CN116486423A - Financial ticketing data processing method based on image recognition - Google Patents
Financial ticketing data processing method based on image recognition Download PDFInfo
- Publication number
- CN116486423A CN116486423A CN202310446267.2A CN202310446267A CN116486423A CN 116486423 A CN116486423 A CN 116486423A CN 202310446267 A CN202310446267 A CN 202310446267A CN 116486423 A CN116486423 A CN 116486423A
- Authority
- CN
- China
- Prior art keywords
- character
- scanning
- current
- invoice
- line
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 25
- 238000000034 method Methods 0.000 claims description 34
- 238000012545 processing Methods 0.000 claims description 28
- 238000012512 characterization method Methods 0.000 claims description 23
- 238000003708 edge detection Methods 0.000 claims description 10
- 230000002159 abnormal effect Effects 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 8
- 238000002372 labelling Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 4
- 238000007792 addition Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- 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
- G06V30/41—Analysis of document content
- G06V30/412—Layout analysis of documents structured with printed lines or input boxes, e.g. business forms or tables
-
- 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
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/146—Aligning or centring of the image pick-up or image-field
- G06V30/1475—Inclination or skew detection or correction of characters or of image to be recognised
-
- 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
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/15—Cutting or merging image elements, e.g. region growing, watershed or clustering-based techniques
-
- 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
- G06V30/10—Character recognition
- G06V30/19—Recognition using electronic means
- G06V30/19007—Matching; Proximity measures
- G06V30/19093—Proximity measures, i.e. similarity or distance measures
-
- 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
- G06V30/41—Analysis of document content
- G06V30/418—Document matching, e.g. of document images
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Character Input (AREA)
Abstract
The invention discloses a financial ticket processing method based on image recognition, which mainly uses invoice scanning equipment to scan, recognize and input a large amount of invoice information, is convenient for staff to search, arrange and manage the invoices, and adjusts scanned invoice image information when the invoice scanning and recognition are wrong, so that the scanned invoice image information is scanned, recognized and input after being aligned to a scanning window.
Description
Technical Field
The invention belongs to the technical field of invoice management safety, and relates to a financial ticketing processing method based on image recognition.
Background
Along with the development of economy, an invoice is an indispensable financial bill, so in the prior art, in order to collect paper invoices in storage, an intelligent processing mode of large-batch scanning identification is often adopted (the aim of intelligent archiving processing of the paper invoices for large-batch scanning identification is mainly two, namely, the aim of conveniently storing real scanned bill evidence data is two, and the aim of conveniently calling and retrieving related bill data at any time is two);
during the scanning and identification of a large number of financial notes, the post-pushing scanning of the invoices can be performed one by one at intervals (namely one by one) through matched equipment. However, research finds that when a large number of invoices are scanned, the phenomenon that a plurality of invoices are overlapped due to the fact that a plurality of invoices are pushed at one time is likely to occur, so that the scanned file cannot be normally acquired due to the fact that the plurality of invoices are overlapped, and the extraction and recognition influence is caused.
Disclosure of Invention
The invention provides a financial ticketing processing method based on image recognition and a readable storage medium, which solve the technical problems in the prior art.
The invention provides a financial ticketing processing method based on image recognition, which comprises the following operation steps:
obtaining text information of a plurality of sellers and text information of buyers, and forming the text information of the plurality of sellers into a seller text set; forming text information of a plurality of buyers into a buyer text set;
dividing the basic template invoice into areas according to functions of the basic template invoice to obtain divided subareas, wherein the functions comprise a header, a two-dimensional code, a password area and a table area, the divided subareas are numbered according to a preset sequence and are sequentially marked as 1,2,..k,..v, further obtaining positions corresponding to the subareas, and constructing a position set W (W1, W2,..Wk,..Wv) of each subarea, wherein Wk represents the position of the kth subarea;
when the scanning processing operation is carried out on the layout of the current whole invoice, scanning information after scanning is obtained: the scanning of the whole invoice layout comprises the steps of scanning a header, a two-dimensional code, a password area and a table area of the whole invoice, further obtaining area representation information corresponding to each subarea, and further constructing an information set H (H1, H2, hk, hv) of each subarea, wherein Hk represents the area representation information corresponding to the kth subarea;
And executing scanning information screening processing operation: acquiring and calling subregions of the current whole invoice in a form region of the buyer and the seller, acquiring region characterization information corresponding to the subregions of the buyer and the seller, comparing the region characterization information corresponding to the subregions of the buyer with a buyer text set, judging that the subregions are abnormal subregions if no match exists, judging that the subregions are normal subregions if the match exists, and filtering the normal subregions; comparing the regional characterization information corresponding to the subregion of the seller with the seller text set, judging that the subregion is regarded as an abnormal subregion if no match exists, and regarding as a normal subregion if the match exists, and further filtering the normal subregion; if the subarea is abnormal subarea, the subarea is marked as a marked area, then the image information in the current whole invoice scanning information corresponding to the marked area is processed, and the whole invoice information is scanned and recorded after confirming that the image information is aligned to the scanning window.
