Transportieren eines zu bedruckenden Gegenstands |
DE102012200580A1 (de) | 2011-01-26 | 2012-07-26 | Siemens Aktiengesellschaft | Verfahren und Anordnung zum Identifizieren eines transportierten Gegenstands mit einem Etikett |
US9144827B2 (en) | 2010-09-23 | 2015-09-29 | Solystic | Postal sorting method with user processing information being recovered between sorting passes |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FR2918199B1 (fr) * | 2007-06-26 | 2009-08-21 | Solystic Sas | Procede de traitement d'envois postaux exploitant l'identification virtuelle des envois avec le readressage |
FR2920678B1 (fr) * | 2007-09-07 | 2009-10-16 | Solystic Sas | Procede de traitement d'envois postaux avec des codes client associes a des empreintes numeriques. |
AU2011200830B2 (en) * | 2011-02-25 | 2014-09-25 | Canon Kabushiki Kaisha | Method, apparatus and system for modifying quality of an image |
US8818023B2 (en) * | 2011-03-25 | 2014-08-26 | Siemens Industry, Inc. | Bulk region of interest learning |
US9147178B2 (en) * | 2012-01-16 | 2015-09-29 | International Business Machines Corporation | De-identification in visual media data |
EP2923775B1 (fr) * | 2014-03-26 | 2020-07-08 | NEC Corporation | Système de Tri Postal |
US10031925B2 (en) * | 2014-10-15 | 2018-07-24 | Thinkcx Technologies, Inc. | Method and system of using image recognition and geolocation signal analysis in the construction of a social media user identity graph |
KR101742637B1 (ko) * | 2015-05-15 | 2017-06-01 | 주식회사 세중에스앤씨 | 반송 우편물 자동처리 시스템 및 방법 |
FR3037691B1 (fr) * | 2015-06-16 | 2018-04-20 | Solystic | Appariement d'images d'articles postaux avec des descripteurs de singularites du champ de gradient |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE19947259C1 (de) * | 1999-09-30 | 2000-09-28 | Siemens Ag | Verfahren und Vorrichtung zum Sortieren von Sendungen |
DE20218212U1 (de) * | 2002-08-12 | 2003-12-24 | Mit Management Intelligenter Technologien Gmbh | Vorrichtung zur automatischen Adressblockmarkierung |
FR2883493A1 (fr) * | 2005-03-24 | 2006-09-29 | Solystic Sas | Procede de traitement d'envois incluant une gestion des empreintes numeriques des envois |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6360001B1 (en) * | 2000-05-10 | 2002-03-19 | International Business Machines Corporation | Automatic location of address information on parcels sent by mass mailers |
FR2866252B1 (fr) * | 2004-02-18 | 2006-04-21 | Solystic | Procede pour trier des envois postaux en plusieurs passes de tri |
ITTO20040299A1 (it) * | 2004-05-07 | 2004-08-07 | Elsag Spa | Metodo di riconoscimento postale |
-
2007
- 2007-05-30 FR FR0755342A patent/FR2916661B1/fr not_active Expired - Fee Related
-
2008
- 2008-05-22 ES ES08805830T patent/ES2348009T3/es active Active
- 2008-05-22 WO PCT/FR2008/050885 patent/WO2008152277A2/fr active Application Filing
- 2008-05-22 PT PT08805830T patent/PT2155406E/pt unknown
- 2008-05-22 DE DE602008001985T patent/DE602008001985D1/de active Active
- 2008-05-22 AT AT08805830T patent/ATE475492T1/de active
- 2008-05-22 CA CA2688163A patent/CA2688163C/fr not_active Expired - Fee Related
- 2008-05-22 JP JP2010509870A patent/JP4785986B2/ja not_active Expired - Fee Related
- 2008-05-22 CN CN2008800181090A patent/CN101678404B/zh not_active Expired - Fee Related
- 2008-05-22 DK DK08805830.0T patent/DK2155406T3/da active
- 2008-05-22 US US12/282,438 patent/US8126204B2/en not_active Expired - Fee Related
- 2008-05-22 EP EP08805830A patent/EP2155406B1/fr not_active Not-in-force
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE19947259C1 (de) * | 1999-09-30 | 2000-09-28 | Siemens Ag | Verfahren und Vorrichtung zum Sortieren von Sendungen |
