CN107944252A - A kind of method of information seal impression uniqueness characteristic extraction - Google Patents
A kind of method of information seal impression uniqueness characteristic extraction Download PDFInfo
- Publication number
- CN107944252A CN107944252A CN201711367747.0A CN201711367747A CN107944252A CN 107944252 A CN107944252 A CN 107944252A CN 201711367747 A CN201711367747 A CN 201711367747A CN 107944252 A CN107944252 A CN 107944252A
- Authority
- CN
- China
- Prior art keywords
- image
- printed text
- point
- feature point
- mrow
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
- G06F21/31—User authentication
- G06F21/36—User authentication by graphic or iconic representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/64—Protecting data integrity, e.g. using checksums, certificates or signatures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Security & Cryptography (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computer Hardware Design (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Evolutionary Computation (AREA)
- Bioethics (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Image Processing (AREA)
- Collating Specific Patterns (AREA)
Abstract
The present invention proposes a kind of method of printed text uniqueness characteristic extraction so that the printed text from same seal can be also resolved.This method comprises the following steps：(1) printed text Image Acquisition：Printed text image is obtained by image acquisition equipment, after analog signal sample quantization, view data is stored in the matrix form, generates Direction array；(2) printed text image preprocessing：The Direction array image of generation is pre-processed, local noise interference is eliminated, makes that print images are clear, edge is obvious；(3) encrypted feature point information extraction：According to the location of encrypted feature point in image, size, gray scale accounting and duty cycle this four elements, validity feature point is differentiated, wherein duty cycle characterization oozes print degree；The value of the minutia parameter, i.e. this four elements of validity feature point is extracted, constructs two dimensional character code matrix, is the uniqueness characteristic of the printed text.
Description
Technical field
The present invention relates to a kind of feature extracting method of information seal.
Background technology
Printed text：Refer in particular to physical seal when not illustrating and affix one's seal using ink paste be formed on information carrier (such as paper)
Seal word and graphical information " trace ".
Chinese patent ZL 20112008576.7 (seal for carrying coding information) discloses a kind of novel physical seal, I
Be referred to as information seal, i.e., printed text pattern on seal face includes legacy identification word and the embodiment encoding of graphs letter of setting
The pattern of breath (outline border of seal body working face is provided with annular code area).Affixed one's seal on the carrier using ink paste in seal
Show legacy identification word and embody the pattern of encoding of graphs information.Scanning obtains these encoding of graphs information can be with Equity Investors
Established and contacted by network, to express the actual wishes of Equity Investors.Patent application CN201410692609X (is based on network communication
With the authentication seal information processing system of image recognition) give the application scheme of the information seal, patent application CN
2014108092073 (a kind of information seals and its printed text image information processing method) give the information extraction and identification of optimization
Method.
With the development of computer technology, information seal be able to may also be engraved a seal by being turned over easily according to printed text, and the true and false is difficult
Distinguish.Therefore, patent application CN2016108740484 (a kind of encrypted information seal and its encryption method) carries out such seal
Anti-counterfeiting design.
But, for the confirmation of sealed files validity, whether conventional physical seal or above- mentioned information seal, only
It is true only to confirm to impress, and is also inadequate sometimes.Such as by taking the power of attorney as an example, if other people illegally obtain seal in the power of attorney
On impress, then although it is true (rather than imitation) to impress, it is apparent that being invalid.For another example, donor is misled into, and is running counter to
It is in the case of its subjective desire to have covered a power of attorney, then this power of attorney should also be invalid more.
In technical field of software development, in order to ensure the software of software download was not changed by people or was broken in download
It is bad, can verify that (MD5) technology is confirmed by file fingerprint at present, i.e., by certain algorithm, to file data information into
Row verification computing, has drawn the hexadecimal number (verification and) of one 32.As long as slightly being changed since original information has, lead to
After crossing md5 computings, as a result can all there is very big change, so as to ensure the integrality of file data and security.
It is common to consider perhaps just if simply using for reference above-mentioned file fingerprint verification (MD5) technology for information seal
It is to increase single check code (encoding of graphs) in the code area of information seal, is solved after the electronic pictures information of printed text is obtained
Analysis obtains check information.It will be clear that this scheme is still no fundamentally to solve the problems, such as sealed files validation.
The content of the invention
The present invention is based on encryption information seal, proposes a kind of method of printed text uniqueness characteristic extraction so that from same
The printed text of seal can be also resolved.On this basis, the present invention can realize the quick confirmation of sealed files validity.
The present invention inventive concept be：
According to the principle in terms of ichnology, seal is during impressing, due to dynamics, angle, material and background of impressing
Etc. difference, the random difference of microcosmic upper mark can be formed, its reason is similar to " there is no identical two feet in the world
Print ".Normally impress operation, all there are fine distinction with other printed texts for the printed text obtained each time.Rely on seal design
Uniqueness and mark randomness, we are impressed image by scanning, and extraction is affixed one's seal " uniqueness " mark to be formed every time
Feature, mark feature (hereinafter referred to as：Printed text fingerprint) with impressing, file is mapped, establish " fingerprint " shelves for file of impressing
Case, so as to identify printed text and the uniqueness for file of impressing with " printed text fingerprint ".
The present invention is proposed based on patent application CN2016108740484 (a kind of encrypted information seal and its encryption method)
Encryption information seal, with ichnology technology to printed text image carry out printed text fingerprint feature information extraction so that printed text refers to
Unique correspondence is formed between line and printed text.
Technical scheme is specific as follows：
A kind of method of information seal impression uniqueness characteristic extraction, printed text are impressed from above encryption information seal,
This method comprises the following steps：
(1) printed text Image Acquisition；
Printed text image is obtained by image acquisition equipment, after analog signal sample quantization, stores image in the matrix form
Data, generate Direction array；
(2) printed text image preprocessing；
The Direction array image of generation is pre-processed, local noise interference is eliminated, makes that print images are clear, edge is bright
It is aobvious；
(3) encrypted feature point information extraction；
According to the location of encrypted feature point in image, size, gray scale accounting and duty cycle this four elements, differentiation has
Characteristic point is imitated, wherein duty cycle characterization oozes print degree；Extract the minutia parameter of validity feature point, i.e. this four elements
Value, construct two dimensional character code matrix, be the printed text uniqueness characteristic.
On the basis of above scheme, the present invention has further made following optimization：
Step (3) first carries out step in detail below before validity feature point is differentiated：
3.1) pass point region detection
By the grey scale change amount of detection image, the region where encrypted feature point and the angle of each encrypted feature point are defined
Spend position；
3.2) pass point projection localization
Using the center of circle of seal image as subpoint, ray projection is outwards carried out, projection angle is since 0 degree, to 360 degree
Cut-off, sampling form Gray Projection band, position of the encrypted feature point in Gray Projection band are then calculated, according to this position
Put and a region cutting is encrypted, formation sequence encryption dot image；
3.3) pass point fitting and interpolation processing
The profile of encrypted feature point is found using least square fitting, and is compensated and is fitted by bilinear interpolation algorithm
Point pixel, obtains the characteristic point boundary profile of continuous continuously smooth.
The minutia parameter：Position is angle coordinate of the validity feature dot center position on mark circumference；Gray scale
Accounting is all grey value profile accountings in validity feature point fitted area；Size is characterized a little in the area of mark annular region
Accounting；The statistical nature that print degree uses gradient is oozed, Grad is higher, and representative oozes that print degree is low, and image is more clear, otherwise oozes print
Degree is higher.
Step 3.1) pass point region detection, using Canny algorithms.
Step 3.1) pass point region detection, is specially：
Pretreated view data is subjected to gaussian filtering first：
fs(x, y)=f (x, y) * G (x, y)
F (x, y) represents pretreated view data, and G (x, y) represents two-dimensional Gaussian function, namely convolution operation number, fs
The image of (x, y) for convolution after smooth；σ is the variance of setting；
Then following Canny algorithms are performed：
3.1.1 single order local derviation finite difference formulations gradient) is used
Make difference coefficient (the approximate substitution difference quotients of Δ f/ Δs x)The change rate of gray scale is sought, is taken respectively adjacent on x and y directions
Pixel makes the difference, instead of asking for x and y directions single order local derviation；
3.1.2) magnitude image is carried out to apply non-maxima suppression
By angular divisions into four direction scope:Horizontal (0 °), -45 °, vertical (90 °) ,+45 °；If central point is (i.e.：Visit
Ask a little) along its direction the gradient magnitude of neighborhood it is maximum, then retain；Otherwise, suppress；
3.1.3) detection of dual threashold value-based algorithm and connection edge
Choose high threshold TH and Low threshold TL；
The greatest gradient amplitude in the image after non-maxima suppression is taken out, redefines high-low threshold value；
Point less than TL is abandoned, assigns 0；The point that will be greater than TH marks immediately, these points are marginal point, assign 1；
It will be greater than TL, o'clock determined less than TH using 8 connected regions, i.e.,：It can just be received when being only connected with TH pixels,
As marginal point, 1 is assigned.