Specifically, image information in the current whole invoice scanning information corresponding to the labeling area is processed, and the method specifically comprises the following steps:
Dividing the scanned image information of the current overlapped invoice to obtain an uppermost invoice image as a target image; taking a transverse straight line where the current scanning window is positioned as a preset horizontal line and taking the preset horizontal line as a comparison line of the long-side edge line of the table;
obtaining a long-side edge line and a short-side edge line of a table according to the table edge of a table area in a scanning identification target image, and preliminarily judging whether the first initial condition of the alignment position is met or not according to the comparison line of the long-side edge line and the long-side edge line of the table; the first initial condition is that the current long-side edge line of the table and the contrast line of the long-side edge line of the table are in parallel relation;
if the scanning identification form long-side edge line is not parallel to the comparison line of the form long-side edge line, the target image is adjusted, and the adjusted target image which accords with the second initial condition is scanned, identified and recorded, comprising the following steps:
determining character boundaries (character boundaries) within the fields of the form field by edge detection;
acquiring character boundaries of two continuous current characters in a form area, determining two continuous character intervals, judging whether the two continuous character intervals are equal to a preset character interval value, if so, regarding the two continuous characters as the same row, and further judging whether a plurality of discontinuous continuous two character semanteme in the same row has financial bill related semanteme; the two continuous character semanteme is regarded as a phrase, and a plurality of discontinuous two continuous character semantemes in the same row are regarded as a plurality of discontinuous phrases in the same row; (namely, two continuous character semanteme forms a phrase, the semantic judgment is more accurate by a plurality of discontinuous phrases in the same row, and the situation that the semantic judgment is inaccurate exists for the single two continuous character semantic judgment indeed, for example, chinese semanteme has financial bill association semanteme, business semanteme has financial bill association semanteme, chinese semanteme does not have financial bill association semanteme, so that the semantic judgment needs to be more accurate for a plurality of discontinuous phrases in the same row);
If a plurality of discontinuous word groups in the same row are judged to have financial bill association semantics, determining that any two characters in the same row with the financial bill association semantics are a first character and a second character in sequence, and further determining that a long-side edge line of a table in a target table area in a target image with parallel character extension vector directions from the first character to the second character is a first edge of the target table area in the target image;
determining a vertical line with a first edge of a target table area of the target image as a second edge of the table, enabling the first edge of the table and the second edge of the table to rotate clockwise (namely, rotating the target image acquired through whole scanning), and driving the target image to rotate clockwise to adjust the position of the target image;
simultaneously detecting whether the first edge of a target table area of the rotated target image is parallel to a preset horizontal line (namely a contrast line of a long-side edge line of the table) in real time, if so, regarding the table edge as being at an alignment position, and further performing scanning identification entry on the invoice serving as the uppermost layer of the target image after finishing adjustment;
the second initial condition is that the distance between any two current continuous characters in the same row is consistent with a preset character distance value, and a plurality of discontinuous phrases in the same row accord with the association semantics of the financial bill.
Specifically, judging that two continuous character semantics have financial bill association semantics specifically includes the following steps:
presetting a plurality of financial bill related vocabulary sets and combining a semantic library constructed by the plurality of financial bill related vocabulary sets;
firstly, acquiring current two continuous characters, matching the current two continuous characters with a current semantic library, and judging that the current two continuous characters have financial bill associated semantics if the semantics match is successful; and then acquiring a corresponding financial bill related vocabulary set successfully matched with the current two continuous characters.
If the two characters are the same, the current character is determined to be the first character to the second character, and the long-side edge line of the table with the parallel vector directions of the first character to the second character is determined to be the first edge of the table.
Specifically, the method for determining the character boundary in the table area by the edge detection method specifically comprises the following steps:
performing binarization processing on the target image to obtain a black-and-white image;
defining the first edge of the table area and the second edge of the table area as X axis and Y axis respectively, taking X, Y axis direction as basic direction, making the identification point start from X axis coordinate as 0, scanning downwards, scanning, traversing with width of 1px, if the RGB value of the current pixel point is 0, recording the X axis coordinate A at this time 1 Then continuing to scan downwards to obtain the next X-axis coordinate A 2 ；
Identifying the point from A 1 +1 starting, scanning longitudinally through each pixel point on the whole target image, if the RGB value of the current pixel point of the scanning is 255, recording the X-axis coordinate B at the moment 1 The method comprises the steps of carrying out a first treatment on the surface of the Then continuing to scan downwards to obtain the next X-axis coordinate B 2 ；
In the X interval (A) 1 ，(B-1) 1 ) In (3), the identification point is from the Y coordinateThe scanning is transversely carried out on 0 point, the R value of each point is judged, the scanning is stopped until the RGB value is equal to 0, and the Y-axis coordinate C at the moment is recorded 1 The method comprises the steps of carrying out a first treatment on the surface of the Then continuing to scan and traverse to the right to acquire the next Y-axis coordinate C 2 ；
In the X interval (A) 1 ，(B-1) 1 ) In (2), the identification point is from C 1 Starting transverse scanning at +1, judging R value of each pixel point, and stopping scanning to record Y-axis coordinate D at the moment if the R value is equal to 255 1 The method comprises the steps of carrying out a first treatment on the surface of the Then continuing to scan and traverse to the right to acquire the next Y-axis coordinate D 2 ；
Acquiring the abscissa and the ordinate of the left, right, upper and lower boundary points of the character, namely Z 1 (A 1 ，(B-1) 1 ,C 1 ,(D-1) 1 ) Is a character boundary.
Specifically, the continuous two-character distance calculating mode is a difference between a boundary coordinate of a continuous second character and a boundary coordinate of a first character, and the calculating formula is as follows:
Z 2 (A 2 ，(B-1) 2 ,C 2 ,(D-1) 2 )-Z 1 (A 1 ，(B-1) 1 ,C 1 ,(D-1) 1 )
wherein Z is 2 Character boundary coordinates for the second character; z is Z 1 Is the first character boundary coordinates.
The present invention provides a computer readable storage medium having stored thereon a computer program which when executed by a processor implements the steps of the image recognition based financial ticketing method.
Compared with the prior art, the invention has at least the following beneficial effects:
the invention provides a financial ticket processing method based on image recognition and a readable storage medium, wherein the financial ticket processing method based on image recognition recognizes text information and image information in a whole invoice through scanning the whole invoice and stores the text information and the image information into a database, so that staff can conveniently input invoice codes or other basic information so as to search relevant invoice scanning data, and ticket evidence data can be obtained. However, for an entire invoice scan, an important identification and specification scan is required if an invoice overlap is encountered, particularly in an important sub-region (e.g., a seller or purchaser sub-region of a form area). In this regard, the financial ticketing processing method based on image recognition provided by the embodiment of the invention screens scanned invoice images, eliminates compliance scanned invoice images, and automatically processes non-compliance invoice images, so that automatic scanning is more convenient. The invention processes the non-compliant invoice image, accurately acquires the invoice image information, and compares the invoice image information with Chinese information in the invoice image information, so that the processing is accurate.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of main steps of a financial ticketing processing method based on image recognition according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of an embodiment of processing image information in scan information of a current overlapped invoice when detecting that corresponding information in a sub-area is wrong in the financial ticketing processing method based on image recognition according to the embodiment of the invention;
FIG. 3 is a schematic diagram of an invoice scanning window in a financial ticketing processing method based on image recognition according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of overlapping and misaligned arrangement of multiple invoices in a scanning window in a financial ticketing processing method based on image recognition according to an embodiment of the present invention;
FIG. 5 is a flowchart illustrating an implementation of a step of processing a target image when a long edge line of a form is not parallel to a reference line of the long edge line of the form according to the financial ticketing processing method based on image recognition provided by the embodiment of the present invention;
FIG. 6 is a flowchart of an implementation of a step of determining that two consecutive character semantics have financial ticket association semantics in the financial ticket processing method based on image recognition according to the embodiment of the present invention;
FIG. 7 is a schematic diagram of a second flow chart of an embodiment of processing image information in scan information of a current overlapping invoice when detecting that corresponding information in a sub-area is wrong in the financial ticketing processing method based on image recognition according to the embodiment of the present invention;
FIG. 8 is a second flowchart of an implementation of steps for processing a target image when the edge line of the long side of the form is not parallel to the reference line of the edge line of the long side of the form in the financial ticketing processing method based on image recognition according to the embodiment of the present invention;
FIG. 9 is a second flowchart of an implementation of a step of determining that two consecutive character semantics have financial bill related semantics in the financial bill processing method based on image recognition according to the embodiment of the present invention;
FIG. 10 is a flowchart illustrating steps performed in determining character boundaries in a form area by an edge detection method according to an embodiment of the present invention;
fig. 11 is a schematic diagram of an in-region character boundary simulation of a table region determined by an edge detection method according to an embodiment of the present invention.
Detailed Description
The foregoing is merely illustrative of the principles of the invention, and various modifications, additions and substitutions for those skilled in the art will be apparent to those having ordinary skill in the art without departing from the principles of the invention or from the scope of the invention as defined in the accompanying claims.
Example 1
Referring to fig. 1, a first embodiment of the present invention provides a financial ticketing processing method based on image recognition, which includes the following steps:
step S1, in an initial state, firstly, utilizing invoice tax control equipment to collect a plurality of invoices for buyers and sellers in an initialized mode;
s2, obtaining text information of a plurality of sellers and text information of a purchaser, and forming a seller text set from the text information of the plurality of sellers; forming text information of a plurality of buyers into a buyer text set; (firstly, the invoice tax control equipment is utilized to perform initialized collection on a plurality of invoice buyers and sellers through a data interface mode, and the other mode is that the buyers and the sellers are preset and stored, common client first side information is manually input, and the like, so that text information of the plurality of sellers and text information of the buyers are obtained, and a seller text set and a buyer text set (hereinafter collectively called text set) are constructed;
Step S3, dividing the basic template invoice into areas according to the functions of the basic template invoice to obtain divided subareas, wherein the functions comprise a header, a two-dimensional code, a password area and a table area, numbering the divided subareas according to a preset sequence, sequentially marked as 1,2,..k,..v, and further obtaining the corresponding positions of all the subareas, and constructing a set of sub-region positions W (W1, W2,..wk,..wv), wk representing the position where the kth sub-region is located;
step S4, when the scanning processing operation is carried out on the layout of the current whole invoice, scanning information after scanning is obtained: the scanning of the whole invoice layout comprises the steps of scanning a header, a two-dimensional code, a password area and a table area of the whole invoice, further obtaining area representation information corresponding to each subarea, and further constructing an information set H (H1, H2, hk, hv) of each subarea, wherein Hk represents the area representation information corresponding to the kth subarea;
the form area comprises key information such as buyers and sellers, but the most key is for the form area; simultaneously, the purchaser and the seller correspond to respective important information, namely, regional characterization information of the purchaser and regional characterization information of the seller; the seller comprises a seller text message and a official seal corresponding to the seller; the buyers comprise the buyer text information, the corresponding name or business item (name or business item total name item) of the buyers and the corresponding quantity of the name or business item of the buyers; in other words, the purchaser includes purchaser text information, the corresponding article items of the purchaser, and the number of the purchaser article items;
Step S5, executing scanning information screening processing operation: acquiring and calling subregions of a buyer and a seller in a form region of a current whole invoice, acquiring region characterization information corresponding to the subregions of the buyer and the seller, comparing the region characterization information corresponding to the subregions of the buyer with a buyer text set, judging that the subregions are abnormal subregions if no match exists, judging that the subregions are normal subregions if the match exists, and filtering the normal subregions (if the match exists, representing that the text similarity of the region characterization information corresponding to the subregions of the buyer and a certain buyer information in the buyer text set is larger than a standard threshold value); comparing the regional characterization information corresponding to the subregion of the seller with the seller text set, judging that the subregion is regarded as an abnormal subregion if no match exists, and regarding the subregion as a normal subregion if the match exists, and further filtering the normal subregion (the regional characterization information corresponding to the subregion representing the seller is more than a standard threshold value in the text set of the seller if the match exists, or the regional characterization information corresponding to the regional representation region is completely consistent with the text information of a seller in the text set of the seller if the match exists); if the subarea is abnormal subarea, the subarea is marked as a marked area, the image information in the current whole invoice scanning information corresponding to the marked area is processed, and the whole invoice information is scanned and recorded after confirming that the image information is aligned to the scanning window.
In step S5, acquiring a region characterization information set corresponding to a subregion of the current whole invoice, and then calling the subregions where the buyers and sellers are located in the form region of the current whole invoice, acquiring region characterization information corresponding to the subregions of the buyers and sellers, comparing the region characterization information corresponding to the subregions of the buyers with a buyer text set, wherein under the condition of normal non-overlapping, the scanned region characterization information set corresponding to the subregions of the current whole invoice has a certain matching relationship with the text set stored in the text form; if the scanning of the corresponding information (such as the seller text information and the buyer text information corresponding to the form area in the subarea) in the subarea is correct, the subarea is regarded as a normal subarea, the normal subarea is filtered, and the whole invoice is filtered (part of the invoice is only in a form area which is not shielded, other areas such as a password area are shielded or overlapped, and of course, invoice overlapping can be recognized by matching the profit password area, but the invoice overlapping of the non-form area is not recognized and is not considered in the embodiment of the invention;
if the corresponding information in the sub-region is wrong (namely, when the invoice is scanned, the corresponding region characterization information of the sub-region of the scanned whole invoice is not corresponding to the text set of the corresponding position possibly caused by overlapping of a plurality of invoices), the scanned information image is processed, and then the whole invoice is scanned again and the identification and input operation is carried out.
The financial ticketing processing method based on image recognition recognizes the text information and the image information in the whole invoice through scanning the whole invoice and stores the text information and the image information in a database, so that staff can conveniently input an invoice code or other basic information, and related invoice scanning data can be retrieved, and the acquisition of bill evidence data is realized.
The analysis of the technical scheme can be as follows: the invention relates to a financial ticketing processing method based on image recognition, which is characterized in that an invoice tax control device or preset text information is used for dividing a basic template invoice into areas according to functions of the invoice, so as to obtain divided subareas, wherein the functions comprise a table head, a two-dimensional code, a password area and a table area, the divided subareas are numbered according to a preset sequence, the numbers are sequentially marked as 1, 2. When scanning, scanning processing operation is carried out on the layout of the current whole invoice, and scanned scanning information is obtained: the scanning of the whole invoice layout comprises the steps of scanning the header, the two-dimensional code, the password area and the table area of the whole invoice, further obtaining the area characterization information corresponding to each subarea, and further constructing each subarea information set; then screening and judging whether the region characterization information corresponding to the scanned sub-region is matched with a pre-stored or called text set; if the regional characterization information corresponding to the subregion is accurate, the subregion is regarded as a normal subregion, and then the normal subregion is filtered and stored, and then the corresponding whole invoice is filtered and stored, namely, the invoice is stored in a warehouse;
If the corresponding regional characterization information in the subarea is wrong (the scanned whole invoice is not corresponding due to overlapping of multiple invoices when the invoice is scanned), the scanned whole invoice is marked, the image information in the current whole invoice scanning information corresponding to the marked region is processed again, the whole invoice is scanned again, and the identification and input operation is carried out.
The image information in the current whole invoice scanning information corresponding to the labeling area is processed, and one of the specific embodiments is one of the specific embodiments;
as shown in fig. 2, in step S5, if the corresponding information in the sub-area is wrong, the scanned information image is processed, and then the whole invoice is scanned again and the identification entry operation is performed (when the corresponding information in the sub-area is detected to be wrong, the current invoice is considered to be the current overlapped invoice, and the image information in the scanned information of the current overlapped invoice is processed), specifically including the following steps:
step S51: referring to fig. 4, multiple invoices (multiple invoices at this time are overlapped and not placed in alignment) in a scanning window are divided, and the image information of the scanned current overlapped invoice is obtained as a target image; taking a transverse straight line where the current scanning window is positioned as a preset horizontal line and taking the preset horizontal line as a comparison line of the long-side edge line of the table;
Referring to fig. 3, the outer frame is the current scanning window, and the inner frame is the whole invoice scanning image under the alignment condition;
the transverse straight line where the current scanning window is located is used as a preset horizontal line, and the current scanning window scans a rectangular frame image. Explaining a transverse straight line of the current scanning window, presetting a horizontal line for the current overlapped invoice obtained from the window scanning range of the rectangular current scanning window, and also understanding a physical reference horizontal line which is arranged from the left side edge to the right side edge by taking the outer frame of the current overlapped invoice as a rectangle;
step S52: scanning the table edge of the table area in the identification target image to obtain a table long side edge line and a table short side edge line, and primarily judging whether the first initial condition of the alignment position is met according to the comparison line of the table long side edge line and the table long side edge line (namely, if the table long side edge line is parallel to the comparison line of the table long side edge line, the table area of the target image is primarily considered to be likely to meet the alignment position condition, namely, the comparison line of the table long side edge line and the table long side edge line is parallel when the table long side edge line is completely aligned or completely turned 180 degrees);
Step S53: if the scan identification table long edge line is not parallel to the contrast line of the table long edge line (the table long edge line is not parallel to the contrast line of the table long edge line), the target image is adjusted, and the scan identification entry is performed, see fig. 5, comprising the following steps:
step S531: determining character (or simply character) boundaries in the table area by an edge detection method;
step S532: judging whether the distance between two continuous characters is equal to a preset character distance value or not (if not, the continuous characters of the two continuous characters are not characters belonging to the same row, and filtering the characters), if yes (the continuous characters of the same row can be judged), further judging whether the semantics of the continuous characters have the related semantics of financial notes or not (if not, filtering the semantics of the continuous characters);
step S533: if the two continuous character semantics have the financial bill association semantics (preset association semantics), determining that the two continuous characters with the financial bill association semantics are a first character and a second character in sequence, and further determining that the long-edge line of the target table area in the target image from the first character to the second character is the first edge of the target table area of the target image;
Step S534: determining a vertical line with a first edge of a target table area of the target image as a second edge of the table (namely, a short edge line of the table), enabling the first edge of the table and the second edge of the table to rotate clockwise (processing is carried out by using an image rotation tool, namely, the target image acquired by whole scanning is rotated), and driving the target image to rotate clockwise to carry out target image position adjustment;
step S535: and detecting whether the first edge of the target table area of the rotated target image is parallel to a preset horizontal line in real time, if so, regarding the table edge as being in an aligned position, and further performing scanning identification entry on the uppermost invoice serving as the target image after finishing adjustment.
According to the scheme, when the specific operation is performed, the corresponding information in the detection subarea is wrong, the current invoice is considered to be the current overlapped invoice, the image information in the scanning information of the current overlapped invoice is processed, the target image is obtained through image segmentation, parallel lines are preset to serve as comparison lines of table edge lines, the long-side edge lines and the short-side edge lines are obtained through scanning and identifying the edge lines of the table area in the target image, and whether the comparison lines of the long-side edge lines and the table edge lines meet a first initial condition or not is judged, namely, the long-side edge lines are parallel to the comparison lines of the table edge lines; if not, the target image is regarded as not being aligned with the scanning window, and the target image is subjected to rotation processing:
And if so, determining that the edge line of the long edge of the table of the target table area in the target image in which any two character extension vector directions of the continuous characters are parallel is the first edge of the target table area of the target image, and determining that the vertical line with the first edge of the table is the second edge of the table.
And detecting whether a first edge line of a table area of the rotated target image is parallel to a preset horizontal line in real time, if so, considering that the table edge is at an alignment position, and then completely identifying the information of the whole invoice by the scanning image, and carrying out scanning identification and input on the information.
Specifically, referring to fig. 6, in step S533, it is determined that two consecutive character semantics have financial instrument association semantics, which specifically includes the steps of:
s5331, presetting a plurality of financial bill related vocabulary sets, and combining a semantic library constructed by the plurality of financial bill related vocabulary sets;
step S5332, firstly, acquiring current two continuous characters, matching the current two continuous characters with a current semantic library, and judging that the current two continuous characters have financial bill association semantics if semantic matching is successful; and then acquiring a corresponding financial bill related vocabulary set successfully matched with the current two continuous characters.
If the two characters are the same, the current character is determined to be the first character to the second character, and the long-side edge line of the table with the parallel vector directions of the first character to the second character is determined to be the first edge of the table.
By analyzing the scheme, according to the financial ticketing processing method based on image recognition, through presetting a plurality of financial bill related word sets, merging a semantic library constructed by the plurality of financial bill related word sets, and through matching the acquired plurality of two continuous characters with the semantic library, if the matching is successful, the long edge line of the table with the parallel vector directions from any first character to second character of the plurality of two continuous characters can be acquired as the first edge of the table.
The second embodiment of the method processes the image information in the current whole invoice scanning information corresponding to the labeling area;
in this embodiment, the processing of the image information in the current whole invoice scan information corresponding to the labeling area, see fig. 7, specifically includes the following steps:
step S54: dividing the scanned image information of the current overlapped invoice to obtain an uppermost invoice image as a target image; taking a transverse straight line where the current scanning window is positioned as a preset horizontal line and taking the preset horizontal line as a comparison line of the long-side edge line of the table;
Step S55: obtaining a long-side edge line and a short-side edge line of a table according to the table edge of a table area in a scanning identification target image, and preliminarily judging whether the first initial condition of the alignment position is met or not according to the comparison line of the long-side edge line and the long-side edge line of the table; the first initial condition is that the current long-side edge line of the table and the contrast line of the long-side edge line of the table are in parallel relation;
step S56: if the scan identification table long edge line is not parallel to the comparison line of the table long edge line, the target image is adjusted, and the scan identification entry is performed on the adjusted target image which meets the second initial condition, see fig. 8, and the method comprises the following steps:
step S561: determining character boundaries (character boundaries) within the fields of the form field by edge detection;
step S562: acquiring character boundaries of two continuous current characters in a form area, determining two continuous character intervals, judging whether the two continuous character intervals are equal to a preset character interval value, if so, regarding the two continuous characters as the same row, and further judging whether a plurality of discontinuous continuous two character semanteme in the same row has financial bill related semanteme; the two continuous character semanteme is regarded as a phrase, and a plurality of discontinuous two continuous character semantemes in the same row are regarded as a plurality of discontinuous phrases in the same row;
Step S563: if a plurality of discontinuous word groups in the same row are judged to have financial bill association semantics, determining that any two characters in the same row with the financial bill association semantics are a first character and a second character in sequence, and further determining that a long-side edge line of a table in a target table area in a target image with parallel character extension vector directions from the first character to the second character is a first edge of the target table area in the target image;
step S564: determining a vertical line with a first edge of a target table area of the target image as a second edge of the table, enabling the first edge of the table and the second edge of the table to rotate clockwise, and driving the target image to rotate clockwise to adjust the position of the target image;
step S565: simultaneously detecting whether the first edge of a target table area of the rotated target image is parallel to a preset horizontal line in real time, if so, regarding the table edge as being in an aligned position, and further performing scanning identification entry on the uppermost invoice serving as the target image after finishing adjustment;
the second initial condition is that the distance between any current continuous two characters in the same row is consistent with the preset character distance value, and a plurality of discontinuous phrases in the same row accord with the association semantics of the financial bill.
In the implementation process of step S563 in the above specific embodiment, the determining that the plurality of discontinuous phrases in the same line have the financial document associated semantics, see fig. 9, specifically includes the following steps:
step S5631: presetting a plurality of financial bill related vocabulary sets and combining a semantic library constructed by the plurality of financial bill related vocabulary sets;
step S5632: firstly, a plurality of discontinuous word groups in the same row are obtained, matching is carried out according to the plurality of discontinuous word groups in the same row and a current semantic library, and if semantic matching is successful, the plurality of discontinuous word groups in the same row currently have financial bill associated semantics; and then acquiring a corresponding financial bill related vocabulary set successfully matched with a plurality of discontinuous phrases in the same row.
Specifically, as shown in fig. 10, in step S531 corresponding to the first embodiment or step S561 corresponding to the second embodiment, the boundary of the character (or simply, the character) in the table area is determined by the edge detection method, and the method includes the following steps:
step S5311: performing binarization processing on the target image (the binarization processing mode is common knowledge and is not repeated), so as to obtain a black-and-white image (at the moment, an RGB value of 255 images is white or an RGB value of 0 images is black);
Step S5312: defining the first edge of the table area and the second edge of the table area as X axis and Y axis respectively, taking X, Y axis direction as basic direction to make the identification point from X,Starting from the origin of the Y-axis intersection point, scanning downwards, wherein the scanning traversing width is 1px; first for one character Z of two continuous characters 1 Scanning and traversing character boundaries of the pixel, if the RGB value of the scanned current pixel point is 0, recording the coordinate A at the moment 1 Thereby obtaining Z 1 Coordinates a of the left boundary point of (2) 1 (Z 1 Is the complete coordinate A 1 ) Then continuing to scan downwards to obtain the pixel point coordinate A of the next character 2 ；
Step S5313: identifying the point from A 1 +1 starting, scanning longitudinally through each pixel point on the whole target image, if the RGB value of the current pixel point of the scanning is 255, recording the coordinate B at the moment 1 According to the coordinates B at this time 1 Thereby obtaining Z 1 Coordinates of the right boundary point of (B-1) 1 (i.e., B-1) 1 Is the original coordinate point B 1 The abscissa of (1) is the original coordinate point B 1 Is the ordinate value of (2) ) The method comprises the steps of carrying out a first treatment on the surface of the Then continuing to scan and traverse to the right along the X-axis direction to obtain the pixel point coordinate B of the next character 2 ；
Step S5314: in M interval (A) 1 ，(B-1) 1 ) In the method, the identification point is longitudinally scanned from the point with the Y coordinate being 0, the R value of each point is judged, the scanning is stopped until the RGB value is equal to 0, and the coordinate C at the moment is recorded 1 Thereby obtaining Z 1 Coordinates C of the upper boundary point of (2) 1 (i.e. Z 1 Coordinates C of the upper boundary point of (2) 1 ) The method comprises the steps of carrying out a first treatment on the surface of the Then continuing scanning downwards along the Y-axis direction to obtain the pixel point coordinate C of the next character 2 ；
Step S5315: in M interval (A) 1 ，(B-1) 1 ) In (3), the identification point is from coordinate point C 1 +1 (coordinate point C) 1 +1 is the original coordinate point C 1 The X-axis value of (C) is unchanged, and the original coordinate point C 1 The Y-axis value +1 of (2) to obtain a coordinate point C 1 Y-axis value of +1), the R value of each pixel is determined, and if the R value is equal to 255, the scanning is stopped to record the coordinate D at that time 1 (i.e. Z 1 Coordinates of the lower boundary point (D-1) 1 )；
Step S5316: according to the current character Z 1 Coordinates of four boundary points of the left boundary point, the right boundary point, the upper boundary point and the lower boundary point, and calculating the current character Z according to the coordinates of the four boundary points 1 Is described as Z 1 (A 1 ，(B-1) 1 ,C 1 ,(D-1) 1 ) Is a character boundary;
step S5317: then repeating the above steps to obtain another character Z of the two continuous characters 2 Is described as Z 2 (A 2 ，(B-1) 2 ,C 2 ,(D-1) 2 ) Is a character boundary.
By analyzing the scheme, the first edge of the table area is defined as the X axis, the second edge of the table area is defined as the Y axis, each pixel point is traversed in sequence from the origin of the X, Y-axis intersection point, and one character Z of two continuous characters is obtained 1 The coordinates of the boundary points, i.e. Z 1 (A 1 ，(B-1) 1 ,C 1 ,(D-1) 1 ) Further, repeat Z 1 Coordinates of the boundary point to thereby obtain the other character Z of the two consecutive characters 2 Character boundary point coordinate of Z 2 (A 2 ，(B-1) 2 ,C 2 ,(D-1) 2 )。
Specifically, the continuous two-character distance calculating mode is a difference between a boundary coordinate of a continuous second character and a boundary coordinate of a first character, and the calculating formula is as follows:
Z 2 (A 2 ，(B-1) 2 ,C 2 ,(D-1) 2 )-Z 1 (A 1 ，(B-1) 1 ,C 1 ,(D-1) 1 )；
wherein Z is 2 Character boundary coordinates for the second character; z is Z 1 Is the first character boundary coordinates.
In summary, the invention obtains a large amount of invoice information through scanning, is convenient for staff to manage and search the large amount of invoices, and when the scanned images are inaccurate due to overlapping and even rotating placement of a plurality of invoices, the uppermost layer is a target image (namely the uppermost invoice) obtained by dividing the scanned invoice image, and the table line of the table area in the scanned target image recognizes that the placement of the whole invoice is not aligned;
and further determining that the continuous characters are the same row of characters through the spacing of boundary points of the continuous characters in the table area, further determining that the continuous characters in the same row of characters accord with financial related semantic information, further determining that edge lines which are parallel to any character vector direction straight lines in the current same row of characters are first edge lines (namely long edge lines), further determining second edge lines (namely short edge lines) which are connected with the first edge lines and are vertical to the first edge lines, and performing scanning identification and input invoice information by rotating the first edge lines and the second edge lines (namely rotating target images acquired through whole scanning) to target image alignment scanning frames.
Example two
The present invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the image recognition-based financial ticketing method in embodiment one.
The foregoing is merely illustrative and explanatory of the principles of the invention, as various modifications and additions may be made to the specific embodiments described, or similar thereto, by those skilled in the art, without departing from the principles of the invention or beyond the scope of the appended claims.
Claims (9)
1. A financial ticketing processing method based on image recognition is characterized in that: the method comprises the following steps:
obtaining text information of a plurality of sellers and text information of buyers, and forming the text information of the plurality of sellers into a seller text set; forming text information of a plurality of buyers into a buyer text set;
dividing the basic template invoice into areas according to functions of the basic template invoice to obtain divided subareas, wherein the functions comprise a header, a two-dimensional code, a password area and a table area, the divided subareas are numbered according to a preset sequence and are sequentially marked as 1,2,..k,..v, further obtaining positions corresponding to the subareas, and constructing a position set W (W1, W2,..Wk,..Wv) of each subarea, wherein Wk represents the position of the kth subarea;
When the scanning processing operation is carried out on the layout of the current whole invoice, scanning information after scanning is obtained: the scanning of the whole invoice layout comprises the steps of scanning a header, a two-dimensional code, a password area and a table area of the whole invoice, further obtaining area representation information corresponding to each subarea, and further constructing an information set H (H1, H2, hk, hv) of each subarea, wherein Hk represents the area representation information corresponding to the kth subarea;
and executing scanning information screening processing operation: acquiring and calling subregions of the current whole invoice in a form region of the buyer and the seller, acquiring region characterization information corresponding to the subregions of the buyer and the seller, comparing the region characterization information corresponding to the subregions of the buyer with a buyer text set, judging that the subregions are abnormal subregions if no match exists, judging that the subregions are normal subregions if the match exists, and filtering the normal subregions; comparing the regional characterization information corresponding to the subregion of the seller with the seller text set, judging that the subregion is regarded as an abnormal subregion if no match exists, and regarding as a normal subregion if the match exists, and further filtering the normal subregion; if the subarea is abnormal subarea, the subarea is marked as a marked area, then the image information in the current whole invoice scanning information corresponding to the marked area is processed, and the whole invoice information is scanned and recorded after confirming that the image information is aligned to the scanning window.