DE20218212U1 (de) * | 2002-08-12 | 2003-12-24 | Mit Management Intelligenter Technologien Gmbh | Vorrichtung zur automatischen Adressblockmarkierung |
FR2883493A1 (fr) * | 2005-03-24 | 2006-09-29 | Solystic Sas | Procede de traitement d'envois incluant une gestion des empreintes numeriques des envois |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102010013220A1 (de) * | 2010-03-29 | 2011-09-29 | Siemens Aktiengesellschaft | Verfahren und Vorrichtung zum Transportieren eines zu bedruckenden Gegenstands |
EP2371461A1 (fr) | 2010-03-29 | 2011-10-05 | Siemens Aktiengesellschaft | Procédé et dispositif de transport d'un objet devant être imprimé |
US8467569B2 (en) | 2010-03-29 | 2013-06-18 | Siemens Aktiengesellschaft | Method and apparatus for transporting an article to be printed |
US9144827B2 (en) | 2010-09-23 | 2015-09-29 | Solystic | Postal sorting method with user processing information being recovered between sorting passes |
DE102012200580A1 (de) | 2011-01-26 | 2012-07-26 | Siemens Aktiengesellschaft | Verfahren und Anordnung zum Identifizieren eines transportierten Gegenstands mit einem Etikett |
Also Published As
Publication number | Publication date |
---|---|
CA2688163C (fr) | 2014-10-07 |
WO2008152277A3 (fr) | 2009-02-12 |
CN101678404B (zh) | 2013-10-30 |
DK2155406T3 (da) | 2010-11-22 |
CN101678404A (zh) | 2010-03-24 |
US20100232642A1 (en) | 2010-09-16 |
DE602008001985D1 (de) | 2010-09-09 |
EP2155406A2 (fr) | 2010-02-24 |
FR2916661B1 (fr) | 2009-07-03 |
EP2155406B1 (fr) | 2010-07-28 |
PT2155406E (pt) | 2010-10-14 |
JP2010528386A (ja) | 2010-08-19 |
US8126204B2 (en) | 2012-02-28 |
ES2348009T3 (es) | 2010-11-26 |
FR2916661A1 (fr) | 2008-12-05 |
JP4785986B2 (ja) | 2011-10-05 |
ATE475492T1 (de) | 2010-08-15 |
CA2688163A1 (fr) | 2008-12-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP2155406B1 (fr) | Procédé de traitement d'envois incluant une classification graphique des signatures associées aux envois | |
EP1718423B1 (fr) | Procede pour trier des envois postaux en plusieurs passes de tri | |
EP2190594B1 (fr) | Procédé de traitement d'envois postaux avec des codes client associés à des empreintes numériques | |
EP1519796B2 (fr) | Chronomarquage d'objets postaux par signature d'image et machine de tri postal associee | |
EP1866105B1 (fr) | Procédé de traitement d'envois incluant une gestion des empreintes numériques des envois | |
EP1850973B1 (fr) | Procede de traitement d'envois postaux avec une detection des occurrences d'attributs ocr | |
EP2791866B1 (fr) | Procédé de traitement d'envois postaux avec génération de modèles numériques de filantes sur borne interactive | |
EP2822709B1 (fr) | Procédé de suivi d'envois postaux pour mesurer des temps de parcours dans un système de tri postal multi réseaux | |
EP3903230B1 (fr) | Appariement structurel d'images par hachage de descripteurs de singularités du gradient | |
EP3205412B1 (fr) | Methode pour trier des articles de courrier pre-tries |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 200880018109.0 Country of ref document: CN |
|
WWE | Wipo information: entry into national phase |
Ref document number: 12282438 Country of ref document: US |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 08805830 Country of ref document: EP Kind code of ref document: A2 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2008805830 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2688163 Country of ref document: CA |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2010509870 Country of ref document: JP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |