CN109086766A - A kind of multi threshold fusion crown word number extracting method based on integrogram - Google Patents
A kind of multi threshold fusion crown word number extracting method based on integrogram Download PDFInfo
- Publication number
- CN109086766A CN109086766A CN201810574416.2A CN201810574416A CN109086766A CN 109086766 A CN109086766 A CN 109086766A CN 201810574416 A CN201810574416 A CN 201810574416A CN 109086766 A CN109086766 A CN 109086766A
- Authority
- CN
- China
- Prior art keywords
- image
- word number
- crown word
- threshold
- obtains
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/153—Segmentation of character regions using recognition of characters or words
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07D—HANDLING OF COINS OR VALUABLE PAPERS, e.g. TESTING, SORTING BY DENOMINATIONS, COUNTING, DISPENSING, CHANGING OR DEPOSITING
- G07D7/00—Testing specially adapted to determine the identity or genuineness of valuable papers or for segregating those which are unacceptable, e.g. banknotes that are alien to a currency
- G07D7/20—Testing patterns thereon
- G07D7/2016—Testing patterns thereon using feature extraction, e.g. segmentation, edge detection or Hough-transformation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The multi threshold fusion crown word number extracting method based on integrogram that the invention discloses a kind of, the White-light image including crown word number image one side and normalization including (1) acquisition object to be identified obtain the gray level image comprising crown word number image-region；(2) Low threshold binary conversion treatment is carried out to the first image and obtains the first image, and project to obtain the second image by row, column；(3) integrogram is calculated to the second image and calculates the Haar-Like H feature in integrogram；(4) the anti-binary conversion treatment of Low threshold is carried out to the second image, obtains third image only comprising prefix sign character, confidence matrix is calculated according to third image；(5) the anti-binary conversion treatment of high threshold is carried out to the second image and obtains the 4th image；The 5th image is obtained using the background texture that Haar-Like H characteristic coordinates remove in the 4th image between prefix sign character；(6) mask process is done to the 5th image using confidence matrix, removes the texture in prefix sign character, obtains clear crown word number image；This method is suitable for a variety of conditions and realizes that crown word number is accurately extracted.
Description
Technical field
The invention belongs to technical field of image processing, more particularly, under a kind of complex background based on the more of integrogram
Threshold Fusion crown word number extracting method.
Background technique
With economic rapid and healthy, need to ticket (including bank note, license, check, property right with public credit
Card etc.) effectively supervised, this kind of ticket can all have crown word number unique identification, to the supervision of crown word number to national national defense safety with
Financial economic security has great significance.Most of automatic ticket identification machine, which does not have, currently on the market accurately identifies prefix
Number ability；The machine that crown word number extracts registering capacity is extracted towards the crown word number under the simple background such as pure color mostly,
Complex texture and have the crown word number recovery rate of obvious stain extremely low for having, be unable to satisfy it is a variety of under the conditions of extraction and identification want
It asks.
The method for extracting crown word number to banknote figure in the prior art first carries out level correction processing to banknote image to reduce
Error caused by banknote image position occurs at random in source images, then extracts characteristic value, determines face amount towards then calculating spy
The gray value of value indicative is matched, and the method that this kind of banknote figure to scanning extracts crown word number can give the true and false knowledge of bank note
Other strong support；However the crown word number that can not be applied under the conditions of having complex texture or having obvious stain is extracted.
The existing technology to the processing of dirty crown word number is mostly the technology classified, by extracting in banknote image laterally containing hat
The precise region of font size and longitudinal precise region containing crown word number, the character of the relatively more lateral precise region containing crown word number
The stain number of the corresponding position character of stain number and longitudinal precise region containing crown word number, according to comparison result to prefix
Number classify, can Accurate classification bank note dirty crown word number and normal crown word number, but the dirty crown word number figure for sorting out
As no longer being identified.
Summary of the invention
Aiming at the above defects or improvement requirements of the prior art, the present invention provides a kind of multi-thresholds based on integrogram to melt
Crown word number extracting method is closed, its object is to realize in complex background or the crown word number image for having obvious stain to crown word number
Effectively extract.
Purpose to realize the present invention provides a kind of multi-threshold based on integrogram and melts according to one aspect of the present invention
Crown word number extracting method is closed, is included the following steps:
(1) White-light image including crown word number image one side of object to be identified is acquired, and is normalized, is obtained
The first image comprising crown word number image-region；
(2) gray level image is carried out binary conversion treatment to obtain using first threshold including imperfect crown word number but not wrapping
The first image containing background obtains the second image, the as crown word number of coarse positioning by row projection and column projection to the first image
Image；
(3) integrogram is calculated to the second image, and calculates the Haar-Like H feature in integrogram；
(4) the anti-binary conversion treatment of Low threshold is carried out to the second image, obtains third image only comprising prefix sign character, root
It is calculated according to the third image and obtains confidence matrix M；
(5) the anti-binary conversion treatment of high threshold is carried out to the second image, obtains the 4th image；Utilize Haar-Like H feature
The background texture that coordinate removes in the 4th image between prefix sign character obtains the 5th image；
(6) mask process is done to the 5th image using confidence matrix M, removes the texture in prefix sign character, it is clear to obtain
Crown word number image.
Preferably, above-mentioned multi threshold fusion crown word number extracting method, step (1) include: bright to normalized image progress
Degree compensates and obtains the denomination information of image, navigates to the gray level image comprising crown word number according to denomination information.
Preferably, above-mentioned multi threshold fusion crown word number extracting method, the method that step (2) obtains the second image include such as
Lower sub-step:
(2.1) it projects to obtain line number group RowSum [i] by carrying out line direction to the first image, wherein i is the first image
Row number；
Column direction is carried out to the first image to project to obtain columns group ColSum [j]；Wherein j is the column number of the first image；
(2.2) coboundary, lower boundary, left margin and the right margin of the first image are determined；It is specific as follows:
It searches for from top to bottom, finds the condition of satisfaction " array RowSum [i]+RowSum [i+1] > T21And RowSum [t] >
T31" point, the coboundary Yu of the first image will be determined as by these lines for constituting of point；
Constantly search downwards, until occurring being unsatisfactory for " array RowSum [i]+RowSum [i+1] > T21And RowSum [t] >
T31" point, the lower boundary Yd of the first image will be determined as by these lines for constituting of point；
It searches for from left to right, finds the condition of satisfaction " array ColSum [t]+ColSum [t+1] > T22And ColSum [t] >
T32" point, the left margin Xt of the first image will be determined as by these lines for constituting of point；
It turns left search from the right side, finds the condition of satisfaction " array ColSum [t]+ColSum [t-1] > T22And ColSum [t] >
T32" point, the right margin Xr of the first image will be determined as by these lines for constituting of point；
Wherein, T21Refer to the first row boundary threshold, T31Refer to the second row bound threshold value, T22Refer to first row boundary threshold,
T32Refer to secondary series boundary threshold；
(2.3) upper left corner A (Xt, Yu), the lower right corner B of crown word number of crown word number are found by positioning the region of crown word number
(Xr,Yd)；The subgraph comprising crown word number image is found in former gray level image from upper left corner C (Xt-10, Yu-10) to bottom right
The region of angle D (Xr-10, Yd-10) constitutes accurate crown word number image, i.e. the second image.
Preferably, above-mentioned multi threshold fusion crown word number extracting method,
Its first threshold T=average*w；
The gray average of whole picture gray level image
Wherein, I1(i, j) is that coordinate is the gray value of the point of (i, j) on gray level image, and w is the weight of mean value.
Preferably, above-mentioned multi threshold fusion crown word number extracting method, step (3) includes following sub-step:
(3.1) indicated in integrogram from the upper left corner of the second image to the cumulative of the gray value of point (x, y) with A (x, y) and；
Wherein, I2(i, j) be on the second image coordinate be (i, j) point gray value；
(3.2) a two-dimensional array Arr [H] [W] is created for integrogram, and all initialized；Wherein H is the second image
Height, W be the second image width；
(3.3) it is calculate by the following formula the integrogram of the second image the first row,
Arr [0] [x]=Arr [0] [x-1]+I [0] [x]；
Wherein I [0] [x] indicates that the coordinate of the second image Img2 is the gray value of the point of (x, 0)；
(3.4) integrogram is all calculated to all the points on the second image Img2 by following formula, obtains complete integrogram
Arr；
Arr [y] [x]=Arr [y] [x-1]+Arr [y-1] [x]-Arr [y-1] [x-1]+I [y] [x].
Preferably, above-mentioned multi threshold fusion crown word number extracting method, step (4) includes following sub-step:
(4.1) second threshold T is used to the second image5Anti- binary conversion treatment is carried out, obtains only including prefix sign character
Third image；
Wherein, second threshold
M is the line number of the second image, and N is the columns of the second image；I (i, j) is the point that coordinate is (i, j) on the second image
Gray value, a is weight；
(4.2) confidence matrix M [C] [D] is calculated, wherein C is the line number of third image, and D is the columns of third image；
To the confidence matrix of initialization into line scans, the point that M [i] [j] value be 0 but front and back existence value is 1 then will
Its corresponding value M [i] [j] is set to X, thus obtains the confidence matrix scanned through space；
Column scan is carried out to the confidence matrix scanned through space, the point that M [i] [j] value be 0 but existence value is 1 up and down,
Then X will be set to by value M [i] [j] accordingly, and obtain the confidence matrix M scanned by row, column.
Preferably, above-mentioned multi threshold fusion crown word number extracting method, step (5) are as follows: third threshold value is used to the second image
T6 carries out anti-binary conversion treatment and obtains the 4th image；The non-character regional value of the 4th image is set by Haar-Like H feature
It is 0, removes the background texture in the 4th image between prefix sign character and obtain the 5th image, extracts knot for final crown word number
Fruit.
Preferably, above-mentioned multi threshold fusion crown word number extracting method is sat as M [i] [j]=X to corresponding in the 5th image
1 point is designated as then to retain；As M [i] [j]=0, the value of the 5th image respective coordinates (i, j) is directly set 0, the hat extracted
Font size image.
In general, through the invention it is contemplated above technical scheme is compared with the prior art, can obtain down and show
Beneficial effect:
Multi threshold fusion crown word number extracting method provided by the invention based on integrogram, using Haar-Like H feature
And multi threshold fusion has carried out effective removal to crown word number background image；Block region energy is based on by Haar-Like H feature
It is enough to be effectively concerned about character zone, non-character region is removed, so intercharacter threadiness complex texture and dotted stain can be
It is effectively removed in Haar-Like H characteristic extraction procedure；The linear texture in dropping character can be removed by confidence matrix
Or dotted stain；And compare, crown word number of the tradition based on single threshold value binaryzation is extracted, due to cannot be by character and back
Scape is kept completely separate, and is split using crown word number projecting method to crown word number, causes Character segmentation fracture occur or adhesion is asked
Topic, extreme influence to subsequent crown word number recognition accuracy.
Detailed description of the invention
Fig. 1 is that the process of the embodiment of the multi threshold fusion crown word number extracting method provided by the invention based on integrogram is shown
It is intended to；
Fig. 2 is the Haar-Like H characteristic pattern in embodiment；
Fig. 3 is the Haar-Like H integrogram in embodiment；
Fig. 4 is the crown word number image Img in embodiment；
Fig. 5 is the anti-binary picture Img1 in embodiment at threshold value T1；
Fig. 6 is the transverse projection of Img1 in embodiment；
Fig. 7 is the longitudinal projection of Img1 in embodiment；
Fig. 8 is the third image Img3 for carrying out binary conversion treatment acquisition in embodiment using Low threshold T5；
Fig. 9 is the 4th image Img4 for carrying out anti-binary conversion treatment acquisition in embodiment using high threshold T6；
Figure 10 is that the 4th image Img4 in embodiment passes through the 5th image obtained after Haar-Like H characteristic processing
Img5；
Figure 11 is the crown word number image bwImg finally identified in embodiment.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.As long as in addition, technical characteristic involved in the various embodiments of the present invention described below
Not constituting a conflict with each other can be combined with each other.
The process of multi threshold fusion crown word number extracting method provided by the invention based on integrogram, referring to Fig.1；It ties below
The example that crown word number extraction is carried out to 50 dollars of images of see-through plate is closed, is specifically illustrated, referring to Fig. 2~Figure 11, to identify
Intermediate image in journey.
(1) acquisition 50 dollars of images of see-through plate include the White-light image of crown word number image one side, and are normalized to 950*400
The size of pixel；Accurate crown word number image is obtained by the approximate region that coarse positioning finds crown word number, coarse positioning includes following son
Step:
(1.1) luminance compensation is carried out to normalized image, and obtains the denomination information of image, navigated to according to denomination information
One gray level image Img comprising crown word number, that size is 50*250 pixel, referring to Fig. 4；
(1.2) pass through Low threshold T1Binaryzation is carried out to the gray level image and obtains the first image Img1, the first image packet
Containing imperfect crown word number, background is not included, referring to Fig. 5；
The gray average of whole picture gray level image
Thresh=average*w
Wherein, average is the gray average of whole picture gray level image, I1(i, j) is that coordinate is (i, j) on gray level image
Point gray value, w be mean value weight；T in this example1It is the value of thresh when w takes 0.6；Character zone is substantially clear at this time
It is clear, but background is vanished from sight substantially, only a small number of dotted stains.
(1.3) line direction, column direction projection are carried out to above-mentioned first image Img1, obtains the second image Img2, as slightly
The crown word number image of positioning；The step specifically:
(1.3.1) obtains projecting to obtain array RowSum [i], (i to the first image Img1 progress line direction by projection
=1,2,3 ..., 50), wherein i be crown word number image row number；
It carries out column direction to the first image Img1 to project to obtain array ColSum [j], (j=1,2,3 ..., 250)；Wherein j
It is transverse projection, the longitudinal projection of Img1 respectively referring to Fig. 6,7 for the column number of crown word number image；
(1.3.2) determines coboundary, lower boundary, left margin and the right margin of the first image Img1；It is specific as follows:
It searches for from top to bottom, by meeting condition " array RowSum [i]+RowSum [i+1] > T21And RowSum [t] >
T31" point constitute crown word number image coboundary Yu；The coboundary of Yu behavior character zone in image, the coboundary are one
Straight line；T in example21Take 30, T31Take 15；
Constantly search downwards, until occurring being unsatisfactory for " array RowSum [i]+RowSum [i+1] > T21And RowSum [t] >
T31" point, the lower boundary Yd of crown word number image is determined as by these straight lines for constituting of point；
If meeting Yd-Yu > length, Yu is determined for coboundary, Yd is lower boundary；Wherein length refers to character
Pixels tall；
If being unsatisfactory for Yd-Yu > length, abandon and continue searching, until meeting Yd-Yu > length, or searches
Image bottom is still unsatisfactory for, then determines that there is no crown word numbers in image, quote exception；
It searches for from left to right, finds the condition of satisfaction " array ColSum [t]+ColSum [t+1] > T22And ColSum [t] >
T32" point constitute crown word number image left margin Xt；
It turns left search from the right side, finds the condition of satisfaction " array ColSum [t]+ColSum [t-1] > T22And ColSum [t] >
T32" point constitute crown word number image right margin Xr；In embodiment, T22Take 6, T32Take 3.
(1.3.3) finds the upper left corner A (Xt, Yu), the lower right corner B of crown word number of crown word number by positioning the region of crown word number
(Xr,Yd)；The subgraph comprising crown word number image is found in former gray level image Img from upper left corner C (Xt-10, Yu-10) to the right side
The region of inferior horn D (Xr10, Yd-10) constitutes accurate crown word number image, i.e. the second image Img2.
(2) integrogram is calculated to the second image Img2 and calculates the Haar-Like H feature in integrogram, including to second
Image Img2 calculates the step of integrogram, calculates the step of Haar-Like H feature in integrogram；
The step of calculating integrogram to the second image Img2 is as follows:
(2.1) indicated in integrogram from the upper left corner of the second image to the cumulative of the gray value of point (x, y) with A (x, y) and；
Wherein, I2(i, j) be on the second image coordinate be (i, j) point gray value.
(2.2) a two-dimensional array Arr [H] [W] is created for integrogram, and all initialized；Wherein H is the second image
The height of Img2, W are the width of the second image Img2；
(2.3) it is calculate by the following formula the integrogram of the second image Img2 the first row,
Arr [0] [x]=Arr [0] [x-1]+I [0] [x]；
Wherein I [0] [x] indicates that the coordinate of the second image Img2 is the gray value of (x, 0)；
(2.4) integrogram is all calculated to all the points on the second image Img2 by following formula, obtains complete integrogram
Arr；
Arr [y] [x]=Arr [y] [x-1]+Arr [y-1] [x]-Arr [y-1] [x-1]+I [y] [x].
The step of calculating Haar-Like H feature by integrogram Arr and obtain set of coordinates C:
(2.5) the Haar-Like H feature of suitable image is determined according to crown word number feature；Dollar prefix is directed in embodiment
Number, the crown word number character boundary selected is 22*14, and two Haar-Like H features of use are respectively as follows: H1Size 40*38,
For positioning the position of letter with crown word number, H2Size 40*32, for positioning the position of similar character.
(2.6) by carrying out H to the second image Img21Feature and H2The traversal of feature obtains one group of coordinate, and record is wherein full
Sufficient H1>T4、H2>T4Coordinate (xi, yi), i=1,2 ..., k }；
Wherein k is Haar-Like number of the condition that meets；
(2.7) since the Haar-Like H feature of crown word number image is front and back connection, by the above-mentioned coordinate for meeting condition
In the peak value of Haar-Like H form one group of coordinate C={ (xi, yi), i=1,2 ..., 9 }, wherein m is prefix sign character
Number.
(3) specific as follows to the second image Img2 progress anti-binary conversion treatment of Low threshold to calculate confidence matrix:
(3.1) Low threshold T is used to the second image Img25Anti- binaryzation is carried out, the only comprising prefix sign character is obtained
Background texture is not included in three image Img3, third image Img3, character is more complete；
Wherein, second threshold
M is the line number of the second image, is the columns that 50, N is the second image in embodiment, is 250 in embodiment；I(i,j)
Be on the second image coordinate be (i, j) point gray value, i=1,2,3 ..., 50, j=1,2,3 ..., 250；A is weight, real
Apply is 0.5 in example；Obtained image is referring to Fig. 8.
(3.2) confidence matrix M [C] [D] is calculated, wherein C is the line number of third image Img3, and D is third image Img3's
Columns；
To confidence matrix initialisation, the value of initialization is identical as the value of third image Img3；
Formula specific as follows:
M [i] [j]=Img3 [i] [j], i ∈ { 1,2 ..., C }, j ∈ { 1,2 ..., D }；Img3 [i] [j] is to refer to third
Image Img3 coordinate (i, j) value.
To the confidence matrix of initialization into line scans, the point that M [i] [j] value be 0 but front and back existence value is 1 then will
Its corresponding value M [i] [j] is set to X, thus obtains the confidence matrix scanned through space.
Column scan is carried out to the confidence matrix scanned through space, the point that M [i] [j] value be 0 but existence value is 1 up and down,
Then X will be set to by value M [i] [j] accordingly, and obtain the confidence matrix M by rank scanning.
(4) confidence matrix based on step (2) set of coordinates C obtained and step (3) acquisition obtains final crown word number image
BwImg, specific as follows:
(4.1) anti-binaryzation is carried out using high threshold T6 to the second image Img2, obtains the 4th image Img4, referring to figure
9；By set of coordinates C combination Haar-Like H feature, the non-character regional value of the 4th image Img4 is set to 0, i.e. removal word
Background texture between symbol obtains the 5th image Img5, and referring to Fig.1 0.In this step, threshold value T6Using acquisition T in step (1.2)1's
Method, difference are that w takes 0.8, it is therefore intended that character in image is allowed to keep complete；W can according to the specific feature of image of processing come
It determines.
(4.2) the confidence matrix M obtained using step (3) does mask process to the 5th image Img5, removes crown word number word
Texture in symbol,
I.e. as M [i] [j]=X, the point for being 1 to respective coordinates in the 5th image Img5 then retains, as M [i] [j]=0,
The value of 5th image Img5 respective coordinates (i, j) is directly set 0, the crown word number bwImg extracted mentions for final crown word number
It takes as a result, referring to Fig.1 1.
The above-mentioned crown word number extracting method based on integrogram and multi threshold fusion that embodiment mentions, using Haar-Like H
Feature and multi threshold fusion have carried out effective removal to crown word number background image；For line present in crown word number background patterns
Reason problem, stain problem are attained by good treatment effect, and can be adjusted according to background feature, scalability
By force, be able to satisfy crown word number extract require, can effectively solve ticket intelligent recognition machine adapt to it is a variety of under the conditions of crown word number extract
Problem achievees the purpose that accurately identify crown word number, and feature is simple, and calculating speed is fast, and robustness is good.
As it will be easily appreciated by one skilled in the art that the foregoing is merely illustrative of the preferred embodiments of the present invention, not to
The limitation present invention, any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should all include
Within protection scope of the present invention.
Claims (8)
1. a kind of multi threshold fusion crown word number extracting method based on integrogram, which comprises the steps of:
(1) White-light image including crown word number image one side of object to be identified is acquired, and is normalized and is included
The gray level image of crown word number；
(2) binary conversion treatment is carried out using first threshold to the gray level image to obtain comprising imperfect crown word number but not comprising back
First image of scape obtains the second image by row projection and column projection to the first image；
(3) integrogram is calculated to the second image, and calculates the Haar-Like H feature in integrogram；
(4) the anti-binary conversion treatment of second threshold is carried out to the second image, obtains third image only comprising prefix sign character, according to
The third image, which calculates, obtains confidence matrix；
(5) anti-binary conversion treatment is carried out using third threshold value to the second image, obtains the 4th image；Utilize the Haar-Like
The background texture that H characteristic coordinates remove in the 4th image between prefix sign character obtains the 5th image；
(6) mask process is done to the 5th image using the confidence matrix, removes the texture in prefix sign character, obtained clear
Clear crown word number image.
2. multi threshold fusion crown word number extracting method as described in claim 1, which is characterized in that the step (1) includes: pair
Normalized image carries out luminance compensation and obtains the denomination information of image, positions to obtain comprising crown word number according to the denomination information
Gray level image.
3. multi threshold fusion crown word number extracting method as described in claim 1, which is characterized in that the step (2) obtains the
The method of two images includes following sub-step:
(2.1) it projects to obtain line number group RowSum [i] by carrying out line direction to the first image, wherein i is the row of the first image
Number；
Column direction is carried out to the first image to project to obtain columns group ColSum [j]；Wherein j is the column number of the first image；
(2.2) coboundary, lower boundary, left margin and the right margin of the first image are determined；It is specific as follows:
It searches for from top to bottom, finds the condition of satisfaction " array RowSum [i]+RowSum [i+1] > T21And RowSum [t] > T31"
The line being made of these points is determined as the coboundary Yu of the first image by point；
Constantly search downwards, until occurring being unsatisfactory for " array RowSum [i]+RowSum [i+1] > T21And RowSum [t] > T31"
The line being made of these points is determined as the lower boundary Yd of the first image by point；
It searches for from left to right, finds the condition of satisfaction " array ColSum [t]+ColSum [t+1] > T22And ColSum [t] > T32"
The line being made of these points is determined as the left margin Xt of the first image by point；
It turns left search from the right side, finds the condition of satisfaction " array ColSum [t]+ColSum [t-1] > T22And ColSum [t] > T32"
The line being made of these points is determined as the right margin Xr of the first image by point；
Wherein, T21Refer to the first row boundary threshold, T31Refer to the second row bound threshold value, T22Refer to first row boundary threshold, T32It is
Refer to secondary series boundary threshold；
(2.3) upper left corner A (Xt, Yu) of the crown word number and lower right corner B of crown word number is found by positioning the region of crown word number
(Xr,Yd)；The subgraph comprising crown word number image is found in former gray level image from upper left corner C (Xt-10, Yu-10) to bottom right
The region of angle D (Xr-10, Yd-10) constitutes accurate crown word number image, as the second image.
4. multi threshold fusion crown word number extracting method as claimed in claim 3, which is characterized in that the first threshold T=
average*w；
The gray average of whole picture gray level image
Wherein, I1(i, j) is that coordinate is the gray value of the point of (i, j) on gray level image, and w is the weight of mean value.
5. multi threshold fusion crown word number extracting method as claimed in claim 1 or 2, which is characterized in that the step (3) includes
Following sub-step:
(3.1) indicated in integrogram from the upper left corner of the second image to the cumulative of the gray value of point (x, y) with A (x, y) and；
Wherein, I2(i, j) be on the second image coordinate be (i, j) point gray value；
(3.2) a two-dimensional array Arr [H] [W] is created for integrogram, and all initialized；Wherein H is the height of the second image
Degree, W are the width of the second image；
(3.3) it is calculate by the following formula the integrogram of the second image the first row,
Arr [0] [x]=Arr [0] [x-1]+I [0] [x]；
Wherein I [0] [x] indicates that the coordinate of the second image Img2 is the gray value of the point of (x, 0)；
(3.4) integrogram is all calculated to all the points on the second image Img2 by following formula, obtains complete integrogram Arr；
Arr [y] [x]=Arr [y] [x-1]+Arr [y-1] [x]-Arr [y-1] [x-1]+I [y] [x].
6. multi threshold fusion crown word number extracting method as claimed in claim 1 or 2, which is characterized in that the step (4) includes
Following sub-step:
(4.1) second threshold T is used to the second image5Anti- binary conversion treatment is carried out, third figure only comprising prefix sign character is obtained
Picture；
Second threshold
Wherein, M is the line number of the second image, and N is the columns of the second image；I (i, j) is that coordinate is (i, j) on the second image
The gray value of point, a is weight；
(4.2) confidence matrix M [C] [D] is calculated, wherein C is the line number of third image, and D is the columns of third image；
To the confidence matrix of initialization into line scans, the point that M [i] [j] value be 0 but front and back existence value is 1, then by its phase
The value M [i] [j] answered is set to X, thus obtains the confidence matrix scanned through space；
Column scan is carried out to the confidence matrix scanned through space, the point that M [i] [j] value be 0 but existence value is 1 up and down then will
Corresponding value M [i] [j] is set to X, obtains the confidence matrix M scanned by row, column.
7. multi threshold fusion crown word number extracting method as claimed in claim 1 or 2, which is characterized in that the step (5) are as follows:
Anti- binary conversion treatment is carried out using third threshold value to the second image and obtains the 4th image；By Haar-Like H feature by the 4th
The non-character regional value of image is set to 0, removes the background texture in the 4th image between prefix sign character and obtains the 5th image, is
Final crown word number extracts result.
8. multi threshold fusion crown word number extracting method as claimed in claim 7, which is characterized in that right as M [i] [j]=X
The point that respective coordinates are 1 in 5th image then retains；It is as M [i] [j]=0, the value of the 5th image respective coordinates (i, j) is direct
0 is set, the crown word number image extracted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810574416.2A CN109086766B (en) | 2018-06-06 | 2018-06-06 | Multi-threshold fusion crown word number extraction method based on integral graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810574416.2A CN109086766B (en) | 2018-06-06 | 2018-06-06 | Multi-threshold fusion crown word number extraction method based on integral graph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109086766A true CN109086766A (en) | 2018-12-25 |
| CN109086766B CN109086766B (en) | 2021-03-09 |
Family
ID=64839413
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810574416.2A Active CN109086766B (en) | 2018-06-06 | 2018-06-06 | Multi-threshold fusion crown word number extraction method based on integral graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109086766B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112766249A (en) * | 2019-11-04 | 2021-05-07 | 深圳怡化电脑股份有限公司 | Bill boundary positioning method and device and terminal equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521911A (en) * | 2011-12-16 | 2012-06-27 | 尤新革 | Identification method of crown word number (serial number) of bank note |
| CN102750771A (en) * | 2012-07-13 | 2012-10-24 | 中山大学 | Method for identifying denominations of fifth series of renminbi applied in smart phone |
| CN107590900A (en) * | 2017-09-15 | 2018-01-16 | 深圳怡化电脑股份有限公司 | A kind of crown word number image acquiring method, device, equipment and storage medium |
| CN105354566B (en) * | 2015-10-26 | 2019-01-22 | 江苏国光信息产业股份有限公司 | Character for crown word number identification pollutes minimizing technology |
-
2018
- 2018-06-06 CN CN201810574416.2A patent/CN109086766B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521911A (en) * | 2011-12-16 | 2012-06-27 | 尤新革 | Identification method of crown word number (serial number) of bank note |
| CN102750771A (en) * | 2012-07-13 | 2012-10-24 | 中山大学 | Method for identifying denominations of fifth series of renminbi applied in smart phone |
| CN105354566B (en) * | 2015-10-26 | 2019-01-22 | 江苏国光信息产业股份有限公司 | Character for crown word number identification pollutes minimizing technology |
| CN107590900A (en) * | 2017-09-15 | 2018-01-16 | 深圳怡化电脑股份有限公司 | A kind of crown word number image acquiring method, device, equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 窦艳芳: ""复杂背景下纸币号码识别技术的研究"", 《中国优秀硕士学位论文全文数据库 (电子期刊) 信息科技辑》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112766249A (en) * | 2019-11-04 | 2021-05-07 | 深圳怡化电脑股份有限公司 | Bill boundary positioning method and device and terminal equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109086766B (en) | 2021-03-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105930791B (en) | The pavement marking recognition methods of multi-cam fusion based on DS evidence theory | |
| Babu et al. | Vehicle number plate detection and recognition using bounding box method | |
| CN104751142B (en) | A kind of natural scene Method for text detection based on stroke feature | |
| CN103679118B (en) | A kind of human face in-vivo detection method and system | |
| CN102750540B (en) | Morphological filtering enhancement-based maximally stable extremal region (MSER) video text detection method | |
| CN106056751B (en) | The recognition methods and system of serial number | |
| CN101923640B (en) | Method for distinguishing false iris images based on robust texture features and machine learning | |
| CN105956578A (en) | Face verification method based on identity document information | |
| TW201039248A (en) | Method and system for identifying image and outputting identification result | |
| CN104318238A (en) | Method for extracting crown word numbers from scanned banknote images in banknote detection module | |
| CN102542655B (en) | Note anti-counterfeiting discrimination method based on fiber personality characteristics | |
| CN107464252A (en) | A kind of visible ray based on composite character and infrared heterologous image-recognizing method | |
| CN106875546A (en) | A kind of recognition methods of VAT invoice | |
| CN106845542B (en) | Paper money number intelligent identification Method based on DSP | |
| CN109426814A (en) | A kind of positioning of the specific plate of invoice picture, recognition methods, system, equipment | |
| CN102364496A (en) | Method and system for identifying automobile license plates automatically based on image analysis | |
| CN107122775A (en) | A kind of Android mobile phone identity card character identifying method of feature based matching | |
| CN106529534A (en) | Variable-length license plate character segmentation method based on hybrid tilt correction and projection method | |
| Centeno et al. | Recurrent comparator with attention models to detect counterfeit documents | |
| CN109658440A (en) | A kind of method for tracking target based on target significant characteristics | |
| Akbani et al. | Character recognition in natural scene images | |
| CN107742357A (en) | A kind of recognition methods of paper money number and device | |
| CN109086766A (en) | A kind of multi threshold fusion crown word number extracting method based on integrogram | |
| CN106447910A (en) | Paper money identifying method and apparatus | |
| CN106355743A (en) | Banknote version identification method and 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address |
Address after: 430000 room 04, floor 3, building 3, phase III, international enterprise center, No. 1, Guanggu Avenue, Donghu New Technology Development Zone, Wuhan, Hubei 3602 (Wuhan area of free trade zone) Patentee after: Wuhan Huake heding Information Technology Co.,Ltd. Address before: No.99-74, Taizhong Road, Zijin (Pukou) science and technology entrepreneurship special community, Nanjing, Jiangsu, 210000 Patentee before: NANJING HUAKE HEDING INFORMATION TECHNOLOGY Co.,Ltd. |
|
| CP03 | Change of name, title or address |