CN111189784A - Method for identifying stamping signature sequence based on polarization spectrum imaging - Google Patents

Method for identifying stamping signature sequence based on polarization spectrum imaging Download PDF

Info

Publication number
CN111189784A
CN111189784A CN202010035340.3A CN202010035340A CN111189784A CN 111189784 A CN111189784 A CN 111189784A CN 202010035340 A CN202010035340 A CN 202010035340A CN 111189784 A CN111189784 A CN 111189784A
Authority
CN
China
Prior art keywords
polarization
image
characteristic
area
images
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
Application number
CN202010035340.3A
Other languages
Chinese (zh)
Inventor
张淼
牛思聪
冯迦炆
沈毅
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Harbin Institute of Technology
Original Assignee
Harbin Institute of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Harbin Institute of Technology filed Critical Harbin Institute of Technology
Priority to CN202010035340.3A priority Critical patent/CN111189784A/en
Publication of CN111189784A publication Critical patent/CN111189784A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/21—Polarisation-affecting properties
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/255—Details, e.g. use of specially adapted sources, lighting or optical systems
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/27—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands using photo-electric detection ; circuits for computing concentration
    • 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
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
    • 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
    • G06V30/22—Character recognition characterised by the type of writing
    • G06V30/226—Character recognition characterised by the type of writing of cursive writing
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N2021/1734—Sequential different kinds of measurements; Combining two or more methods
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N2021/1765—Method using an image detector and processing of image signal

Abstract

A stamping signature sequence identification method based on polarization spectrum imaging relates to the field of spectrum information detection, and solves the problems that the existing method is serious in information loss, greatly influenced by environment and poor in judgment accuracy. The method comprises the following steps: firstly, collecting a polarization hyperspectral image of an original sample; secondly, calculating Stokes parameters to obtain corresponding polarization parameter images; thirdly, fusing the obtained polarization degree and polarization angle images to obtain polarization characteristic images and corresponding polarization characteristic vectors; fourthly, carrying out image blocking processing on the polarization intensity image to obtain a characteristic region and a corresponding light intensity characteristic vector; and fifthly, corresponding the characteristic region to the polarization characteristic image, and performing correlation analysis on the characteristic vector of the polarization characteristic image so as to determine the sequence of stamping and signing. The method utilizes the comprehensiveness of the sample polarization spectrum information and combines the characteristic vector correlation analysis to judge the sequence, and is suitable for the sequence identification application of the materials similar to the ink inkpad.

Description

Method for identifying stamping signature sequence based on polarization spectrum imaging
Technical Field
The invention relates to the technical field of spectral imaging and polarization imaging, in particular to a stamping signature sequence identification method based on polarization spectral imaging.
Background
Handwriting is the reflection of writing action as the main expression form of handwritten character symbols. The sign pen is one of the indispensable external conditions for completing writing activities, has the advantages of fluent writing, convenience in carrying, low price and the like as the most popular writing tool in the world at present, and particularly relates to the identification of signature handwriting for most document evidences in various criminal and civil cases. With the increasing of economic crimes and civil dispute cases, the requirements for signature handwriting recognition are more and more, and particularly, the requirements for the identification of the sequence of handwriting and stamping inkpad are also increased gradually. The thickness of the ink or the ink paste on the paper is very thin, but the thickness of the ink or the ink paste can penetrate into the paper, so that the actual effective thickness of the ink or the ink paste is expanded, and the thickness of the ink or the ink paste can be influenced by the thickness of a pen point, which brings great difficulty to the relevant handwriting authentication case.
The polarized hyperspectral imaging technology is applied to the field of document inspection. The spectral imaging technology combines a spectral technology and an imaging method, wherein the imaging provides the intensity of each pixel of an image, the spectral image provides the spectrum of each pixel, and therefore a continuous and narrow-band 3D data set is generated, and two-dimensional geometric space and one-dimensional spectral information of a shot object are detected. The polarization hyperspectral technology organically combines the spectrum and the polarization characteristic, and can acquire the spectrum, the intensity and the polarization characteristic of a certain pixel serving as an object on the basis of the traditional spectrum technology. The existing polarization technology uses full-band polarization, and does not consider the influence of different single bands on polarization, and the polarization characteristics are influenced by wavelength, so that a polarization spectrum can be formed, and related information of an object is also contained in the polarization characteristics.
By adopting polarization hyperspectral detection, more abundant information about the target and the scene can be obtained compared with the traditional intensity image and spectrum image, the identification of the disguised target and the abnormal detection of the tiny target are realized, and the identification capability of the target is improved. Therefore, the invention provides an algorithm for sequentially identifying the handwriting and the stamping inkpad of the sign pen based on the polarization spectrum imaging technology, the distribution of the handwriting and the stamping inkpad in the longitudinal space can be rapidly and accurately detected, and the deep structural information of the handwriting and the stamping inkpad can be mined.
Disclosure of Invention
The invention aims to provide a method for identifying the sequence of stamping and signing based on polarization spectrum imaging, so that the problems of serious information loss, great environmental influence and poor judgment accuracy existing in the conventional method are solved.
The purpose of the invention is realized by the following technical scheme: firstly, controlling a polaroid to perform angular rotation twice in a polarization hyperspectral sampling system to perform shooting at three angles, acquiring a hyperspectral image in a red light range by using a hyperspectral polarization camera at each angle, and repeating the process until the hyperspectral polarized images at the three angles are acquired; processing three polarization images with different angles acquired by a single waveband to obtain corresponding polarization Stokes parameters and polarization parameter images; fusing the polarization angle image and the polarization degree image with the polarization characteristics by adopting a polarization energy fusion method to obtain a polarization characteristic image and polarization characteristic vectors of all pixel points; then, image blocking is carried out on the polarization intensity image by using a dynamic threshold method based on local region segmentation, and a target region is selected according to the gray level change degree of sub-blocks so as to remove regions with overlarge information fluctuation and excessively uneven material distribution and obtain a characteristic region with gentle gray level change and corresponding light intensity characteristic vectors of all pixel points; and finally, corresponding the screened characteristic regions to a polarization characteristic image to perform characteristic vector fusion and correlation analysis, and determining the sequence of the ink inkpad according to the correlation strength among the regions.
The flow chart of the invention is shown in figure 1, and the specific steps are as follows:
the method comprises the following steps: to the originalCollecting images of the initial sample: the method comprises the steps of controlling a polaroid to rotate twice in angle in a polarization hyperspectral sampling system, shooting at three angles of 0 degree, 60 degrees and 120 degrees (0 degree is non-rotation shooting), and collecting red light band images by using a CMOS (complementary metal oxide semiconductor) camera (or CCD) at each angle
Figure BDA0002365662800000022
(lambda represents a wave band, theta represents rotation angles of 0 degrees, 60 degrees and 120 degrees by 0 degree, 60 degrees and 120 degrees respectively), and the process is repeated until the target polarization hyperspectral image is acquired.
In the photographing process, firstly, a light source irradiates an original sample at a certain angle, light reflected by the original sample penetrates through a polaroid and enters a spectral filtering device to screen the wavelength, then, the light is imaged in a CMOS (complementary metal oxide semiconductor) (or charge coupled device) camera, and an image is read in a computer. The light source, the sample and the polaroid form a certain angle to enhance the polarization imaging effect, and the white laser light source is selected to reduce the influence of fluorescence carried by inkpad and ink of a sign pen on the experimental result in the experimental process according to the bleaching effect generated by laser; because the red inkpad and the black ink are used in the invention, a red spectrum wave band is selected for imaging; in addition, in the experimental process, the polaroid is rotated to enable the lens to generate displacement, errors can be generated in the image acquisition process, and the polaroid is rotated twice for measuring images with three angles, so that the method has a positive effect of reducing the errors caused by the lens displacement.
Step two: three different-angle polarization hyperspectral images acquired by each wave band
Figure BDA0002365662800000021
And processing to obtain corresponding polarization Stokes parameters and polarization parameter images.
In the application field of polarization detection technology, the stokes vector S is often used to analyze various polarization states of a plane wave. Stokes vector S ═ I, Q, U, V]TWherein I represents the total radiance of the light, Q represents the difference in the amount of linear polarization in the vertical direction and the horizontal direction, U is the difference in the amount of linear polarization in the + -45 deg. direction, and V represents the elliptical polarization,v is a very small value that can be detected by a detector and the present invention uses linearly polarized light, so this variable is ignored in the present invention. Corresponding parameter calculation is carried out on the image obtained by sampling the polarization hyperspectral system, and a Stokes vector corresponding to the image can be obtained.
In the collection of the polarization hyperspectral image, the light intensity I collected by a cameraλ,aThe relationship with the polarizer rotation angle θ is:
Figure BDA0002365662800000031
wherein Iλ,j、Qλ,j、Uλ,jIs a Stokes vector [ I, Q, U, V]TThe first three parameters, λ wavelength, and j pixel point location (j)1，j2) The polarization data cube of a certain pixel point of the polarization image under the single waveband can be expressed as [ I ]λ,i,Qλ,i,Uλ,i,DoLPλ,Orientλ]TWhen θ is 0 °, 60 °, or 120 °
Figure BDA0002365662800000032
Figure BDA0002365662800000033
Figure BDA0002365662800000034
Based on the Stokes vectors obtained above, the polarization Angle (Angle of polarization) and Degree of polarization (Degree of polarization) parameters can be obtained by the following algorithm, respectively using Aλ,jAnd Dλ,jRepresents:
Figure BDA0002365662800000035
Figure BDA0002365662800000036
in the polarization parameter obtained, Iλ,jSpectral information (also called polarization intensity) representing a polarized hyperspectral image, and a degree of polarization Dλ,jAnd angle of polarization Aλ,jThe polarization characteristics of the image are mainly reflected, the polarization degree image reflects rich texture, edge and other information of the object, and the polarization angle image describes the state characteristics and other information of the target object and the background where the target is located. However, the collection of the polarization and hyperspectral images is easily influenced by uncertain factors such as changes of surrounding environments and the like, so that the collected polarization hyperspectral images are poor in imaging effect and low in quality, and therefore difficulty is caused in distinguishing detailed features of targets, especially in the aspect of handwriting identification, the identification of ink-inkpad cross areas is particularly caused. In order to improve the quality of the acquired images, the characteristics of the polarization parameter images are considered, and the polarization degree images are subjected to fusion processing to achieve the purposes of weakening the influence of the surrounding environment on imaging and highlighting the polarization characteristics of the target.
Step three: the polarization angle image and the polarization degree image in the polarization parameter image are subjected to image fusion processing to obtain the polarization characteristic image of the sample on each waveband and the polarization characteristic vector G of each pixel pointj。
Because the gray value of the pixels in the rough area is smaller than that of the pixels in the smooth area in the polarization degree image, the polarization energy corresponding to the two areas has great difference, and based on the energy characteristics of the polarization degree image, the polarization energy weighted fusion algorithm shown as follows is adopted in a single waveband:
Figure BDA0002365662800000041
Figure BDA0002365662800000042
Figure BDA0002365662800000043
wherein, Fλ,DoLP(n)，Fλ,Orient(N) each represent the gray scale value of a single pixel N over the wavelength band λ, and M N represents the range of regions of the same selected location in the polarization degree and polarization angle images.
In order to more obviously express the polarization degree difference of rough and smooth areas in the polarization degree image, the invention is used for Dλ,jThe normalized fusion processing is performed, and the algorithm is as follows:
Figure BDA0002365662800000044
through the fusion process, the polarization characteristic image of the sample on each waveband and the polarization characteristic value G of each pixel point on the single waveband lambda in the embodiment of the invention can be obtainedλ,jFurther obtain the polarization characteristic vector G of each pixel point on the full wave band used by the inventionj，Gλ,jAnd GjThe relationship is as follows:
Gj＝[Gmin,j,Gmin+1，j,…,Gλ,j,…,Gmax-1,j,Gmax,j]T
wherein min and max represent the minimum and maximum values of the target band, respectively.
Step four: and (3) partitioning the image of the polarization intensity image by adopting a dynamic threshold method based on local region segmentation, and selecting a target region according to the severe condition of gray level change in the partitioned block so as to remove regions with too severe information fluctuation and too uneven material distribution and obtain corresponding characteristic regions and light intensity characteristic vectors of all pixel points.
The basic idea of the dynamic threshold method based on local region segmentation is to divide a target image into blocks correspondingly, determine the threshold values of different regions by using the sliding of a local window, and judge the gray level change condition of the regions by comparing the sub-block threshold values with the window threshold values. The method comprises the following specific steps:
step 4-1: when the target local window slides in the target area, different image sub-blocks in the calculation area fall into the window surfaceRatio q of productsiThe corresponding calculation formula is:
qi＝mi/m
where m denotes the total number of pixels in the sliding window, miRepresenting the number of pixels in the target area where the ith sub-block falls in the sliding window;
step 4-2: the threshold passing proportion q corresponding to each sub-blockiAfter weighting and summing, a global threshold value can be obtained, and the calculation method is as follows:
Figure BDA0002365662800000051
wherein n represents the number of image sub-blocks falling into the local window, K1iThe calculation formula is the average absolute deviation corresponding to the ith image sub-block:
Figure BDA0002365662800000052
wherein F (r) represents the gray value of the r-th pixel point in the sub-block, mukAnd expressing the gray average value of the pixel points in the sub-block, and R multiplied by T expresses the size of the sub-block.
Step 4-3: similarly, the average absolute deviation epsilon of the pixels in the local sliding window can be calculated;
step 4-4: by comparing the mean absolute deviation epsilon of the local sliding window pixels with a global threshold K2To determine the change of the pixel gray level in the local window area if ε < α K2If the gray scale change is flat, otherwise, the gray scale change is severe, wherein α represents a comparison coefficient;
and 4-5: obtaining a characteristic area with gentle gray scale change in the target area and a light intensity characteristic vector H corresponding to each pixel point in the area after the step is finishedjThe representation method is as follows:
Hj＝[Imin,j,Imin+1,j,…,Iλ,j,…,Imax-1j,Imaxj]T
wherein min and max represent the minimum and maximum values of the target band, respectively.
Step five: corresponding the obtained gray level change flat area to a polarization characteristic image, and enabling the pixel point polarization characteristic vector G generated in the step three to bejAnd step four, generating a pixel point light intensity characteristic vector HjCarrying out pixel-by-pixel combination to obtain a combined characteristic vector OjJudging the characteristic vectors O of the pixel points of the inkpad area, the ink area and the mixed area of the inkpad area and the ink area by using Kendall Rank (Kendall Rank) correlation coefficientsjThe correlation between the inkpad and the ink handwriting is judged according to the strength of the consistency level correlation.
The Kendel grade correlation coefficient is a statistical method for measuring the correlation between two variables, and has two vectors Oj1,Oj2The number of the contained elements is N, and the ith values of the two variables are respectively used
Figure BDA0002365662800000053
And (4) showing. O isj1And Oj2The corresponding elements in the set form an element pair set Oj12From corresponding elements
Figure BDA0002365662800000061
And (4) forming. When set Oj12Any two of
Figure BDA0002365662800000062
And
Figure BDA0002365662800000063
the two elements are considered to be identical when they are arranged in the same order. If it is
Figure BDA0002365662800000064
And is
Figure BDA0002365662800000065
Or
Figure BDA0002365662800000066
And is
Figure BDA0002365662800000067
When two elements are not identical. When it occurs
Figure BDA0002365662800000068
Or
Figure BDA0002365662800000069
Then the two elements are neither inconsistent nor consistent. The specific calculation formula is as follows:
Figure BDA00023656628000000610
wherein W represents Oj12The number of consistency element pairs owned; y represents Oj12The number of inconsistent element pairs is obtained.
The value range of the Kendel correlation coefficient is between-1 and 1, and when the value of tau is-1, the two variables have completely opposite level correlation; when tau is 1, the two variables have completely consistent level correlation; when tau is between 0 and 1, if tau is in the range of 0.8 to 1.0, the two variables have extremely strong consistent grade correlation, the two variables have strong consistent grade correlation between 0.6 and 0.8, the two variables have medium consistent grade correlation between 0.4 and 0.6, the consistent grade correlation between 0.2 and 0.4 is weak, and the consistent grade correlation or the irrelevant is extremely weak between 0 and 0.2; when τ is 0, it indicates that the two variables are independent of each other. In the invention, the Kendel correlation coefficients of the feature vectors are weighted, the consistency level correlation strength is judged according to the weighted result, and the sequence of the ink and the inkpad is finally determined.
Compared with the prior art, the invention has the following advantages:
the invention adopts the consistent correlation strength of the polarization hyperspectral characteristic vectors of all the substances to judge the sequence of the inkpad and the ink handwriting, and no similar method is adopted for identifying the sequence of the handwriting at present. The method comprises the steps of obtaining information representing handwriting as much as possible by extracting the polarization hyperspectral characteristics of the handwriting, greatly reducing the influence of environmental change on information acquisition more possibly by polarization parameter image fusion, ensuring the accuracy of the obtained information, avoiding the influence of deviation caused by the intermiscibility of substances and greatly reducing the misjudgment rate.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
FIG. 2 is a sample graph used in an embodiment of the present invention.
FIG. 3 is a polarization diagram of an embodiment of an imaging system acquired at different angles of a single band; in the figure, (a) is the embodiment pixel point at 0 ° polarization image; (b) the pixel point is a 60-degree polarized image for the embodiment; (c) the pixel point is at 120 ° polarization image for the example.
FIG. 4 is a diagram of polarization parameters for a single band in accordance with an embodiment of the present invention; in the figure, (a) to (e) respectively correspond to a polarized light intensity image, a linear polarization horizontal and vertical difference image, a linear polarization plus-minus 45-degree difference image, a polarization degree image and a polarization angle image of the pixel point of the embodiment.
FIG. 5 is a polarization signature diagram of an embodiment of the present invention.
FIG. 6 is a polarization intensity grid partitioning diagram based on dynamic thresholding for local region segmentation.
Detailed Description
Embodiments of the present invention are further described below with reference to the accompanying drawings and specific examples. The following embodiments illustrate specific embodiments of the present invention, using morning light MG-6139 ink and an i-shaped plate 683 inkpad with a diameter of 7.5cm as examples:
the original sample picture is made by using the sequence of writing with the ink I and the inkpad seal, as shown in figure 2, the upper left corner is signed first and then sealed, namely the inkpad is on the ink, and the lower right corner is signed first and then sealed, namely the inkpad is on the ink.
Executing the step one: use ofThe white light source irradiates an original sample, the spectral distribution of the white light source is closest to that of natural light, the polarization intensity of light in each direction is uniform at light speed, the requirement on the invention principle is met, the included angle between incident light and reflected light is larger than 90 degrees in the light source irradiation process, and the polarization information of the reflected light is obvious under the angle. Then, the polaroid is controlled to rotate for two times in the polarization hyperspectral sampling system, shooting is carried out at three angles of 0 degree, 60 degree and 120 degree respectively in the red light wave band (615 and 650nm, the adjusting step length is 1nm) of the light source light, and a hyperspectral polarization camera is used for collecting images of the infrared wave band at each angle
Figure BDA0002365662800000071
(lambda represents a wave band, theta represents a rotation angle and represents 0 degrees, 60 degrees and 120 degrees by 0, 60 and 120 degrees), taking a pixel point j with a 630nm wavelength coordinate (452 and 205) as an example, as shown in fig. 3, repeating the image acquisition process until the target polarization hyperspectral image is acquired.
And (5) executing the step two: according to the polarization image data I collected by the cameraλ,a(0°)，Iλ,a(60°)，Iλ,a(120 degree), corresponding Stokes parameters [ I, Q, U, V ] are solved]TAnd degree of polarization AλAngle of polarization DλThe invention does not need to calculate the elliptical polarization V, and the calculation method of other parameters comprises the following steps:
Figure BDA0002365662800000072
Figure BDA0002365662800000073
Figure BDA0002365662800000074
at 630nm pixel point (452,205) in the example of the invention, I630,(452，205)＝0.442029，Q630,(452,205)＝0.745831，U630,(452,205)0.800000, the corresponding information of polarization degree and angle can be obtained from the above parameters：
Figure BDA0002365662800000075
Figure BDA0002365662800000081
Corresponding to available D630,(452，205)＝0.590203，A630,(452,205)0.133045, and the corresponding polarization parameter image is obtained, as shown in fig. 4.
And step three is executed: in order to better analyze the polarization information of the image, the polarization degree image and the polarization angle image are fused to obtain a polarization characteristic image, and in order to ensure the uniform distribution of the polarization information of the image, the polarization characteristic value of the image is obtained by adopting a polarization energy weighted fusion method. The specific calculation method is as follows:
Figure BDA0002365662800000082
Figure BDA0002365662800000083
Figure BDA0002365662800000084
Figure BDA0002365662800000085
where the value of M × N is 2 × 2, the polarization characteristic value G of the sample point can be obtained by the above calculation630,(452,205)Calculating all pixel points to obtain polarization characteristic vector G of the pixel point (0.375901)jAnd polarization signature images over a single wavelength band, as shown in fig. 5.
And step four is executed: image blocking is carried out on the polarized intensity image by adopting a dynamic threshold method based on local region segmentation, a target region is selected according to the drastic situation of gray level change in the molecular block, and if the gray level in the region is changed, the target region is selectedIf the gray scale change of the pixel value is biased to be flat, the area is taken as a target area, otherwise, the area is marked as a background area. In the method, the change condition of the pixel gray level in the local window area is judged by comparing the average deviation with the window threshold value, so that whether the gray level change of the pixel value in the area is biased to be flat or severe can be reflected, and the area with flat change is screened out as a target area. In the calculation of this step, 2 × 2 image blocks are selected as a single sub-block, the image sub-block regions are divided, as shown in fig. 6, 4 image blocks with the same size are selected as a sliding window, epsilon is selected to be 0.5 as a comparison coefficient, and the average absolute deviation of the pixels in the local sliding window is compared with a global threshold through corresponding calculation, so as to obtain a feature region with flat gray scale change and a light intensity feature vector H of each pixel in the feature regionj。
And executing the step five: in the polarization characteristic image, corresponding characteristic areas, namely a pure inkpad area, a pure ink area, an inkpad-ink-first area and an inkpad-ink-first area are selected, the method for respectively calculating Kendall-grade correlation coefficients of pixels among the areas and carrying out weighted comparison is adopted, the sequence of the inkpad and the ink handwriting is judged according to the relation among the correlation coefficients, and the result of the correlation coefficients is shown in a table 1.
From Table 1, it can be seen that the correlation coefficient between the inkpad-first and inkpad-second areas and the pure ink area is 0.6204, between 0.6 and 0.8, indicating that the two areas have strong consistency grade correlation, the correlation coefficient with the pure ink area is 0.0064, the magnitude of the coefficient is between 0 and 0.2, indicating that the consistency grade correlation between the two areas is weak, the correlation coefficient between the inkpad-first and inkpad-second areas and the pure ink area is 0.6635, between 0.6 and 0.8, indicating that the two areas have strong consistency grade correlation, the correlation coefficient with the pure ink is 0.0095, between 0 and 0.2, indicating that the consistency grade correlation between the two areas is weak, and from the above analysis, it can be determined that if the correlation coefficient of the mixed area and a pure substance area is above 0.6, and the correlation coefficient with another pure substance area is between 0 and 0.2, the mixed area has high substance order, the materials with low correlation are arranged below, and for the sequence, the materials with low correlation are arranged in front of the materials with high correlation, so that the sequence of the inkpad and the ink handwriting can be accurately judged.
TABLE 1 correlation coefficient of each feature region pixel
Figure BDA0002365662800000091

Claims (6)

1. A stamping signature sequence identification method based on polarization spectrum imaging is characterized by comprising the following steps:
the method comprises the following steps: shooting an original sample by using a polarization hyperspectral imaging system which is self-assembled in a laboratory, controlling a polaroid to rotate for two angles, shooting at three angles, collecting a hyperspectral image in a red light range by using a hyperspectral polarization camera at each angle, and repeating the process until the hyperspectral polarization images at the three angles are completely collected;
step two: respectively calculating Stokes parameters of the collected sample polarization spectrum images on a single wave band, and calculating according to the image parameters of three angles to obtain light intensity Iλ,jAnd the difference parameter Q of linear polarization quantity in the horizontal and vertical directionsλ,jLinear polarization quantity difference parameter U in positive and negative 45 DEG angle directionsλ,jPolarization degree parameter Dλ,jAnd a polarization angle parameter Aλ,j5 characteristic parameters and five corresponding polarization parameter images are obtained;
step three: further processing the parameters of the polarization degree image to balance the problem of overlarge difference between the polarization degree parameters of a smooth area and a rough area in the image, and performing polarization energy weighted fusion on the obtained polarization degree and polarization angle image to obtain polarization characteristic images on all wave bands and polarization characteristic vectors of pixel points;
step four: image blocking is carried out on the polarization intensity image, and a target area is selected to remove an area with overlarge information fluctuation and excessively uneven material distribution, so that a characteristic area meeting requirements and corresponding light intensity characteristic vectors of all pixel points can be obtained;
step five: and corresponding the screened areas to a polarization characteristic image to perform characteristic vector fusion and correlation analysis, and determining the sequence of the ink inkpad according to the correlation strength among the areas.
2. The method for identifying the stamping signature sequence based on polarization spectrum imaging as claimed in claim 1, wherein the first step specifically comprises:
shooting a target sample by using a polarization hyperspectral imaging system consisting of an LED light source, a polaroid, LCTF, an intelligent camera and a computer, wherein the image comprises different information of inkpad and ink; the spectrum imaging system based on the liquid crystal tunable optical filter has the advantages of fast imaging, uniform light intensity distribution, less stray light, integration of preprocessing function, more suitability for target detection and acquisition and convenience for spectrum data acquisition in the aspect of medical images; the intelligent camera adopts a CMOS (or CCD) image sensor, and the camera is an embedded system and can execute the functions of picture preprocessing and analysis, so that the imaging speed of the microscopic hyperspectral imaging system is high, the intelligent camera can operate in a free operation state or a fixed frame rate mode when acquiring images, and the images obtained by shooting contain original polarization information under different wavelengths of red light wave bands.
3. The method for identifying the stamping signature sequence based on the polarization spectrum imaging as claimed in claim 1, wherein the second step specifically comprises:
the Stokes parameters are used for representing the polarization information of the image, and according to the polarization image parameters collected by the camera, 5 characteristic parameters of light intensity, linear polarization quantity difference parameters in the horizontal and vertical directions, linear polarization quantity difference parameters in the positive and negative 45-degree directions, polarization degree and polarization angle are calculated, and five corresponding polarization parameter images are obtained.
4. The method for identifying the stamping signature sequence based on the polarization spectrum imaging as claimed in claim 1, wherein the third step specifically comprises:
and finally, solving polarization characteristic values through the two types of polarization energy and the corresponding parameters to obtain a polarization parameter vector of a single pixel point and a fusion image.
5. The method for identifying the stamping signature sequence based on the polarization spectrum imaging as claimed in claim 1, wherein the step four specifically comprises:
the method comprises the following steps of carrying out image blocking on a polarization intensity image based on a local region segmentation dynamic threshold method, further screening the region, and removing the region with overlarge information fluctuation and excessively uneven material distribution, wherein the method comprises the following specific steps:
step 4-1: uniformly dividing the polarization intensity image to obtain a plurality of sub-blocks, calculating the average absolute deviation of the sub-blocks in each sub-block, and determining the size of a corresponding local window;
step 4-2: when the local window slides in the whole target image, calculating the window area proportion of different image sub-blocks in the window, and weighting and summing the threshold corresponding to each sub-block according to the proportion to obtain a new threshold;
step 4-3: calculating the average absolute deviation of pixels in the local sliding window by the same method, and judging the change condition of the pixel gray level in the local window area by comparing the average deviation of the sub-block pixels with the calculated window threshold;
step 4-4: comparing the average absolute deviation with the global threshold value, and judging whether the area is a flat area with gray scale change or an area with sharp gray scale change;
and 4-5: after the step is finished, a characteristic area with smooth gray scale change in the target area and a light intensity characteristic vector corresponding to each pixel point in the area are obtained.
6. The method for identifying the stamping and signing sequence based on polarization spectrum imaging as claimed in claim 1, wherein the fifth step specifically comprises:
in the polarization characteristic image, corresponding characteristic areas, namely a pure inkpad area, a pure ink area, an inkpad-ink-first area and an inkpad-ink-first area are selected.
CN202010035340.3A 2020-01-14 2020-01-14 Method for identifying stamping signature sequence based on polarization spectrum imaging Pending CN111189784A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010035340.3A CN111189784A (en) 2020-01-14 2020-01-14 Method for identifying stamping signature sequence based on polarization spectrum imaging

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010035340.3A CN111189784A (en) 2020-01-14 2020-01-14 Method for identifying stamping signature sequence based on polarization spectrum imaging

Publications (1)

Publication Number Publication Date
CN111189784A true CN111189784A (en) 2020-05-22

Family

ID=70706268

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010035340.3A Pending CN111189784A (en) 2020-01-14 2020-01-14 Method for identifying stamping signature sequence based on polarization spectrum imaging

Country Status (1)

Country Link
CN (1) CN111189784A (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE10046199A1 (en) * 2000-09-19 2002-03-28 Leica Microsystems Method for three-dimensional forensic signature analysis for use in determining writing sequence, e.g. at cross points which stroke or line overlays the other, writing implement used, signature age, etc.
CN103105385A (en) * 2013-01-29 2013-05-15 沈阳博兴亚达科技有限公司 Method for spectral fluorescence detection and three-dimensional rotating microscope testing of sequence of crossed writings and seal stamps
WO2017042807A1 (en) * 2015-09-07 2017-03-16 Mobileodt Ltd. Handheld mobile device for detection of biological evidence
CN108107003A (en) * 2017-12-15 2018-06-01 哈尔滨工业大学 Fast illuminated light field-polarization imager and imaging method based on microlens array
CN108414468A (en) * 2017-02-09 2018-08-17 哈尔滨工业大学 Infrared spectrum wave band feature Enhancement Method based on wavelet transformation and nonlinear transformation
CN109387501A (en) * 2018-11-29 2019-02-26 中国政法大学 Utilize the method for quick three-dimensional Raman spectrum reconfiguration technique identification crossing stroke sequence

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE10046199A1 (en) * 2000-09-19 2002-03-28 Leica Microsystems Method for three-dimensional forensic signature analysis for use in determining writing sequence, e.g. at cross points which stroke or line overlays the other, writing implement used, signature age, etc.
CN103105385A (en) * 2013-01-29 2013-05-15 沈阳博兴亚达科技有限公司 Method for spectral fluorescence detection and three-dimensional rotating microscope testing of sequence of crossed writings and seal stamps
WO2017042807A1 (en) * 2015-09-07 2017-03-16 Mobileodt Ltd. Handheld mobile device for detection of biological evidence
CN108414468A (en) * 2017-02-09 2018-08-17 哈尔滨工业大学 Infrared spectrum wave band feature Enhancement Method based on wavelet transformation and nonlinear transformation
CN108107003A (en) * 2017-12-15 2018-06-01 哈尔滨工业大学 Fast illuminated light field-polarization imager and imaging method based on microlens array
CN109387501A (en) * 2018-11-29 2019-02-26 中国政法大学 Utilize the method for quick three-dimensional Raman spectrum reconfiguration technique identification crossing stroke sequence

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
LINTON GODOWN: "Sequence of writings", 《THE JOURNAL OF CRIMINAL LAW, CRIMINOLOGY, AND POLICE SCIENCE》 *
S FANALI AND M. SCHUDEL: "Some Separation of Blak and Red Water—soluble fiber—trip Pen Inks by Capillary Zone Electrophoresis and thin-layer chromatography", 《JOURNAL OF FORENSIC SCIENCES》 *
魏松 等: "偏振光显微镜判断激光打印文件朱墨时序实验初探", 《中国司法鉴定》 *

Similar Documents

Publication Publication Date Title
Zhang et al. GAN-FM: Infrared and visible image fusion using GAN with full-scale skip connection and dual Markovian discriminators
Chahi et al. Local directional ternary pattern: A new texture descriptor for texture classification
Peregrina-Barreto et al. Automatic grain size determination in microstructures using image processing
Raghavendra et al. A new perspective—Face recognition with light-field camera
Uss et al. Maximum likelihood estimation of spatially correlated signal-dependent noise in hyperspectral images
Zia et al. 3D reconstruction from hyperspectral images
Dong et al. Infrared image colorization using a s-shape network
Liang et al. 3D plant modelling via hyperspectral imaging
CN116612472B (en) Single-molecule immune array analyzer based on image and method thereof
CN106951863A (en) A kind of substation equipment infrared image change detecting method based on random forest
Xie et al. Trainable spectral difference learning with spatial starting for hyperspectral image denoising
Runxin et al. Survey on image saliency detection methods
Chen et al. Scene segmentation of remotely sensed images with data augmentation using U-net++
Gupta et al. Fast single image haze removal method for inhomogeneous environment using variable scattering coefficient
Huynh et al. Hyperspectral imaging for skin recognition and biometrics
Kaur et al. Image fusion on digital images using Laplacian pyramid with DWT
CN117237736A (en) Daqu quality detection method based on machine vision and deep learning
Bremananth et al. Wood species recognition using GLCM and correlation
Vangah et al. Global and local characterization of rock classification by Gabor and DCT filters with a color texture descriptor
Donate et al. Classification of textures distorted by waterwaves
Tzortzis et al. Automatic inspection of cultural monuments using deep and tensor-based learning on hyperspectral imagery
CN111189784A (en) Method for identifying stamping signature sequence based on polarization spectrum imaging
CN116883893A (en) Tunnel face underground water intelligent identification method and system based on infrared thermal imaging
Gillis et al. Manifold learning techniques for the analysis of hyperspectral ocean data
Omar et al. GLCM-based metric for image fusion assessment

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
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20200522