CN108965636A - A kind of flatbed scanner and its application - Google Patents
A kind of flatbed scanner and its application Download PDFInfo
- Publication number
- CN108965636A CN108965636A CN201810918811.8A CN201810918811A CN108965636A CN 108965636 A CN108965636 A CN 108965636A CN 201810918811 A CN201810918811 A CN 201810918811A CN 108965636 A CN108965636 A CN 108965636A
- Authority
- CN
- China
- Prior art keywords
- spectrum
- image
- sample
- light source
- picture
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/00519—Constructional details not otherwise provided for, e.g. housings, covers
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/00127—Connection or combination of a still picture apparatus with another apparatus, e.g. for storage, processing or transmission of still picture signals or of information associated with a still picture
- H04N1/00249—Connection or combination of a still picture apparatus with another apparatus, e.g. for storage, processing or transmission of still picture signals or of information associated with a still picture with a photographic apparatus, e.g. a photographic printer or a projector
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/024—Details of scanning heads ; Means for illuminating the original
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/024—Details of scanning heads ; Means for illuminating the original
- H04N1/028—Details of scanning heads ; Means for illuminating the original for picture information pick-up
- H04N1/02815—Means for illuminating the original, not specific to a particular type of pick-up head
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/04—Scanning arrangements, i.e. arrangements for the displacement of active reading or reproducing elements relative to the original or reproducing medium, or vice versa
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a kind of flatbed scanners, including tabular shell structure and the sweep mechanism being contained in shell structure, bearing panel, light source module group, sweep mechanism to be scanned within a preset range, bearing panel is located above sweep mechanism, to carry scanned object, it is with transparent panel, so that sweep mechanism scanning is located at the scanned object on transparent panel, light source module group connects sweep mechanism, for emitting light to sweep mechanism, light source module group can emit the light of multiple and different spectrum.The invention also discloses the schemes that above-mentioned flatbed scanner is applied to image analysis and material identification.Flatbed scanner of the present invention can emit the light of multiple and different spectrum by setting tabular shell structure, form a cheap, easy to carry, easy to operate multispectral scanner.
Description
Technical field
The present invention relates to optical scanner field, especially a kind of flatbed scanner and its application.
Background technique
Spectrum (spectrum) is polychromatic light after dispersion system (such as prism, grating) light splitting, the monochrome opened by dispersion
The pattern that light is arranged successively by wavelength (or frequency) size.In addition to (the wavelength covering 390 to 770 of the visible visible spectrum of human eye
Nanometer), spectrum further includes the longer infrared ray of wavelength and shorter ultraviolet spectra.
It is multispectral by multiband detector detection target surface brightness and spectral information, wave-length coverage compares visible light
Greatly, wave band up to several to dozens of, such as ultraviolet (UV), infrared (IR), microwave etc..
Multi-optical spectrum imaging technology mainly has differences for principle, by mesh the absorption of different wave length light with object
Intensity variation of the object in one group of monochrome discrete wavelength is marked to realize the application demands such as detection, discrimination, compared to using single
Visible light have higher identification, antifalsification, safety and reliability.
But as shown in fig. 6, multispectral imaging equipment price is expensive at present, image taking speed is slow, and operation difficulty is big, is suitble to experiment
Room environmental uses.
Summary of the invention
The present invention provides a kind of cheap, easy to carry, easy to operate flatbed scanner and its application.
To achieve the above object, technical scheme is as follows:
A kind of flatbed scanner including tabular shell structure and the sweep mechanism being contained in shell structure, is held
Section plate, light source module group, to be scanned within a preset range, bearing panel is located above sweep mechanism sweep mechanism, uses
To carry scanned object, with transparent panel, so that sweep mechanism scanning is located at the scanned object on transparent panel, light
Source mould group connects sweep mechanism, and for emitting light to sweep mechanism, light source module group can emit the light of multiple and different spectrum；It sweeps
Instrument is retouched with control system, control system includes processing module and light source emitting module, light source receiving module, and processing module is used
In the operation for controlling other each modules, light source emitting module connects light source module group, for driving light source module group to generate spectral signal,
Light source receiving module connects receiver, and the spectral signal for receiving receiver is converted into electric signal transmission to processing module；
When scanning, the processing module of scanner receives " scanning signal ", is sent out by the preset light source module group of light source emitting module order
Default spectrum is penetrated, meanwhile, processing module passes through motor drive module, control motor rotation, to drive probe round-trip along rack gear
Movement, receiver starts to receive spectral signal, and sends processing module to, completes a kind of scanning of spectrum, other spectrum are by upper
Step is stated successively to scan.
In one embodiment, sweep mechanism includes striplight, receiver, control panel and movement mechanism, band-like
Luminaire is used to receive the spectral emissions signal of light source module group, and forms band-like light and emit to scanned object, and receiver is used
In the reflection light of reception scanned object, and electric signal is converted optical signals to, is transferred to control panel, movement mechanism is for controlling
Striplight processed and receiver move, to scan the full scope of scanned object.
In one embodiment, movement mechanism includes motor, driving cog wheel and rack, motor, transmission gear with it is band-like
Luminaire, receiver, which are fixed on together on a crossbeam, forms probe, and motor is connect with transmission gear, rack gear be it is longitudinally disposed,
It is engaged with gear, the rotary motion of gear is converted into the straight reciprocating motion along rack gear, moves probe along longitudinal direction, to sweep
Retouch the full scope of scanned object.
In one embodiment, control system further includes motor drive module, and motor drive module connects motor, control
Motor forward or reverse, makes probe along predetermined manner longitudinal movement.