2. The financial ticketing processing method based on image recognition according to claim 1, wherein the processing of the image information in the current whole invoice scan information corresponding to the labeling area specifically comprises the following steps:
dividing the scanned image information of the current overlapped invoice to obtain an uppermost invoice image as a target image; taking a transverse straight line where the current scanning window is positioned as a preset horizontal line and taking the preset horizontal line as a comparison line of the long-side edge line of the table;
obtaining a long-side edge line and a short-side edge line of a table according to the table edge of a table area in a scanning identification target image, and preliminarily judging whether the first initial condition of the alignment position is met or not according to the comparison line of the long-side edge line and the long-side edge line of the table; the first initial condition is that the current long-side edge line of the table and the contrast line of the long-side edge line of the table are in parallel relation;
if the scanning identification form long-side edge line is not parallel to the comparison line of the form long-side edge line, the target image is adjusted, and the adjusted target image which accords with the second initial condition is scanned, identified and recorded, comprising the following steps:
Determining character boundaries in the areas of the form area by an edge detection method;
acquiring character boundaries of two continuous current characters in a form area, determining two continuous character intervals, judging whether the two continuous character intervals are equal to a preset character interval value, and if yes, further judging whether the two continuous character semantics have financial bill related semantics;
if the two continuous character semantics have the financial bill association semantics, determining that the two continuous characters with the financial bill association semantics are a first character and a second character in sequence, and further determining that the edge line of the long edge of the table of the target table area in the target image with the parallel character extension vector directions from the first character to the second character is the first edge of the target table area of the target image;
determining a vertical line with a first edge of a target table area of the target image as a second edge of the table, enabling the first edge of the table and the second edge of the table to rotate clockwise, and driving the target image to rotate clockwise to adjust the position of the target image;
simultaneously detecting whether the first edge of a target table area of the rotated target image is parallel to a preset horizontal line in real time, if so, regarding the table edge as being in an aligned position, and further performing scanning identification entry on the uppermost invoice serving as the target image after finishing adjustment;
The second initial condition is that the distance between any current continuous two characters in the same row is consistent with the preset character distance value, and the semantics of any current continuous two characters in the same row accord with the financial bill association semantics.
3. The method for processing financial ticketing based on image recognition according to claim 2, wherein the step of determining that two consecutive character semantics have financial ticket association semantics specifically comprises the following steps:
presetting a plurality of financial bill related vocabulary sets and combining a semantic library constructed by the plurality of financial bill related vocabulary sets;
firstly, acquiring current two continuous characters, matching the current two continuous characters with a current semantic library, and judging that the current two continuous characters have financial bill associated semantics if the semantics match is successful; and then acquiring a corresponding financial bill related vocabulary set successfully matched with the current two continuous characters.
4. The financial ticketing processing method based on image recognition according to claim 3, wherein the processing of the image information in the current whole invoice scan information corresponding to the labeling area specifically comprises the following steps:
Dividing the scanned image information of the current overlapped invoice to obtain an uppermost invoice image as a target image; taking a transverse straight line where the current scanning window is positioned as a preset horizontal line and taking the preset horizontal line as a comparison line of the long-side edge line of the table;
obtaining a long-side edge line and a short-side edge line of a table according to the table edge of a table area in a scanning identification target image, and preliminarily judging whether the first initial condition of the alignment position is met or not according to the comparison line of the long-side edge line and the long-side edge line of the table; the first initial condition is that the current long-side edge line of the table and the contrast line of the long-side edge line of the table are in parallel relation;
if the scanning identification form long-side edge line is not parallel to the comparison line of the form long-side edge line, the target image is adjusted, and the adjusted target image which accords with the second initial condition is scanned, identified and recorded, comprising the following steps:
determining character boundaries in the areas of the form area by an edge detection method;
acquiring character boundaries of two continuous current characters in a form area, determining two continuous character intervals, judging whether the two continuous character intervals are equal to a preset character interval value, if so, regarding the two continuous characters as the same row, and further judging whether a plurality of discontinuous continuous two character semanteme in the same row has financial bill related semanteme; the two continuous character semanteme is regarded as a phrase, and a plurality of discontinuous two continuous character semantemes in the same row are regarded as a plurality of discontinuous phrases in the same row;
If a plurality of discontinuous word groups in the same row are judged to have financial bill association semantics, determining that any two characters in the same row with the financial bill association semantics are a first character and a second character in sequence, and further determining that a long-side edge line of a table in a target table area in a target image with parallel character extension vector directions from the first character to the second character is a first edge of the target table area in the target image;
determining a vertical line with a first edge of a target table area of the target image as a second edge of the table, enabling the first edge of the table and the second edge of the table to rotate clockwise, and driving the target image to rotate clockwise to adjust the position of the target image;
simultaneously detecting whether the first edge of a target table area of the rotated target image is parallel to a preset horizontal line in real time, if so, regarding the table edge as being in an aligned position, and further performing scanning identification entry on the uppermost invoice serving as the target image after finishing adjustment;
the second initial condition is that the distance between any current continuous two characters in the same row is consistent with the preset character distance value, and a plurality of discontinuous phrases in the same row accord with the association semantics of the financial bill.
5. The method for processing financial ticketing based on image recognition as defined in claim 4, wherein the step of determining that a plurality of discontinuous word groups in the same line have financial ticket association semantics comprises the following steps:
presetting a plurality of financial bill related vocabulary sets and combining a semantic library constructed by the plurality of financial bill related vocabulary sets;
firstly, a plurality of discontinuous word groups in the same row are obtained, matching is carried out according to the plurality of discontinuous word groups in the same row and a current semantic library, and if semantic matching is successful, the plurality of discontinuous word groups in the same row currently have financial bill associated semantics; and then acquiring a corresponding financial bill related vocabulary set successfully matched with a plurality of discontinuous phrases in the same row.
6. The method for processing financial ticketing based on image recognition according to claim 2 or 4, wherein the semantic library comprises an enterprise information semantic library, a bank information semantic library, an industry service category information semantic library and a product category information semantic library.