In step 3.2), a Projection Sampling is carried out every 0.1 degree, sampling ray amounts to 3600；Corresponding gray scale is thrown
In shadow bands horizontal direction, 0.1 degree is represented per pixel.
Based on the method for above- mentioned information seal impression uniqueness characteristic extraction, the present invention and then provide one kind file of impressing and have
The method that effect property confirms：By obtaining the uniqueness characteristic of printed text on file of impressing, printed text fingerprint is denoted as；By the printed text fingerprint with
The printed text fingerprint file that information seal Equity Investors pre-establish carries out lookup comparison, if failing to find in printed text fingerprint file
The record being consistent, then judging this to impress, file is invalid；It is conversely, then effective.
The present invention has the following advantages：
The present invention, which can realize, distinguishes the printed text from same seal.And then the present invention can realize that sealed files are effective
Property quick confirmation.
Brief description of the drawings
Fig. 1 is the schematic diagram of Difference Calculation gradient in Canny algorithms.
Fig. 2 is the schematic diagram of pass point area detection result.
Fig. 3 is the schematic diagram that Gray Projection band is formed in pass point projection localization link.
Fig. 4 obtains the schematic diagram of sequential encryption dot image for pass point projection localization link.
Fig. 5 is the schematic diagram for oozing print degree.
Embodiment
The present invention is described in detail with reference to the accompanying drawings and examples.
The extraction of printed text fingerprint is a comprehensive problem for being related to the technologies such as image procossing, pattern match, ichnology.By
The difference for condition of impressing in seal, includes the complexity of background, impress firmly size, ink paste ooze print etc. all can be to seal
Image has an impact so that the image obtained from same seal also has random difference.Present invention is generally directed to add information to compile
The seal of code and encrypted feature point, by mobile phone terminal image acquisition equipment, the seal designs on file of affixing one's seal are taken pictures to obtain
Image, for this image, extracts the characteristic point on seal trace with a series of image processing techniques, forms unique mark
" fingerprint " of printed text.
During file is affixed one's seal each time, size of the encrypted feature point in printed text, ooze print degree, gray scale accounting all has
There is certain randomness, these random characters just constitute printed text fingerprint.
The extraction process of printed text fingerprint mainly includes printed text Image Acquisition, printed text image preprocessing, encrypted feature point information
Extract three main process.Determining whether the factor of validity feature point includes the seal annulus position residing for characteristic point, feature
Size, gray scale accounting and the duty cycle four elements of point.Detail characteristic value (position, size, ash are extracted for validity feature point
Degree accounting, ooze print degree), a two dimensional character code matrix is constructed, this matrix is exactly unique printed text fingerprint of the printed text.When having
When effect characteristic point is less than the 30% of total characteristic point, it is considered as the failure of printed text fingerprint extraction.
1. printed text Image Acquisition
Taken pictures by image acquisition equipments such as mobile phone, scanner, high photographing instruments to region with an official seal affixed in printed text, shape
Into printed text image, the object as printed text fingerprint extraction.
2. printed text image preprocessing
Being oozed the reasons such as print, seal and paper material by firmly size, ink paste due to the process of impressing is influenced, and is made to be analyzed
Printed text picture noise is more and has to characteristic point compared with strong jamming, the extraction of effect characteristics information.Printed text image is by that will simulate
After signal sampling quantifies, computer is stored in the matrix form, after generating Direction array, in order to which the local noise for eliminating stronger is done
Disturb, it is necessary to be pre-processed to the Direction array image of generation.Pretreatment is the premise of printed text fingerprint extraction, and whole work
Basis.
Printed text image preprocessing is the noise removed in printed text image, improves the quality of print images, makes print images clear
It is clear, edge is obvious, in order to improve extraction and store characteristic point accuracy rate.Including region detection of impressing, picture quality judge,
Image enhancement, projection mapping etc..
After being pre-processed to printed text image, more clearly encrypted feature projection zone, region bag can be obtained
Contain encrypted feature point in information seal.Due to the interference of noise, these characteristic points in shape, quantity, position, ooze print degree etc.
Aspect has certain random character.
3. encrypted feature point extracts
" printed text fingerprint " feature extraction：The characteristic point of printed text image is extracted.Due to passing through pretreated refinement
There are substantial amounts of pseudo-random numbers generation on image, the presence of these pseudo-random numbers generations, not only substantially reduces matched speed, also makes " print
Literary fingerprint " performance drastically declines, and causes the rising for refusing rate and misclassification rate by mistake of identifying system.Therefore, should before being extracted
Pseudo-random numbers generation is removed as far as possible, containing substantial amounts of pseudo-random numbers generation for the printed text fingerprint details extracted, this is asked
Topic, it is proposed that a kind of profile information and ooze print degree diagnostic method, effectively removes pseudo-random numbers generation, it will be apparent that reduce pseudo-characteristic
Point.
The basic step of encryption information point extraction is divided into：Pass point region detection, pass point projection localization, pass point fitting
And interpolation processing, characteristic point differentiates and information extraction.
3.1 pass point region detections
Region detection is mainly the monitoring and positioning of the grey scale change amount of image, its essence just extracts discontinuous portion in image
Point feature, region detection the result is that image Segmentation Technology relies on key character.
Before pass point region detection, first using gaussian filtering smoothed image：
Make f (x, y) represent data (input source data), G (x, y) represent two-dimensional Gaussian function (convolution operation number), fs (x,
y)
For image of the convolution after smooth.
fs(x, y)=f (x, y) * G (x, y)
The value of variances sigma=0.8 is set, each coordinate points (x, y) of correspondence are brought into dimensional Gaussian formula G (x, y), are obtained
One weight matrix, normalized weight matrix (the sum of each point divided by weight in matrix), obtains the weight matrix of standard, i.e., high
This template.
Template is set as 3 × 3 matrixes, the weight of corresponding points is multiplied by with the gray value of each pixel, 9 obtained values are asked
With, be exactly central point Gaussian Blur value.
The present invention is as follows using Canny algorithms detection pass point region, specific method：
(1) single order local derviation finite difference formulations gradient is used
Make difference coefficient (the approximate substitution difference quotients of Δ f/ Δs x)The change rate of gray scale is sought, is taken respectively adjacent on x and y directions
Pixel makes the difference, instead of asking for x and y directions single order local derviation.
(2) magnitude image is carried out applying non-maxima suppression
By angular divisions into four direction scope:Horizontal (0 °), -45 °, vertical (90 °) ,+45 °.If central point is (i.e.：Visit
Ask a little) along its direction the gradient magnitude of neighborhood it is maximum, then retain；Otherwise, suppress.
(3) detection of dual threashold value-based algorithm and connection edge
Choose high threshold TH and Low threshold TL, ratio 2:1 or 3:1.(generally taking TH=0.3/0.2, TL=0.1)
The greatest gradient amplitude in the image after non-maxima suppression is taken out, redefines high-low threshold value.I.e.：TH×Max,
TL×Max.(given it is of course possible to oneself)
Point less than TL is abandoned, assigns 0；The point that will be greater than TH marks (these points are exactly marginal point) immediately, assigns 1.
TL is will be greater than, less than o'clock being determined (i.e. using 8 connected regions for TH：It can just be received when being only connected with TH pixels,
As marginal point, assign 1).
After carrying out region detection by Canny algorithms, the encrypted feature point region in mark can be good at being defined place
Reason.As shown in Figure 2.
3.2 pass point projection localizations
What the present invention took is method of projection on ray, i.e., using the center of circle of seal image as subpoint, outwards carries out ray throwing
Shadow.Projection angle to 360 degree of cut-offs, a Projection Sampling is carried out every 0.1 degree, sampling ray amounts to 3600 since 0 degree
Bar.Due to after pretreatment, image gray processing, so the figure that projection is formed is Gray Projection band, as shown in Figure 3.
In horizontal direction, 0.1 degree is represented per pixel.
In 3.1 encryption section detection process, the angle position of each pass point according to horizontal direction it has been determined that there be not picture
Position of the pass point in projection zone is calculated in 0.1 degree of element, and a region cutting is encrypted according to this position.Formation sequence
Dot image is encrypted, as shown in Figure 4.
3.3 pass points are fitted and interpolation processing
The present invention finds the profile of pass point using least square fitting.The algorithm is by minimizing square of error
With the optimal function matching for finding data.
minb∣∣Ab-Y∣∣2,A∈Cn×m,Y∈Cn
The particular solution of above formula is the generalized inverse matrix of A and the product of Y, this is also the solution of two least norms at the same time, its general solution is
Particular solution adds the kernel of A, i.e. A+Y is the solution of least square method.Oval with least square fitting, for ellipse, its is general
Equation is
Ax+Bxy+Cy2+ Dx+Ey=1
We have some coordinate (x put(i), y(i)), it is desirable to ask for its parameter
P=[A, B, C, D, E]TSo that coordinate is brought into equation, meets equation as far as possible, i.e., its desired output is 1.
Point coordinates is extended to eigenmatrix X, the i-th row there are 5 row, is respectively
((x(i))2,x(i)y(i),(y(i))2,x(i),y(i)), it would be desirable to output is write as b=[1,1 ... 1]T
The immediate elliptic parameter that b vectors can be asked by formula.
If Constrained, it is desirable to which known point is within all ellipses tried to achieve, then requires the estimation of all the points to export
Less than 1, then an initial parameter P can be tried to achieve using a upper formula, then filter out all asking using parameter p
The point of oval outside is obtained, then reconstructs X matrix using these external points, re-uses a upper formula to try to achieve new ginseng
Count, this process of Reusability, the point outside ellipse can be fewer and fewer, it is known that the point outside ellipse is seldom.The ellipse so obtained, both
In the outside of all known points, and it is most to be fitted all known points.
Interpolation is to increase a kind of method of image pixel size in the case where not generating pixel, main in the present invention to use
In compensation match point pixel.Interpolation algorithm make use of the correlation of four pixels around original image pixels point to be treated
Property, it is calculated by bilinear algorithm.For a purpose coordinate, it is obtained in original image by reflection method backward
Corresponding floating-point coordinate (i+u, j+ ν), wherein i, j are nonnegative integer, and u, ν are the floating number in [0,1] section, then this pixel
Value f (i+u, j+ ν) can be (i, j) by coordinate in original image, (i+1, j), (i, j+1), corresponding to (i+1, j+1) around four
The value decision of a pixel, i.e.,：
F (i+u, j+ ν)=(1-u) × (1- ν) × f (i, j)+(1-u) × ν × f (i, j+1)+u × (1- ν) × f (i+1,
j)+u×ν×f(i+1,j+1)
Wherein, f (i, j) represents the pixel value at source images (i, j) place, and so on.Interpolation algorithm can be described as follows：
(1) size of new images is obtained by original image and scale factor, and creates new images.
(2) original image (x ', y ') place is mapped to by some pixel (x, y) of new images.
(3) (xx, yy) is obtained to (x ', y ') rounding and is obtained
Arrive, (xx, yy), (xx+1, yy), the value of (xx, yy+1) and (xx+1, yy+1).
(4) obtain the value of pixel (x, y) using bilinear interpolation and write back new images.
(5) repeat step (2) is write until all pixels of new images.
3.4 characteristic points differentiate and information extraction
Each characteristic point has specific position, size, shape, angle of inclination in design.During impressing, by
In the interference of noise, the information of characteristic point has lost, can also produce some pseudo-random numbers generations.Before characteristic point information is extracted
First have to carry out the rejecting processing of pseudo-random numbers generation.
The Rule of judgment of pseudo-random numbers generation includes：Seal annulus position residing for characteristic point, the size of characteristic point, gray scale accounting
And duty cycle this four elements.This four elements is weighted summation scoring according to 0.4,0.1,0.2,0.3 weights respectively, always
It is considered as pseudo-random numbers generation to divide the point less than 80 points.Seal theory characteristic point has 68, when validity feature point is less than 30%, depending on
Extract and fail for " printed text fingerprint ".
" printed text fingerprint " is extraction minutia parameter, constitutive characteristic value sequence, as printed text in validity feature point
Unique features code.Minutia includes：Position, size, gray scale accounting, ooze four information of print degree.
Position：Angle coordinate of the characteristic point regional center position on mark circumference.
Gray scale accounting：All grey value profile accountings in characteristic point fitted area.
Size：Area accounting of the characteristic point in mark annular region.
Ooze print degree：Characteristic point is theoretically elliptic region, but due to oozing print, the region is by three parts structure
Into, and outer boundary, ooze print transition region, characteristic point inner spare area.Different print degree reflections of oozing have not to print transition region is oozed
Same width.As shown in figure 5, print degree is oozed from left to right from large to small.The main method that print degree is oozed in measurement is exactly gradient
Statistical nature, usual Grad is higher, and picture oozes that print degree is low, and image is more clear, otherwise it is higher to ooze print degree.
Each validity feature point includes 4 characteristic details values above, and all characteristic points form a two dimensional character code square
Battle array, this matrix is exactly unique " the printed text fingerprint " of printed text.
Table 1
| Validity feature point sequence number | Position | Size | Gray scale accounting | Ooze print degree |
| 1 | 10 | 0.1 | 23 | 0.12 |
| 2 | 12.5 | 0.12 | 35 | 0.15 |
| 5 | 45 | 0.09 | 36 | 0.21 |
| 6 | 126.5 | 0.13 | 26 | 0.15 |
| … | ||||
| n | 345 | 0.11 | 45 | 0.18 |
With the example that upper table 1 is " printed text fingerprint ".
Claims (7)
1. a kind of method of information seal impression uniqueness characteristic extraction, the printed text are impressed from encryption information seal, its
It is characterized in that, comprises the following steps：
(1) printed text Image Acquisition；
Printed text image is obtained by image acquisition equipment, after analog signal sample quantization, stores view data in the matrix form,
Generate Direction array；
(2) printed text image preprocessing；
The Direction array image of generation is pre-processed, local noise interference is eliminated, makes that print images are clear, edge is obvious；
(3) encrypted feature point information extraction；
According to the location of encrypted feature point in image, size, gray scale accounting and duty cycle this four elements, differentiate effectively special
Point is levied, wherein duty cycle characterization oozes print degree；The minutia parameter of extraction validity feature point, i.e. this four elements
Value, constructs two dimensional character code matrix, is the uniqueness characteristic of the printed text.
2. the method for information seal impression uniqueness characteristic extraction according to claim 1, it is characterised in that：Step (3)
Before validity feature point is differentiated, step in detail below is first carried out：
3.1) pass point region detection
By the grey scale change amount of detection image, the region where encrypted feature point and the angle position of each encrypted feature point are defined
Put；
3.2) pass point projection localization
Using the center of circle of seal image as subpoint, ray projection is outwards carried out, projection angle is cut since 0 degree to 360 degree
Only, sampling forms Gray Projection band, position of the encrypted feature point in Gray Projection band is then calculated, according to this position
A region cutting, formation sequence encryption dot image is encrypted；
3.3) pass point fitting and interpolation processing
The profile of encrypted feature point is found using least square fitting, and match point picture is compensated by bilinear interpolation algorithm
Element, obtains the characteristic point boundary profile of continuous continuously smooth.
3. the method for information seal impression uniqueness characteristic extraction according to claim 2, it is characterised in that：The details
Characteristic parameter：Position is angle coordinate of the validity feature dot center position on mark circumference；Gray scale accounting is validity feature point
All grey value profile accountings in fitted area；Size is characterized a little in the area accounting of mark annular region；Print degree is oozed to adopt
With the statistical nature of gradient, Grad is higher, and representative oozes that print degree is low, and image is more clear, otherwise it is higher to ooze print degree.
4. the method for information seal impression uniqueness characteristic extraction according to claim 2, it is characterised in that：Step 3.1)
Pass point region detection, using Canny algorithms.
5. the method for information seal impression uniqueness characteristic extraction according to claim 4, it is characterised in that：Step 3.1)
Pass point region detection, is specially：
Pretreated view data is subjected to gaussian filtering first：
<mrow>
<mi>G</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mn>2</mn>
<msup>
<mi>&pi;&sigma;</mi>
<mn>2</mn>
</msup>
</mrow>
</mfrac>
<msup>
<mi>e</mi>
<mfrac>
<mrow>
<mo>-</mo>
<mrow>
<mo>(</mo>
<msup>
<mi>x</mi>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mi>y</mi>
<mn>2</mn>
</msup>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mn>2</mn>
<msup>
<mi>&sigma;</mi>
<mn>2</mn>
</msup>
</mrow>
</mfrac>
</msup>
</mrow>
fs(x, y)=f (x, y) * G (x, y)
The pretreated view data of f (x, y) expressions, G (x, y) expression two-dimensional Gaussian functions, namely convolution operation number, fs (x,
Y) image for convolution after smooth；σ is the variance of setting；
Then following Canny algorithms are performed：
3.1.1 single order local derviation finite difference formulations gradient) is used
Make difference coefficient (the approximate substitution difference quotients of Δ f/ Δs x)The change rate of gray scale is sought, takes adjacent pixel on x and y directions to do respectively
Difference, instead of asking for x and y directions single order local derviation；
3.1.2) magnitude image is carried out to apply non-maxima suppression
By angular divisions into four direction scope:Horizontal (0 °), -45 °, vertical (90 °) ,+45 °；If central point is (i.e.：Accessing points)
It is maximum in the gradient magnitude of the neighborhood along its direction, then retain；Otherwise, suppress；
3.1.3) detection of dual threashold value-based algorithm and connection edge
Choose high threshold TH and Low threshold TL；
The greatest gradient amplitude in the image after non-maxima suppression is taken out, redefines high-low threshold value；
Point less than TL is abandoned, assigns 0；The point that will be greater than TH marks immediately, these points are marginal point, assign 1；
It will be greater than TL, o'clock determined less than TH using 8 connected regions, i.e.,：It can just be received when being only connected with TH pixels, become
Marginal point, assigns 1.
6. the method for information seal impression uniqueness characteristic extraction according to claim 2, it is characterised in that：
In step 3.2), a Projection Sampling is carried out every 0.1 degree, sampling ray amounts to 3600；Corresponding Gray Projection band
In horizontal direction, 0.1 degree is represented per pixel.
A kind of 7. method for file availability confirmation of impressing, it is characterised in that：Using the information seal impression described in claim 1
The method of uniqueness characteristic extraction, obtains the uniqueness characteristic of printed text on file of impressing, is denoted as printed text fingerprint；By the printed text fingerprint
The printed text fingerprint file pre-established with information seal Equity Investors carries out lookup and compares, if failing to search in printed text fingerprint file
To the record being consistent, then judging this to impress, file is invalid；It is conversely, then effective.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711367747.0A CN107944252B (en) | 2017-12-18 | 2017-12-18 | Method for extracting uniqueness characteristic of seal impression of information seal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711367747.0A CN107944252B (en) | 2017-12-18 | 2017-12-18 | Method for extracting uniqueness characteristic of seal impression of information seal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107944252A true CN107944252A (en) | 2018-04-20 |
| CN107944252B CN107944252B (en) | 2021-07-06 |
Family
ID=61941877
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711367747.0A Active CN107944252B (en) | 2017-12-18 | 2017-12-18 | Method for extracting uniqueness characteristic of seal impression of information seal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107944252B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109299605A (en) * | 2018-11-19 | 2019-02-01 | 珠海实跃科技有限公司 | A kind of application method and system of seal protector |
| CN110287936A (en) * | 2019-07-02 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Image detecting method, device, equipment and storage medium |
| CN110717157A (en) * | 2019-09-06 | 2020-01-21 | 平安科技（深圳）有限公司 | User authentication method, device, computer equipment and storage medium |
| CN114387194A (en) * | 2021-12-15 | 2022-04-22 | 北京达佳互联信息技术有限公司 | Image processing method, image processing apparatus, electronic device, medium, and program product |
| CN116645591A (en) * | 2023-05-31 | 2023-08-25 | 杭州数盒魔方科技有限公司 | Pixel value-based electronic contract seal picture PS trace identification method and system |
| TWI841997B (en) * | 2021-07-26 | 2024-05-11 | 大陸商矩網科技有限公司 | Encryption seal and strip data secondary encryption method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102009016429A1 (en) * | 2009-04-04 | 2010-10-07 | Kempchen Dichtungstechnik Gmbh | Static flat seal i.e. flange seal, has optically readable marking designed as barcode and directly formed on material of seal, where barcode is formed in two-dimension with matrix-like sample and corresponds to character string |
| CN101894260A (en) * | 2010-06-04 | 2010-11-24 | 北京化工大学 | Method for identifying forgery seal based on feature line randomly generated by matching feature points |
| CN206114546U (en) * | 2016-09-19 | 2017-04-19 | 南京航空航天大学 | Printing defect vision detection system |
-
2017
- 2017-12-18 CN CN201711367747.0A patent/CN107944252B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102009016429A1 (en) * | 2009-04-04 | 2010-10-07 | Kempchen Dichtungstechnik Gmbh | Static flat seal i.e. flange seal, has optically readable marking designed as barcode and directly formed on material of seal, where barcode is formed in two-dimension with matrix-like sample and corresponds to character string |
| CN101894260A (en) * | 2010-06-04 | 2010-11-24 | 北京化工大学 | Method for identifying forgery seal based on feature line randomly generated by matching feature points |
| CN206114546U (en) * | 2016-09-19 | 2017-04-19 | 南京航空航天大学 | Printing defect vision detection system |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109299605A (en) * | 2018-11-19 | 2019-02-01 | 珠海实跃科技有限公司 | A kind of application method and system of seal protector |
| CN110287936A (en) * | 2019-07-02 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Image detecting method, device, equipment and storage medium |
| CN110287936B (en) * | 2019-07-02 | 2021-08-10 | 北京字节跳动网络技术有限公司 | Image detection method, device, equipment and storage medium |
| CN110717157A (en) * | 2019-09-06 | 2020-01-21 | 平安科技（深圳）有限公司 | User authentication method, device, computer equipment and storage medium |
| CN110717157B (en) * | 2019-09-06 | 2023-10-24 | 平安科技（深圳）有限公司 | User authentication method, device, computer equipment and storage medium |
| TWI841997B (en) * | 2021-07-26 | 2024-05-11 | 大陸商矩網科技有限公司 | Encryption seal and strip data secondary encryption method |
| CN114387194A (en) * | 2021-12-15 | 2022-04-22 | 北京达佳互联信息技术有限公司 | Image processing method, image processing apparatus, electronic device, medium, and program product |
| CN116645591A (en) * | 2023-05-31 | 2023-08-25 | 杭州数盒魔方科技有限公司 | Pixel value-based electronic contract seal picture PS trace identification method and system |
| CN116645591B (en) * | 2023-05-31 | 2024-01-05 | 杭州数盒魔方科技有限公司 | Pixel value-based electronic contract seal picture PS trace identification method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107944252B (en) | 2021-07-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107944252A (en) | A kind of method of information seal impression uniqueness characteristic extraction | |
| US10410309B2 (en) | Classification and authentication of identification documents using a convolutional neural network | |
| Ferrer et al. | Robustness of offline signature verification based on gray level features | |
| CN108345880A (en) | Invoice recognition methods, device, computer equipment and storage medium | |
| Otori et al. | Texture synthesis for mobile data communications | |
| US20190377970A1 (en) | Digital image generation through an active lighting system | |
| CN104063706A (en) | Video fingerprint extraction method based on SURF algorithm | |
| CN101894260A (en) | Method for identifying forgery seal based on feature line randomly generated by matching feature points | |
| CN104680130A (en) | Chinese character recognition method for identification cards | |
| CN106096601A (en) | The method and system of character types in a kind of automatic detection bill | |
| CN108509992A (en) | A kind of seal impression Weigh sensor method for anti-counterfeit based on machine learning techniques | |
| CN108509870A (en) | A kind of Eriocheir sinensis uniqueness recognition methods based on images match | |
| CN107180422A (en) | A kind of labeling damage testing method based on bag of words feature | |
| CN106326887A (en) | Method and device for checking optical character recognition result | |
| CN110059722A (en) | Checking method, device, equipment and the readable storage medium storing program for executing of seal image | |
| CN107038445A (en) | A kind of binaryzation and dividing method for Chinese character identifying code | |
| Das et al. | Multi‐script versus single‐script scenarios in automatic off‐line signature verification | |
| CN109492696A (en) | Seal authenticity identification method, apparatus and system | |
| CN106778498A (en) | A kind of method for strengthening Fingerprint recognition | |
| CN105404868A (en) | Interaction platform based method for rapidly detecting text in complex background | |
| CN105430218A (en) | Image color recognition method and device | |
| Ferrer et al. | Signature verification using local directional pattern (LDP) | |
| Pu et al. | Fractional-order spatial steganography and blind steganalysis for printed matter: Anti-counterfeiting for product external packing in Internet-of-Things | |
| Ferrer et al. | Realistic synthetic off-line signature generation based on synthetic on-line data | |
| CN110533704B (en) | Method, device, equipment and medium for identifying and verifying ink label |
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 |