A kind of image-recognizing method, which is characterized in that use flatbed scanner as described above, image-recognizing method packet
It includes:
Intelligent selection scanning optical spectrum, including Image Acquisition and pretreatment, including flatbed scanner is used to utilize a variety of light
The multiple spectrum pictures of scanning collection are composed, and dimensional standard and denoising are carried out to each spectrum picture of acquisition；Intelligent Selection
Spectrum is selected for subsequent analysis, is analyzed by traditional machine learning algorithm or feature consistency, in multiple scanning optical spectrums
Selection retains relatively more complete spectrum to partial information is altered；
The generation of multiple spectrum picture features, including take weighted value: FF=k1*F1+k2*F2+ ...+kn*Fn, wherein quilt
The weight setting of the spectrum picture of selection is relatively large；Or matrix is arranged in by FF=[F1, F2 ..., Fn], wherein FF
Only comprising the spectrum picture selected, then take characteristic value as multispectral characteristics of image by dimensionality reduction；
It is swept by above-mentioned multispectral image characteristic extracting method using flat using the spectrum picture of intelligent selection
Retouch the content that instrument identification target object is modified.
A kind of image-recognizing method, which is characterized in that use flatbed scanner as described above, image-recognizing method packet
It includes:
Intelligent selection scanning optical spectrum, including Image Acquisition and pretreatment, including flatbed scanner is used to utilize a variety of light
The multiple spectrum pictures of scanning collection are composed, and dimensional standard and denoising are carried out to each spectrum picture of acquisition；Intelligent Selection
Spectrum is selected for subsequent analysis, is analyzed by traditional machine learning algorithm or feature consistency, in multiple scanning optical spectrums
Selection retains relatively more complete spectrum to partial information is altered；
3D convolutional network structure is designed, input is (n1*w*h), wherein n1 is the quantity of image, and h and w are image respectively
Length and width；Multiple spectrum image is arranged in that same order { M1, M2 ..., Mn }, is then enter into 3D convolution
Network；In 3D convolutional network, the picture of input is finally mapped to by full articulamentum by 3D convolution, the pond 3D calculation process；
Subsequent identifying processing is used for according to the result of full articulamentum；
Using the scanning optical spectrum of intelligent selection, the image of target object is scanned, by above-mentioned full articulamentum, use is flat
The content that scanner identification target object is modified.
A kind of substance identification, which is characterized in that use flatbed scanner as described above, substance identification packet
It includes:
Master sample feature is established, and the multispectral image of collected master sample is acquired using flatbed scanner,
And extract the feature of multispectral image；
The Image Acquisition of sample to be examined collects the multispectral image of sample to be examined, spectrum using flatbed scanner
Wavelength when wavelength is established with master sample feature is identical；
Above-mentioned collected sample to be examined is handled, the characteristics of image of sample to be examined is extracted, the characteristics of image
Classification is identical as master sample；
Material identification, through master sample feature compared with sample to be examined feature, judge measuring samples whether with standard
Sample is identical.
The beneficial effects of the present invention are: flatbed scanner of the present invention passes through setting tabular shell structure, and can emit
The light of multiple and different spectrum forms a cheap, easy to carry, easy to operate multispectral scanner.
Detailed description of the invention
Fig. 1 is the stereogram exploded view of flatbed scanner of the embodiment of the present invention.
Fig. 2 is that flatbed scanner shown in Fig. 1 combines top view.
Fig. 3 is that flatbed scanner shown in Fig. 1 removes the top view after upper cover and bearing panel.
Fig. 4 is that flatbed scanner shown in Fig. 1 removes the bottom view after bottom plate.
Fig. 5 is the block diagram of flatbed scanner of embodiment of the present invention control system.
Fig. 6 is the schematic diagram of existing multispectral imaging equipment.
Fig. 7 is the schematic diagram that the embodiment of the present invention is applied to image after altering when document judicial expertise.
Fig. 8 is the schematic diagram that the embodiment of the present invention is applied to original image when document judicial expertise.
Fig. 9 is the schematic diagram that the embodiment of the present invention is applied to the reparation of ancient books file.
Figure 10 is the schematic diagram of window sliding of embodiment of the present invention method.
Specific embodiment
With reference to the accompanying drawing and example, the present invention will be further described.
In the present embodiment, as shown in Figures 1 to 4, flatbed scanner include framework 1, sweep mechanism 2, bearing panel 3,
Upper cover 4, light source module group 5, bottom plate 6.The horizontal profile ratio A4 paper of the present embodiment flatbed scanner is slightly larger, is mainly used for scanning and is less than
Or the sheet article (such as paper document) equal to A4 paper size.It is of course also possible to as needed, increased or decrease flat bed scanning
The size of instrument, for scanning the article for being less than or equal to A1 to A5 paper size or other sizes.
1 overall size of framework is close with scanner, more slightly larger than A4 paper, including bottom 11 and by 11 outer rim of bottom
Upwardly extend four side walls 13 to be formed.Bottom 11 is surrounded by multiple pillars 15 for being used to support main circuit board with other parts,
Middle part raises upward, and upper surface is covered with light partition (not shown), and aluminium foil material can be used in light partition, closes for building
Suitable lighting environment.
Sweep mechanism 2 is scanned to (the present embodiment is A4 paper range) within a preset range, is fixed in framework,
Including probe 21, rack gear 23, control panel 25.Probe 21 is the main component of flatbed scanner, to be laterally arranged, including
Crossbeam (not indicating) and the striplight (not shown) being fixed on crossbeam, receiver (not shown), motor (not shown),
Transmission gear (not shown).Striplight is transversely distributed and towards upward, for receiving 5 spectral emissions of light source module group letter
Number, and form band-like light and emit to scanned object.Receiver and striplight are distributed along crossbeam side by side, for receiving quilt
The reflection light of object is scanned, and converts optical signals to electric signal, is transferred to control panel, electric machine structure is for driving driving cog
Wheel rotation, controls beam structure horizontal movement along longitudinal direction.
Rack gear 23 be it is longitudinally disposed, engaged with transmission gear, it is past that the rotary motion of gear is converted to the straight line along rack gear
Multiple movement, moves probe 21 along longitudinal direction, to scan the full scope of scanned object (such as document of A4 paper size).
Multiple electronic component (not shown)s such as chip of the control panel 25 including circuit board (not shown), on circuit board,
And the protection board 251 positioned at circuit board, it is connected with elements such as motor, light source module group 5, receiver, communication devices, is used for
Motor movement is controlled, the spectral signal that receiver obtains is converted into electric signal, image is ultimately formed and uploads to host computer and (do not draw
Out), spectrum picture and subsequent analysis are checked for user.Control panel 25 can also have signal processing function, and receiver is transmitted
The electric signal come in carries out brief analysis processing.Certainly, analysis processing work is mainly come by host computer, such as computer, mobile phone
At this moment processing needs to initially set up communication connection to ensure that data are transmitted.Communication mode can be USB, and wired ethernet etc. is wired
Transmission mode is also possible to the wireless transmission methods such as bluetooth, WIFI.
Bearing panel 3 is fixed in framework 1 and is located at 2 top of sweep mechanism, to carry scanned object.Bearing panel
3 include frame shape fixed plate 31 and the transparent panel 33 for being fixed on 31 middle part of fixed plate.Transparent panel 33 is A4 paper size, to
The light for allowing lower section striplight to project is by reaching scanned object, and the light for allowing scanned object to reflect is through directive
Receiver.31 one end of fixed plate is equipped with a key switch 35, for opening, the power supply of disconnected scanner.
Upper cover 4 is articulated in 1 one end of framework, to be covered on scanned object, prevents exterior light from entering scanning area, seeks
Make metastable light environment.
Light source module group 5 is fixed between 1 bottom 11 of framework and bottom plate 6, striplight is connected, for emitting light to band
Shape luminaire.In the present embodiment, light source module group 5 includes 6 transmitting light sources 51, emits the light of different spectrum respectively.Other
In embodiment, light source module group can also only set a transmitting light source, and the light of different spectrum is emitted by filter element.Or it sets
Multiple transmitting light sources, part emit light source and emit monochromatic spectrum, and part emits light source and emits different spectrum by filter element
Light.
Bottom plate 6 is fixed on 1 bottom of framework below framework 1, to protect internal element.Bottom plate lower part is single equipped with power supply
Member, the power demands for scanner.
Above-mentioned framework 1, upper cover 4, bottom plate 6 constitute tabular shell structure, for accommodating above-mentioned sweep mechanism 2, loading end
The elements such as plate 3, light source module group 5.
As shown in figure 5, the control system of scanner include processing module and light source emitting module, light source receiving module,
Communication module, power management module, key trigger module, motor drive module.These modules are all disposed on control panel, are passed through
Conducting wire is connected with related hardware.
Processing module is the brain of control system, for controlling the operation of other each modules.Processing module can also be to original
Beginning scan data is briefly handled (such as denoising, smooth, spectrum fusion), these final data can be transferred to upper
Machine.
Light source emitting module connects light source module group, for driving light source module group to generate spectral signal.Because the present embodiment has 6
The light source of a difference spectrum, so processing module can successively issue scan instruction and (it is automatic control 6 spectrum respectively when scanning
Successively scan), can also according to the control signal of host computer, select wherein certain it is several give light source emitting module, sent out by light source
Penetrate module control light source module group.
Light source receiving module connects receiver, and the spectral signal for receiving receiver is converted into electric signal transmission to place
Manage module.
Communication module connection communication device, for being carried out data transmission using wired or wireless way with host computer.
Power management module connects power supply, for being managed to power supply, such as obtains remaining capacity information, avoids overcharging
Etc..
Key trigger module connects key switch, and for controlling processing module by key switch, starting or closing are swept
Retouch operation.
Motor drive module connects motor, controls motor forward or reverse, makes probe along predetermined manner longitudinal movement.
A kind of general spectrum round trip, processing module scanning speed can also be controlled by controlling the revolving speed of motor and scanning walks
It is long.
Typical operating process is illustrated below.User presses switch key, and host computer opens control software, even
It is connected to scanner, selects " starting to scan " in upper computer software, the control module of scanner receives scanning signal at this time, leads to
It crosses light source emitting module order light source module group and emits a kind of default spectrum, while receiver reception is reflected back from scanned object
Spectral signal, send processing module to, processing module carries out transferring data to host computer after centainly handling；Processing module is logical
Motor drive module is crossed, control motor rotation completes a kind of sweeping for spectrum to drive probe to make a back and forth movement along rack gear
It retouches, then switches other spectrum and be scanned.
Compared with existing multispectral imaging equipment, the present embodiment flatbed scanner is had the advantages that
1, structure is simple, multiple spectrum cameras is integrated in a scanner, other elements are equal in addition to light source module group
It can share, relatively existing multiple spectrum cameras are separately connected the structure of transmitting light source, and structure is significantly simple；
2, cheap, opposite existing equipment, structure is few, and multiple element can use the universal element of scanner
Part；
3, small in size, it is easy to carry, can carry it is outgoing be scanned, being different from existing multispectral imaging equipment can only be
Laboratory implementation, therefore larger range of application can be obtained.
The present embodiment flatbed scanner can be applied to image recognition, be described as follows.
The present embodiment is acted on based on the multispectral different detections to different objects, is used for identification or expert assignment.By
Can scan and not seen under visible light conditions or unsharp content in the spectrum using different wave length, when file content by
Damage, by modification, correction or erasing etc. (as shown in Figure 7, Figure 8, wherein Fig. 8 is original image, and Fig. 7 is the image after altering), or
When person is unintelligible, multispectral scanner acquisition image can be used, and be based on multispectral image, carry out image characteristics extraction and oneself
Dynamic analysis selects information to retain complete spectrum picture and is used for subsequent processing, on the one hand can reduce the data volume of image analysis,
On the other hand interference of the incomplete spectrum picture of information to subsequent analysis is reduced.
Specifically, first being swept based on multispectral scanner to sample to be examined (being document to be identified in the present embodiment)
It retouches, obtains the multispectral image of document to be checked, image preprocessing and feature extraction are carried out to each spectrum picture.
On the one hand it by multispectral consistency analysis, intelligently picks out and complete spectrum picture is retained to target information；Separately
On the one hand, based on the spectrum picture of intelligent selection, carry out deeper into image analysis, such as the identification of file, identification and auxiliary
Help reparation etc..Scanning optical spectrum intelligent selecting method and image analysis method are specifically described below.
Step S51, Image Acquisition and pretreatment.
Dimensional standard is carried out using the multiple spectrum pictures of multiple spectrum scanning collection, and to each spectrum picture of acquisition
Change and denoising, such as brightness zero-mean: I '(x,y)=I(x,y)- avg (I), wherein avg (I) is the luminance mean value of image.
Step S52, is analyzed by feature consistency, and selection target information retains more complete light in multiple scanning optical spectrums
Spectrum is used for subsequent analysis.
According to multispectral theory, different spectrum can detect the different chemically and physically attribute of target.Due to the correction of file
Or person's handwriting desalination, certain spectrum (such as visible light) may detect the real information of fall short, but use other wavelength
Spectrum is possible to get the content that these spectrum (visible light) are not seen.Further, the incomplete spectrum of these information
On the one hand image does not have good facilitation to subsequent analysis (such as identification, identification etc.), on the other hand also will affect mesh
The extraction for marking feature, influences subsequent analysis.Therefore, method proposes a kind of scanning optical spectrum intelligent selecting methods, first to more
Kind spectrum picture carries out intelligent classification, and selection retains more complete spectrum picture to target information and carries out subsequent analysis.
Due to different detection targets, light spectrum image-forming also can be different, and therefore, this programme is based primarily upon the same target
Different spectrum between class spacing analyzed:
Each spectrum picture is divided into many image using window sliding method by step S521, image block segmentation
Block, window sliding method is as shown in Figure 10, if window size is (w, h), sliding step T is then distinguished as (I for onex,Iy)
Image, it is availableA difference sub-block；Meanwhile also using the sub-block of different resolution, example
Such as: 8*8,16*16,32*32 realize multi-scale division；
Step S522, the feature extraction of image block extract characteristics of image, image to each image block obtained above
Feature can be Gabor wavelet, Haar, Hog, Sift, the diversified forms such as textural characteristics；Wherein, Gabor wavelet is a kind of common
Multiresolution texture characteristic extracting method, calculation is the convolution of Gabor wavelet kernel function and original image.
The product of Gabor wavelet core one oval Gaussian envelope and complex plane wave, is defined as follows:
Wherein, u and v defines direction and the scale of Gabor core, and z=(x, y) indicates that pixel coordinate, ‖ * ‖ are modulus fortune
It calculates, wavelet vectorskv=kmax/ f,kmaxFor maximum frequency, f be kernel function in a frequency domain
Breadth coefficient.In scheme, take Gabor wavelet for 5chidu, 8 directions, i.e. v ∈ { 0,1,2,3,4 }, u ∈ 0,1,2,3,4,5,6,
7 }, other parameters σ=2 π, kmax=pi/2,
Step S523, feature vector dimensionality reduction, there are many image block and intrinsic dimensionality obtained above, carried out using algorithm to it
Space projection and Feature Dimension Reduction, this kind of algorithms for example: PCA, LDA etc. mainly introduce the dimension reduc-ing principle of PCA here, step
Suddenly include:
1, original feature vector is lined up to the matrix X of n row m columnn*m, every a line is the feature vector of an image block；
2, by Xn*mEach row carry out zero averaging, that is, subtract the mean value of this line；
3, seek Xn*mCovariance matrix Cn*n:
Cn*n=(ci,j,ci,j=cov (Xi,Xj))
4, find out the characteristic value and corresponding feature vector of covariance matrix；
5, by feature vector by corresponding eigenvalue size from top to bottom by rows at matrix, k row composition matrix p before taking,
As dimensionality reduction data.
Step S524, the information integrity analysis of original image.Due to may only have regional area in an entire image
Information it is imperfect, this programme is judged using local features: successively to different spectrum, the image block of same area is believed
Breath consistency is analyzed, specific steps are as follows:
Consistency analysis is clustered using feature of the K-means clustering method to different spectrum picture blocks, and information is complete
Spectrum be one kind, the spectrum of loss of learning is another kind of.
Step S525, the judgement of spectrum integrality and mark.
The judgment basis of spectrum integrality are as follows: set in some image block, the light of information complete spectrum and loss of learning
Composing quantity is respectively n1And n2, work as n1> > n2When, by n1One kind be determined as information complete area, n2One kind be determined as in vain
N is worked as in region1≈n2When, it is inactive area by the regional determination；
Spectrum integrality mark: by the way of locally marking, each image subblock is labeled, notation methods
Are as follows: 0 indicates inactive area, and 1 indicates the complete region of information.Then for some spectrum picture Mi, its information integrity mark
Note are as follows:
IAi,r,n={ k1,k2,k3,....kn,},ki=0 or1
Wherein i indicates the sequence of spectrum, and r indicates the resolution ratio of image subblock, i.e. 8*8,16*16, and 32*32, n indicate image
The quantity of sub-block.
The image for being labeled as 1 region is to retain more complete spectrum picture to target information, and when subsequent analysis is only selected
Them are selected, to not only can be reduced the workload of subsequent analysis, but also is avoided that the incomplete spectrum picture of information to subsequent analysis
Interference.
Step S53 carries out subsequent image analysis tasks, such as image authentication, mesh based on the complete spectrum picture of information
Mark not etc..Different according to specific application, method for subsequent processing can be " conventional machines study " method, be also possible to " deep
Degree study " method, both of which only consider that the complete spectrum of information carries out feature extraction and analysis.
Below by taking image authentication as an example, it is illustrated.Image authentication can be applied to the identification of ancient books file, such as width Gu
The famous painting in generation, it is smudgy, but by the comparison with duplication file, the true and false of the famous painting can be identified.
Image authentication is actually the similarity for comparing object to be measured and master sample；This programme mainly uses a kind of distance
The mode of measurement is compared, and belongs to " conventional machines study " method.The calculation method of sample distance has very much: Euclidean distance,
Mahalanobis distance, included angle cosine, manhatton distance etc., by taking mahalanobis distance as an example, if two sample vectors are respectively X1And X2, association
Variance matrix is denoted as S, mean vector μ, then the mahalanobis distance between two of them sample vector are as follows:
Step S531, the comparison of image subblock.The image subblock between sample to be tested and master sample is compared one by one, is compared
Method are as follows:
1, definition: Ni,r,nIndicate spectrum MiResolution ratio is the number of the invalid image subblock (not comparing) of r；Ki,r,nIt indicates
Spectrum MiResolution ratio is the number of the effective image sub-block of r；Di,r,nSpectra re-recorded MiResolution ratio be r effective image subblock and
The mahalanobis distance of master sample；
2, spectrum picture M is fetched one by oneiImage subblock Veci,r,k, wherein i indicates the sequence of spectrum, and r indicates image
The resolution ratio of block, k indicate the corresponding serial number of the image subblock；
3, fetch IAi,r,k, work as IAi,r,kWhen=0, without comparison, Ni,r,nAdd 1, works as IAi,r,kWhen=1, compare,
Ki,r,nAdd 1, Di,r,nIn addition the mahalanobis distance of the sub-block；
4, finally, image MiQualification result when resolution ratio is r is denoted as: N number of image subblock altogether, wherein Ni,r,nA nothing
Effect, Ki,r,nIt is a effective, Mean Mahalanobis distance di,r=Di,r,n/Ki,r,n。
Step S532, amalgamation judging.Qualification result is the fusion of all spectrum picture results, it is contemplated that participates in comparing
Spectrum quantity is more, and comparison result is more credible, takes the confidence factor of spectrum are as follows: ηi,r=Di,r,n/ N is carried out comprehensive in accordance with the following steps
Close judgement:
1, the comparison result of different spectrum equal resolutions is counted, final score is denoted as:
Wherein, r indicates resolution ratio, ηi,rIndicate spectrum MiConfidence factor when resolution ratio is r, di,rFor spectrum MiDividing
Resolution be r when and standard picture mahalanobis distance；
2, setting threshold value is Tth, work as Score > Tth, be determined as it is identical, as Score < Tth, it is determined as difference.
It is spectrum intelligent selecting method to be described so that document is identified as an example and based on the complete spectrum picture of information above
Image analysis (identification) method.It is appreciated that the present invention is based on multispectral image analysis method can be applied to it is other a variety of
Occasion.Such as the reparation for ancient books file, as shown in figure 9, even if ancient books file has not seen ink traces (such as by naked eyes
On the left of Fig. 9), but by multispectral scanner, it can select automatically and internally hold the slightly complete spectrum picture of visible or information, identify
Picture material (on the right side of such as Fig. 9), is repaired for assisting.Again for example for fingerprint recognition, although the finger on classified papers or bank note
Line is wiped free of or covers, although complete information cannot be obtained under visible light conditions, by the spectral scan of other wavelength, and can
To obtain clearer image.
The present embodiment flatbed scanner applies also for material identification, is described as follows.
The present embodiment is based on multispectral substance identification, is spectrally shown not according to different chemical components
Same spectral information chooses the image of different wave length from the multispectral data of substance, these images, which respectively correspond, to be highlighted not
With chemical material, so as to facilitate from these images the various inks of identification, pigment (seal stamp-pad ink), paper or other
Material.By the way that the spectral information of these materials is compared with the same time with the spectral information of kind material, so that it may identify
The true and false of document or ancient books file out.And if there is the spectra database of each material, then the spectral information that can also will be observed that
It is compared with the spectra database information of known materials, can more identify the information such as brand, the place of production, the age of document file output.
It may include steps of specifically, the present embodiment is based on multispectral substance identification.
Step S61, master sample feature are established.The step for be collected when being embodied according to the difference of the application, under
Face is described in detail.
Step S611, the collection of sample data: being collected according to the difference of task, such as to identify the one of document material
Cause property etc. can design from ink, paper etc. and carry out multilayer identification, in collection material, carry out respectively to each sample
Mark, content include: the writing time of materials A, and ink brand used etc. indicates the mark to every a material with Y,
Middle Y=[y1, y2 ..., yn], yn indicate each corresponding attribute value.
Step S612, the acquisition of sample image: the spectrum picture of the every a sample of acquisition, based on multispectral camera to sample
This carries out shooting sampling (spectral wavelength that uses may be selected 300 nanometers, 500 nanometers, 700 nanometers etc. a variety of), so each
Spectrum can acquire a sub-picture, be set to M, wherein M=[M1, M2, M3 ..., Mn], and Mn is some spectrum picture；
Step S613, the image characteristics extraction of master sample: by be then based on the aspect ratio of image to identifying realizing or
Identification, therefore, extracts the feature of each spectrum picture, these features can be selected according to usage scenario: edge feature, Hog
Histogram, Gabor wavelet, Haar small echo, SIFT feature etc.；
Step S614, the classification of standard sample database and builds library: with RGB, the differences such as gray scale or depth image, for each
A target has N kind difference spectrum picture to be described, by taking the identification of document material as an example, firstly for certain a kind of document material
Material, has collected many samples, secondary purpose is analyzed from ink material, writing time, paper age etc., so
And in the sample of collection, it is consistent there may be paper material, but it is inconsistent to write pen and ink；Or writing pen and ink are consistent, and
The different sample of paper material.If described using every a kind of sample as main body, can there are many different combinations, sample
Description it is unintelligible so that follow-up data treatment process is more complicated；Therefore, the present embodiment from more fine-grained sample attribute (such as
Ink material, writing time, paper age) classify and builds library.
Step S62, the Image Acquisition of sample to be examined.If can control sample to be examined and master sample in same light source environment
Middle collection can simplify the data handling procedure in later period, be conducive to improve accuracy of identification.However accomplish that this point difficulty is bigger,
Improved from several respects: 1) continuing to use as far as possible and collect similar environment with master sample；2) pass through video camera light filling unit
It is made up.
Step S63 handles the image of sample to be examined.May include following steps:
Step S631, image standardization: sample-size is consistent with standard sample database first, normalizes secondly by brightness
(wherein brightness is pixel R, the average value of tri- color of G, B) and setting contrast algorithm (average brightness * contrast rating) are to figure
As being normalized.The wherein normalized operation of brightness is: by the average brightness of each pixel subtracted image, setting contrast
It is by being adjusted multiplied by contrast rating.
Step S632, the extraction of characteristics of image to be checked, treatment process of this process with master sample.
Step S64, material identification.This example provides two kinds of identification methods: it is based on sample distance and is based on Attribute Recognition, under
Face is described respectively.
The discrimination process based on sample distance is introduced first.Discrimination process be actually in master sample find and to
Sample originally most similar sample.It can be with the following method:
Step S6411, the distance based on sample characteristics are compared: distance mathematically has many representation methods, such as
Euclidean distance, COS distance etc., wherein Euclidean distance are as follows:
Step S6412, by taking the identification of document material as an example, available three distances of every one kind spectrum, are pen and ink respectively
Material, paper age and paper material may also more other describe, here every a kind of light certainly for different task
Spectrum apart from comparison result are as follows: R=[R1, R2, R3]；
Step S6413, for each attribute R1-R3, if only take a distance it is nearest as a result, being easy to be acquired
The interference of process causes the confidence level of comparison result lower, therefore, borrows the thinking of K-NN here, but be not exclusively the side K-NN
Method, each attribute of each spectrum picture, such as R1, R2, R3 take K most similar distances, such as: spectrum picture
The paper age comparison result (R2) of Mi takes { d1, d2 ..., dk } a near preceding as a result, assigning this K different from K-NN method
The different weight of a ranking results, distance is more close (sequence is more forward), and the weight of acquirement is bigger, it is assumed that the weight coefficient taken
Are as follows: { e1, e2 ..., ek }, finally by the weighting apart from comparison result relevant to the attribute of all spectrum pictures:
Wherein, dijIndicate the distance of the sequence j of spectrum Mi, ejFor the weight for the j that sorts.The R2 attribute of final sample to be tested
Determine result are as follows: the classification of corresponding maximum score value.For example, the R2 attribute (paper age) of sample to be tested corresponding 2015 points
Value is maximum, then determining result are as follows: the paper age of sample to be tested is 2015.
Discrimination process based on Attribute Recognition, the method that " manual Feature extraction~+ conventional machines learning method " can be used
It is further identified, can also be distinguished based on deep learning method, these methods are all relatively mature.So
And target is described since there are many different spectrum pictures, treatment process can be different.Final recognition result is a variety of
The fusion results of spectrum picture.
Step S6421, the fusion of multispectral image
There are two types of classification methods for image co-registration, and one is forward direction fusions, i.e., are first merged before image compares, Ling Yishi
Backward fusion, i.e. each image are used to obtain a comparison result, finally merge comparison result.
The method of " manual Feature extraction~+ conventional machines learning method " is using preceding to fusion, and when specific implementation, this programme is set
The characteristics of image extracted indicates that wherein F=[f1, f2 ..., fn] indicates the spectrum picture feature extracted with F, will
The feature of all spectrum lines up a matrix, FF (i, j)=[F1, F2 ..., Fj]T, utilize space projection or dimension reduction method
(PCA, LDA) etc. carries out dimensionality reduction to FF matrix and characteristics extraction, the feature after taking dimensionality reduction input to traditional machine learning mould
Type is trained；Here, the characteristics of image of manual extraction may is that the edge feature of binary image, HOG histogram feature,
Haar wavelet character, SIFT feature etc.；Machine learning algorithm may is that SVM, neural network, decision tree etc.；
For deep learning method, to fusion method after, each image after the operation of convolutional neural networks,
A kind of comparison result is obtained, all spectrum all compares one time, using the most result of frequency of occurrence as final comparison result；This
In, deep learning network can be by transfer learning or oneself design and progressive alternate is fitted to obtain；
Step S6422, model training: the input of " manual Feature extraction~+ conventional machines learning method " is the feature extracted
Vector, the input of " deep learning " are pictures；Output is exactly the result of identification or classification.About the training of model, due to adopting
Sample is described with the feature of more grain refined, therefore, the mark of sample is each attribute for it；Training process depends on
How many a sample attribute Y=[y1, y2 ..., yn] are selected, each yn requires one disaggregated model of training；
Step S6423, target identification: to sum up, final recognition result be described with sample attribute, such as: the sample
Originally have which type of y1 attribute, which type of y2 attribute ..., which type of yn attribute.By taking above-mentioned paper as an example, certain paper sample
Originally it may determine that have generation production year in 2015, the writing ink time in 2017, ink material of heroic brand etc..
The present invention is based on multispectral substance identifications to identify substance by the multispectral image feature of substance, is reflecting
When the detection such as true and false of other authoring (printing) time, signature, analysis and identify highest accuracy rate up to 99% or more, and operates
Very simple, it is economical.
Claims (10)
1. a kind of flatbed scanner, which is characterized in that including tabular shell structure and be contained in sweeping in shell structure
Mechanism, bearing panel, light source module group are retouched, for sweep mechanism to be scanned within a preset range, bearing panel is located at scanning machine
Above structure, to carry scanned object, with transparent panel, so that sweep mechanism scanning is located at being swept on transparent panel
Object is retouched, light source module group connects sweep mechanism, and for emitting light to sweep mechanism, light source module group can emit multiple and different spectrum
Light, the light of multiple and different spectrum is respectively formed by monochromatic source, or is formed by polychromatic source by filter element.
2. flatbed scanner according to claim 1, which is characterized in that sweep mechanism includes striplight, receives
Device, control panel and movement mechanism, striplight are used to receive the spectral emissions signal of light source module group, and formed band-like light to
Scanned object transmitting, receiver are used to receive the reflection light of scanned object, and convert optical signals to electric signal, transmit
To control panel, movement mechanism is moved for controlling striplight and receiver, to scan the full scope of scanned object.
3. flatbed scanner according to claim 2, which is characterized in that movement mechanism include motor, transmission gear with
Rack gear, motor, transmission gear, which are fixed on a crossbeam together with striplight, receiver, forms probe, motor and transmission
Gear connection, rack gear be it is longitudinally disposed, engage with gear, by the rotary motion of gear be converted to along rack gear linear reciprocation transport
It is dynamic, move probe along longitudinal direction, to scan the full scope of scanned object.
4. flatbed scanner according to claim 3, which is characterized in that scanner has control system, control system
Including processing module and light source emitting module, light source receiving module, processing module is used to control the operation of other each modules, light
Source transmitting module connects light source module group, and for driving light source module group to generate spectral signal, light source receiving module connects receiver, uses
Electric signal transmission is converted into processing module in the spectral signal for receiving receiver；When scanning, processing module is sent out by light source
It penetrates module command light source module group and emits a kind of spectrum, while receiver receives the spectral signal reflected back from scanned object,
Send processing module to；Processing module passes through motor drive module, control motor rotation, to drive probe to make once along rack gear
Back and forth movement, completes a kind of scanning of spectrum, then switches other spectrum and is scanned.
5. a kind of scanning optical spectrum intelligent selecting method characterized by comprising
Image is acquired using multispectral scanner, and size normalised and denoising is carried out to each spectrum picture, it is described
Multispectral scanner is using the flatbed scanner as described in any in Claims 1-4；
It is analyzed by feature consistency, selection target information retains relatively more complete spectrum picture in multiple spectrum pictures.
6. scanning optical spectrum intelligent selecting method according to claim 5, which is characterized in that spectrum picture selection includes: figure
As block segmentation, each spectrum picture is divided into many image subblocks；Image block characteristics extract, and mention to each above-mentioned image block
Take characteristics of image；Feature vector dimensionality reduction carries out space projection and Feature Dimension Reduction to above-mentioned image block characteristics using algorithm；Spectrogram
The information integrity of picture is analyzed, and multispectral image is divided into information is complete and two classes of loss of learning；The judgement of spectrum integrality
Each image subblock is labeled according to spectrum integrality with mark.
7. a kind of image analysis method, which is characterized in that the information obtained based on method according to claim 5 or 6 is complete
Whole spectrum picture carries out.
8. scanning optical spectrum intelligent selecting method according to claim 7, which is characterized in that image analysis method includes: figure
As the comparison of sub-block, the image subblock between sample to be tested and master sample is compared one by one；Amalgamation judging, qualification result are all
Spectrum picture result fusion, take the confidence factor of spectrum are as follows: ηI, r=DI, r, n/ N carries out comprehensive descision in accordance with the following steps:
The comparison result of different spectrum equal resolutions is counted, final score is denoted as:Wherein, r
Indicate resolution ratio, ηI, rIndicate spectrum MiConfidence factor when resolution ratio is r, dI, rFor spectrum MiResolution ratio be r when and standard
The mahalanobis distance of image；Setting threshold value is Tth, as Score > Tth, be determined as it is identical, as Score < Tth, it is determined as difference.
9. one kind is based on multispectral substance identification characterized by comprising
Master sample feature is established, and the multispectral image of collected master sample is acquired, and extracts the spy of multispectral image
Sign；
The Image Acquisition of sample to be examined, collects the multispectral image of sample to be examined, and spectral wavelength is built with master sample feature
Wavelength immediately is identical；
Above-mentioned collected sample to be examined is handled, the characteristics of image of sample to be examined, the classification of the characteristics of image are extracted
It is identical as master sample；
Material identification judges which attribute measuring samples have through master sample feature compared with sample to be examined feature；
Above-mentioned sample standard deviation is acquired using the flatbed scanner as described in any in Claims 1-4.
10. substance identification according to claim 9, which is characterized in that material identification is used based on sample distance
Mode, every one kind spectrum obtain multiple distances, respectively correspond different attribute；To each attribute of each spectrum picture, take
Wherein K near it is preceding as a result, and assign this K sort result different weights, more forward weight is bigger, compares each time
For ranking all multiplied by a weight, final score, which is equal in all spectrum picture comparison results, there is result in preceding K ranking
Adduction, i.e.,Wherein, dijIndicate the distance of the sequence j of spectrum Mi, ejFor sort j weight, finally
The judgement result of the attribute of sample to be tested is the classification of corresponding maximum score value；Or material identification is used based on Attribute Recognition
Mode identifies the attribute of measuring samples, comprising: Multispectral Image Fusion, to obtain feature vector or picture；Model training, it is defeated
Enter the result of feature vector or picture, output identification or classification；Which attribute target identification, final judgement sample have.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810918811.8A CN108965636A (en) | 2018-08-13 | 2018-08-13 | A kind of flatbed scanner and its application |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810918811.8A CN108965636A (en) | 2018-08-13 | 2018-08-13 | A kind of flatbed scanner and its application |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108965636A true CN108965636A (en) | 2018-12-07 |
Family
ID=64469090
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810918811.8A Pending CN108965636A (en) | 2018-08-13 | 2018-08-13 | A kind of flatbed scanner and its application |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108965636A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112161936A (en) * | 2020-09-29 | 2021-01-01 | 杭州海康威视数字技术股份有限公司 | Optical inspection device for cigarette packets |
| CN112161935A (en) * | 2020-09-29 | 2021-01-01 | 杭州海康威视数字技术股份有限公司 | Optical inspection device for cigarette packets |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1188945A (en) * | 1997-01-23 | 1998-07-29 | 陈伯宁 | Automatic signature comparison method and device |
| CN1442997A (en) * | 2002-03-01 | 2003-09-17 | 精工爱普生株式会社 | Image fetching system |
| CN102158624A (en) * | 2011-03-02 | 2011-08-17 | 朱杰 | Double-sided scanning flatbed scanner |
| JP2012060411A (en) * | 2010-09-09 | 2012-03-22 | Hamamatsu Metrix Kk | Multiband one-dimensional line sensor camera |
| CN105842173A (en) * | 2016-06-06 | 2016-08-10 | 南京大学 | Method for identifying hyperspectral material |
| CN106454000A (en) * | 2016-11-18 | 2017-02-22 | 桂林电子科技大学 | Multifunctional scanner based on image acquisition device |
-
2018
- 2018-08-13 CN CN201810918811.8A patent/CN108965636A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1188945A (en) * | 1997-01-23 | 1998-07-29 | 陈伯宁 | Automatic signature comparison method and device |
| CN1442997A (en) * | 2002-03-01 | 2003-09-17 | 精工爱普生株式会社 | Image fetching system |
| JP2012060411A (en) * | 2010-09-09 | 2012-03-22 | Hamamatsu Metrix Kk | Multiband one-dimensional line sensor camera |
| CN102158624A (en) * | 2011-03-02 | 2011-08-17 | 朱杰 | Double-sided scanning flatbed scanner |
| CN105842173A (en) * | 2016-06-06 | 2016-08-10 | 南京大学 | Method for identifying hyperspectral material |
| CN106454000A (en) * | 2016-11-18 | 2017-02-22 | 桂林电子科技大学 | Multifunctional scanner based on image acquisition device |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112161936A (en) * | 2020-09-29 | 2021-01-01 | 杭州海康威视数字技术股份有限公司 | Optical inspection device for cigarette packets |
| CN112161935A (en) * | 2020-09-29 | 2021-01-01 | 杭州海康威视数字技术股份有限公司 | Optical inspection device for cigarette packets |
| CN112161936B (en) * | 2020-09-29 | 2024-05-24 | 杭州海康威视数字技术股份有限公司 | Optical detection device for cigarette packets |
| CN112161935B (en) * | 2020-09-29 | 2024-08-23 | 杭州海康威视数字技术股份有限公司 | Optical detection device for cigarette packets |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sharma et al. | Hyperspectral CNN for image classification & band selection, with application to face recognition | |
| Polak et al. | Hyperspectral imaging combined with data classification techniques as an aid for artwork authentication | |
| Syarif et al. | Enhanced maximum curvature descriptors for finger vein verification | |
| Zou et al. | Illumination invariant face recognition: A survey | |
| US9639748B2 (en) | Method for detecting persons using 1D depths and 2D texture | |
| CN111046703A (en) | Face anti-counterfeiting detection method and device and multi-view camera | |
| CN104102913B (en) | Wrist vena identification system | |
| Luque-Baena et al. | Assessment of geometric features for individual identification and verification in biometric hand systems | |
| Guo et al. | Empirical study of light source selection for palmprint recognition | |
| US11341348B2 (en) | Hand biometrics system and method using digital fingerprints | |
| Wang et al. | Fake modern Chinese painting identification based on spectral–spatial feature fusion on hyperspectral image | |
| CN110637224A (en) | Information search system and program | |
| Mokni et al. | Combining shape analysis and texture pattern for palmprint identification | |
| Chaki et al. | Image color feature extraction techniques: fundamentals and applications | |
| CN109214302A (en) | One kind being based on multispectral substance identification | |
| CN108965636A (en) | A kind of flatbed scanner and its application | |
| Ledoux et al. | Perceptual color hit-or-miss transform: application to dermatological image processing | |
| CN114216867A (en) | Hyperspectral image acquisition and identification device and method | |
| CN116665051B (en) | Method for rescreening metals in garbage based on RGB image reconstruction hyperspectral image | |
| Sarath et al. | A Study on Hyperspectral Remote Sensing | |
| CN116883851A (en) | Calligraphy and painting artwork identification method based on hyperspectral image comparison | |
| CN109858577A (en) | Soybean appearance quality detection device and detection method | |
| Peng et al. | A methodology for palm vein image enhancement and visualization | |
| Zhang et al. | A deep neural network and rule-based technique for fire risk identification in video frames | |
| CN107944426B (en) | Wheat leaf powdery mildew spot marking method based on combination of texture filtering and two-dimensional spectral feature space discrimination |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181207 |
|
| RJ01 | Rejection of invention patent application after publication |