7. The method for processing financial ticketing based on image recognition according to claim 2 or 4, wherein the determining the character boundary in the form area by the edge detection method comprises the following steps:
Performing binarization processing on the target image to obtain a black-and-white image;
defining a first edge of the table area and a second edge of the table area as an X axis and a Y axis respectively, taking a X, Y axis direction as a basic direction, and enabling the identification point to start from an X axis coordinate of 0 to scan downwards, wherein the scanning traversing width is 1px;
first for one character Z of two continuous characters 1 Scanning and traversing character boundaries of the pixel, if the RGB value of the scanned current pixel point is 0, recording the coordinate A at the moment 1 Thereby obtaining Z 1 Coordinates a of the left boundary point of (2) 1 Then continuing to scan downwards to obtain the pixel point coordinate A of the next character 2 The method comprises the steps of carrying out a first treatment on the surface of the Identifying the point from A 1 +1 starting, scanning longitudinally through each pixel point on the whole target image, if the RGB value of the current pixel point of the scanning is 255, recording the coordinate B at the moment 1 According to the coordinates B at this time 1 Thereby obtaining Z 1 Coordinates of the right boundary point of (B-1) 1 (i.e., B-1) 1 Is the original coordinate point B 1 The abscissa of (1) is the original coordinate point B 1 Is the ordinate value of (a);
in M interval (A) 1 ，(B-1) 1 ) In the method, the identification point is longitudinally scanned from the point with the Y coordinate being 0, the R value of each point is judged, the scanning is stopped until the RGB value is equal to 0, and the coordinate C at the moment is recorded 1 Thereby obtaining Z 1 Coordinates C of the upper boundary point of (2) 1 ；
In M interval (A) 1 ，(B-1) 1 ) In (3), the identification point is from coordinate point C 1 Starting longitudinal scanning at +1, judging R value of each pixel point, and stopping scanning to record the coordinate D at the moment if the R value is equal to 255 1 ；
According to the current character Z 1 Coordinates of four boundary points of the left boundary point, the right boundary point, the upper boundary point and the lower boundary point, and calculating the current character Z according to the coordinates of the four boundary points 1 Is described as Z 1 (A 1 ，(B-1) 1 ,C 1 ,(D-1) 1 ) Is a character boundary;
then repeating the above steps to obtain another character Z of the two continuous characters 2 Is described as Z 2 (A 2 ，(B-1) 2 ,C 2 ,(D-1) 2 ) Is a character boundary.
8. The method for processing financial ticketing based on image recognition as set forth in claim 7, wherein the character spacing is a character boundary coordinate of the second character minus a first character boundary coordinate, and the calculation formula is:
Z 2 (A 2 ，(B-1) 2 ,C 2 ,(D-1) 2 )-Z 1 (A 1 ，(B-1) 1 ,C 1 ,(D-1) 1 )；
wherein Z is 2 Character boundary coordinates for the second character; z is Z 1 Is the first character boundary coordinates.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310446267.2A CN116486423A (en) | 2023-04-24 | 2023-04-24 | Financial ticketing data processing method based on image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310446267.2A CN116486423A (en) | 2023-04-24 | 2023-04-24 | Financial ticketing data processing method based on image recognition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116486423A true CN116486423A (en) | 2023-07-25 |
Family
ID=87215144
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310446267.2A Pending CN116486423A (en) | 2023-04-24 | 2023-04-24 | Financial ticketing data processing method based on image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116486423A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116824604A (en) * | 2023-08-30 | 2023-09-29 | 江苏苏宁银行股份有限公司 | Financial data management method and system based on image processing |
-
2023
- 2023-04-24 CN CN202310446267.2A patent/CN116486423A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116824604A (en) * | 2023-08-30 | 2023-09-29 | 江苏苏宁银行股份有限公司 | Financial data management method and system based on image processing |
| CN116824604B (en) * | 2023-08-30 | 2023-11-21 | 江苏苏宁银行股份有限公司 | Financial data management method and system based on image processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105913093B (en) | A kind of template matching method for Text region processing | |
| US6408094B1 (en) | Document image assessment system and method | |
| CN104112128B (en) | Digital image processing system and method applied to bill image character recognition | |
| US7590275B2 (en) | Method and system for recognizing a candidate character in a captured image | |
| CN111476109A (en) | Bill processing method, bill processing apparatus, and computer-readable storage medium | |
| CN110766014A (en) | Bill information positioning method, system and computer readable storage medium | |
| CN108549843A (en) | A kind of VAT invoice recognition methods based on image procossing | |
| US20090290751A1 (en) | Methods and Systems for Detecting Numerals in a Digital Image | |
| US5915039A (en) | Method and means for extracting fixed-pitch characters on noisy images with complex background prior to character recognition | |
| CN112183038A (en) | Form identification and typing method, computer equipment and computer readable storage medium | |
| JP4078009B2 (en) | CHARACTERISTIC RECORDING AREA DETECTION DEVICE FOR FORM, CHARACTER RECORDING AREA DETECTION METHOD FOR FORM, STORAGE MEDIUM, AND FORM FORMAT CREATION DEVICE | |
| JP2015146075A (en) | accounting data input support system, method, and program | |
| CN110490190B (en) | Structured image character recognition method and system | |
| CN109948621B (en) | Image processing and character segmentation method based on picture verification code | |
| CN103034848A (en) | Identification method of form type | |
| CN116486423A (en) | Financial ticketing data processing method based on image recognition | |
| CN113780087A (en) | Postal parcel text detection method and equipment based on deep learning | |
| CN111310682B (en) | Universal detection analysis and recognition method for text file forms | |
| CN111461133A (en) | Express delivery surface single item name identification method, device, equipment and storage medium | |
| US7013045B2 (en) | Using multiple documents to improve OCR accuracy | |
| CN104077562B (en) | A kind of scanning direction determination methods of test paper | |
| CN111738031B (en) | One-dimensional bar code identification method | |
| KR101486495B1 (en) | Shape clustering in post optical character recognition processing | |
| CN110634222B (en) | Bank bill information identification method | |
| JP2003109007A (en) | Device, method and program for classifying slip form and image collating device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |