CN104504639A - Method for realizing digital watermark anti-counterfeiting trademark and detector thereof - Google Patents

Abstract

The invention discloses a method for realizing a digital watermark anti-counterfeiting trademark and a detector thereof, and provides a digital watermark method which combines discrete wavelet transform with discrete cosine transform. The method exhibits good transparency, robustness and viability. The detector which is simple in structure and convenient to use and carry is realized. The hardware of the detector comprises a CMOS (Complementary Metal-Oxide-Semiconductor Transistor) image sensor, a DSP (Digital Signal Processor) chip and a power circuit, wherein the DSP chip adopts a DaVinci-technology DSP chip. The DaVinci-technology DSP chip is used as core development hardware to realize seamless connection between the CMOS image sensor and the DSP, so that the complexity of a detector composition is lowered. Meanwhile, by aiming at the anti-counterfeiting particularity of the trademark, an improved watermark pick-up algorithm and a trademark authenticity identification method are adopted, a detection method of the detector is independently developed, and the detector which has the characteristics of being advanced in technology, high in identification rate, compact in structure and convenient in use is realized.

Description

A kind of digital watermark anti-fake trade mark implementation method and detector thereof

Technical field

The invention belongs to general purpose image data processing technology field, relate to antiforge technology, particularly relate to print domain trademark anti-counterfeit, realize a kind of portable instrumentation of digital watermarking trade mark truth identification.

Background technology

At present, the problem of the printed matter piracies such as certificate, bill, trade mark is more and more serious, and particularly in medicine, food, tobacco and wine, field of beverage, personation packaging is more ferocious sticks up, and often makes people hard to guard against.Bring heavy losses not only to country and enterprise, the economic development of impact society, also directly jeopardizes the health and lives of people.

The major technique adopted in order to the effect playing trademark anti-counterfeit in prior art has: bar codes technique, laser antifalsification technology and electronic trademark.False proof and the laser anti-counterfeit for bar code, due to its core technology wide-scale distribution, causes some enterprises to adopt these technology to manufacture counterfeit trademark on the contrary, these technology are all that observation class is false proof in addition, general is all go to tell truth from falsehood by the vision of people, is convenient to copy, and reliability is not high; Electronic trademark is false proof based on platform of internet of things, and system constructing is complicated.In addition, even if achieve the omnidistance transparent management to commodity circulation by Internet of Things, also still entering of Counterfeit Item cannot be stopped.Counterfeit Item also can carry imitated false electronic trademark and come into the market, because technical progress achievement can be shared, fake producer can crack the cryptographic means of electronic trademark completely, plants the ill-gotten monovalent products identity code of people in false electronic trademark, makes that radio-frequency identification reader is true and false not to be distinguished.Briefly, even if enter the electronic trademark epoch exactly, still there is the problem of personation electronic trademark.

That is, although the various anti-counterfeiting technologies existed at present serve false proof effect to a certain extent, still can not effectively prevent trade mark from palming off the generation of thing and spreading unchecked.Thus the anti-counterfeiting technology promoting the use emerging advanced person is the important means of containment fraud economic criminal activity.

Digital watermarking (Digital Watermarking) technology refers to and in digitized multi-medium data, embeds hidden some identification informations (i.e. digital watermarking) by the method for signal transacting, these identification informations are normally sightless, only have and could be extracted by special detecting device or reader.Digital watermarking is an important research direction of Information Hiding Techniques.

Digital watermarking trademark anti-counterfeit principle of work

Make the anti-false trademark containing digital watermarking, and utilize digital watermark anti-fake trade mark detector to detect its true and false, its process is as follows.

(1) generation of anti-false trademark

By Digital Watermarking Embedded Algorithm, under the control of key K, be embedded into by digital watermark information in the original trademark image that will protect, generate moisture branding image, its theory diagram generated as shown in Figure 1.

Its concrete generative process is, by digital watermark signal W={w (k) } be embedded into original trademark image X={x (k) } in, generate the trademark image X containing watermark _w={ x _w(k) }.Digital Watermarking Embedded Algorithm can select attack resistance algorithm.

(2) true and false of anti-false trademark detects

By key K and digital watermarking extraction algorithm (original trademark image will be used in non-blind watermarking algorithm, and for blind watermarking algorithm, then not need original trademark image), watermark signal contained in the trademark image containing watermark can be extracted.Again by calculating the degree of correlation of watermark signal and the original watermark signal extracted, can differentiate the true and false of trade mark, its true and false testing process as shown in Figure 2.

Digital watermark anti-fake technology has the advantage of self uniqueness, and such as invisibility, information diversity, anti-counterfeiting technology upgrading are fast, high-reliability and high security, cost are low.It is information security field one new have challenging research direction.Therefore, advance digital watermarking trademark anti-counterfeit, the anti-false trademark detector of exploitation digital watermarking, has extremely important social effect and economic worth.

Through retrieval, published patent correlative study invention mainly contains: (1) " print domain trademark anti-counterfeit method based on digital watermark technology ", publication number CN1885342A, disclose a kind of print domain trademark anti-counterfeit method based on digital watermark technology, comprise the generation of anti-false trademark, the generation of self registration template, trade mark authenticity process.But the embedding grammar of watermark is copied in the two-dimensional random position of this article introduction, belong to spatial domain watermark algorithm, the disguise of embed watermark is good not, and the watermark of generation has locality, be difficult to common attack and noise in opposing image procossing, therefore robustness is poor.(2) " embedding grammar of printed matter anti-counterfeit digital water print and detection method thereof ", publication number CN101071499A, describe the singular value of computed image subchannel matrix, then on corresponding subchannel, superpose the embedding grammar of the printed matter anti-counterfeit digital water print of watermarking images and the detection method with digital watermark embedding contrary, but based on the digital watermarking algorithm of svd, anti-shearing attack is not strong, this is because singular value decomposition algorithm itself does not have geometric distortion self-recognizability, does not have that shear resistant causes.(3) " the fake certification device and method based on radio RF recognition technology and digital watermarking ", publication number CN102073962A, describe a kind of demo plant constructive method for differentiating the consumer goods true and false with electronic tag and digital watermarking label, device is by load module, RFID read-write module, image capture module, data processing module, control module, memory module, output module, communication module forms, but just a kind of demo plant, the formation of device is comparatively complicated, selecting simultaneously to Primary Component in device, and adopt which kind of embedding algorithm and extractive technique not to state.

In sum, in the prior art, also can not be applicable to antiforge have that anti-attack ability is strong, confidentiality is high, compact conformation, miniature portable, the feature such as easy to use detecting instrument.

Summary of the invention

The problem to be solved in the present invention is: prior art, for digital watermark anti-fake technology, does not have the pick-up unit of miniature portable, provide a kind of and have that anti-attack ability is strong, confidentiality is high, compact conformation, miniature portable, the feature such as easy to use detecting instrument.

Technical scheme of the present invention is: a kind of digital watermark anti-fake trade mark implementation method, the method that described embedding algorithm adopts wavelet transform to combine with discrete cosine transform, and process is:

1) first, one-level wavelet transformation is carried out to original trademark image, be 4 subgraphs picture breakdown, respectively: low frequency ll channel LL, level detail subgraph HL, vertical detail subgraph LH and diagonal line details subgraph HH, its medium and low frequency ll channel is the low frequency part of original trademark image, level detail subgraph HL, vertical detail subgraph LH and diagonal line details subgraph HH maintains the horizontal edge details of original trademark image respectively, vertical edge details and diagonal angle edge details, it is the HFS of original trademark image, after wavelet transformation, obtain the low frequency ll channel coefficient F of image _l,

2) embed watermark information in low frequency ll channel coefficient, adopts the method for superposition by watermark embedment to ll channel coefficient F _lin, concrete formula is F ' _l=F _l+ α W, in formula, W is watermark information W (j) that will embed, and α is watermark embed strength, and its span is 0 to 0.5;

3) overall discrete cosine transform and Zigzag sequence are carried out to the image of embed watermark information, obtain one group of coefficient by the arrangement of frequency ascending order, to extract wherein L low frequency coefficient as the proper vector U (j) comprising image feature information;

4) use Hash function, generate key K (j), i.e. K (j)=U (j) ⊕ W (j) according to proper vector U (j) and watermark information W (j);

5) according to proper vector U (j), to step 3) discrete cosine transform after newly-generated image carry out inverse discrete cosine transform, obtain the image of transform domain containing watermark, namely new LL subgraph, again discrete wavelet inverse transformation is done to it, obtain the trademark image containing watermark information.

Based on the implementation method of the detector of above-mentioned digital watermark anti-fake trade mark, the hardware of detector comprises cmos image sensor, dsp chip, DDR2 SDRAM storer, FLASH memory, LCD display and power circuit, cmos image sensor, DDR2 SDRAM storer, FLASH memory are connected with dsp chip respectively with LCD display, power circuit is powered, and wherein dsp chip adopts Davinci technology dsp chip;

Dsp chip carries out the process of following extraction watermark to the trademark image of cmos image sensor collection:

First wavelet transform is carried out to the trademark image containing watermark information, again discrete cosine transform is carried out to the low frequency ll channel coefficient that wavelet transform obtains, obtain the proper vector U ' (j) of trademark image to be measured, then according to the key K (j) generated in watermark embed process and the proper vector U ' (j) asked for, Hash function is utilized to obtain watermark information W ' (j), i.e. W ' (j)=K (j) ⊕ U ' (j), normalized correlation coefficient NC is calculated according to original watermark information W (j) and watermark information W ' (j) calculated, and whether judge in tested trademark image containing original watermark information according to NC, realize anti-false trademark to detect.

Further, chromatogram Hydrochloric acid is adopted to be replicated to prevent digital watermark information when judging the true and false of trade mark, be specially: by the digitized processing to trade mark chromatogram, calculate the relation with contents of the chromatogram equivalent of trade mark to be detected and the chromatogram equivalent of standard label sample, with this judge whether be copy after trade mark, thus Certification Mark detect credibility.

As optimal way, dsp chip adopts the DM6437 of TI company, obtain cmos image sensor by the Video processing front end VPFE of the inner integrated video processing subsystem assembly VPSS of dsp chip and gather trademark image data, and by video processing subsystem assembly VPSS, pre-service is carried out to the digital picture gathered, comprise rotation and Slant Rectify, pre-processed results is deposited in SDRAM, then dsp chip carrying out gray processing to view data again, except making an uproar and colour correction process, then extracting watermark information by watermark extraction algorithm.

Method part of the present invention is implemented as the trace routine be carried in detector, after anti-false trademark detector powers on, trace routine loads and brings into operation from FLASH, when detecting digital watermark anti-fake trade mark, start anti-false trademark to detect, first dsp chip controls cmos image sensor and gathers trademark image, and the Video processing front end VPFE passing through the inner integrated video processing subsystem assembly VPSS of dsp chip obtains cmos image sensor collection trademark image data, and pre-service is carried out to the digital picture gathered, result is deposited in DDR2 SDRAM, then dsp chip carries out gray processing to view data again, except making an uproar, correction process, watermark information is extracted again by watermark extraction algorithm, finally do degree of correlation computing, judge the true and false of trade mark, testing result shows on an lcd display.

The present invention for core develops hardware, realizes the seamless link between cmos image sensor and DSP with the dsp chip of Davinci technology, reduces the complicacy that detecting instrument is formed.Simultaneously, for the singularity of trademark anti-counterfeit, adopt the digital watermarking extraction algorithm and trade mark authenticity identification method that improve, the independent development detection method of detecting instrument, the present invention starts with from detection method and detecting instrument two aspect, coordinated by software and hardware achieve that a kind of advanced person that possesses skills, discrimination are high, the detecting instrument of compact conformation, the feature such as easy to use.

Accompanying drawing explanation

Fig. 1 is the theory diagram that digital watermark anti-fake trade mark generates.

Fig. 2 is the true and false testing process of digital watermark anti-fake trade mark.

Fig. 3 is the original trademark image of digital watermarking example.

Fig. 4 is the original watermark image of digital watermarking example.

Fig. 5 is the generation trademark image of digital watermarking example.

Fig. 6 be digital watermarking example not under fire time extract watermarking images.

Fig. 7 is the watermarking images received after shearing attack of digital watermarking example.

Fig. 8 is the watermarking images under fire extracted afterwards of digital watermarking example.

Fig. 9 is the structural representation of detector of the present invention.

Figure 10 is the process flow diagram of the detection method of detector of the present invention.

Figure 11 is image one-level wavelet decomposition figure.

Embodiment

The present invention proposes a kind of simple efficient, digital watermark method that attack tolerant is strong, and combine and achieve the detector that structure is simply easy to use and carry, first introduce digital watermarking algorithm of the present invention below.

For the feature of trademark anti-counterfeit, the present invention proposes the method that a kind of wavelet transform (Discrete Wavelet Transform, DWT) combines with discrete cosine transform (Discrete Cosine Transform, DCT).

The method combines DCT watermarking algorithm strong robustness, good concealment and DWT watermarking algorithm good visual effect, feature that anti-attack ability is strong, makes it to have the good transparency, robustness and robustness.

(1) DWT_DCT watermarking algorithm telescopiny

1. first, carrying out one-level wavelet transformation to original trademark image, is 4 subgraphs picture breakdown, respectively: low frequency ll channel LL, and level detail subgraph HL, vertical detail subgraph LH and diagonal line details subgraph HH, as shown in figure 11.Wherein, low frequency ll channel has concentrated most energy of original trademark image, is also the non-sensitive part of vision, is the low frequency part of original trademark image; Level detail subgraph HL, vertical detail subgraph LH and diagonal line details subgraph HH maintain the horizontal edge details of original trademark image, vertical edge details and diagonal angle edge details respectively, are the HFSs of original trademark image.After wavelet transformation, the low frequency ll channel coefficient F of image can be obtained _l.

2. embed watermark information in low frequency ll channel coefficient.The method of superposition can be adopted by watermark embedment to ll channel coefficient F _lin, concrete formula is F ' _l=F _l+ α W.In formula, W is the watermark information that will embed, and α is watermark embed strength, and its span is 0 to 0.5.

3. overall dct transform and Zigzag sequence are carried out to the image of embed watermark, obtain one group of coefficient by the arrangement of frequency ascending order, to extract wherein L low frequency coefficient as the proper vector U (j) comprising image feature information.

4. use Hash function, generate key K (j), i.e. K (j)=U (j) ⊕ W (j) according to proper vector and watermark information.According to proper vector, IDCT (Inverse Discrete Cosine Transform) conversion is done to the image after overall dct transform, then do IDWT (Inverse Discrete Wavelet Transform) conversion, thus obtain the trademark image containing watermark information.

(2) DWT_DCT watermark extracting and detection

1. first, DWT conversion is carried out to the trademark image containing watermark information, obtain low frequency ll channel coefficient.

2. dct transform is carried out to the low frequency ll channel coefficient obtained after DWT conversion, and carry out Zigzag scanning to obtaining DCT coefficient matrix after dct transform, obtain carrying out by frequency the DCT coefficient matrix that arranges from low to high, L low frequency coefficient before this matrix of coefficients is obtained the proper vector U ' (j) of trademark image to be measured through symbolic operation.

3. according to the key K (j) generated in watermark embed process and the proper vector U ' (j) asked for, Hash function is utilized to obtain watermark information W ' (j), i.e. W ' (j)=K (j) ⊕ U ' (j).And then calculate normalized correlation coefficient NC according to original watermark information W (j) and watermark information W ' (j) calculated, and whether judge in tested trademark image containing watermark information according to NC.

It should be noted that, before execution DWT_DCT watermark extraction algorithm, certain pre-service and correction (as rotated, slant correction etc.) will being carried out, to improve the accuracy rate of testing result to trademark image to be detected.

The validity function of current resist geometric attacks digital watermarking mainly contains: PSNR (Peak Signal-to-NoiseRatio, Y-PSNR) and NC (Normalized Correlation, normalized-cross-correlation function).

Y-PSNR is the important evaluation and test function weighing the image after embedding people's watermark and original image mesopic vision deviation quality, in the overall stochastic error of evaluation, there is good performance, it defines by original image and containing the MSE (MeanSquare Error, square error) between watermarking images.

If original image matrix A={ f (i, j) }, 1≤i≤M, 1≤j≤N, wherein f (i, j) is image pixel value, and M, N are respectively the row, column pixel value size of image.Embed watermark in image A also carries out Image Reconstruction, and obtain containing watermarking images B=f* (i, j), the square error MSE computing formula between image A and B is

$\mathrm{MSE}=\frac{1}{M..N}\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{J=1}{\overset{N}{\mathrm{\Σ}}}{\left|\right[f(i,j)-f^{*}(i,j)\left]\right|}^2$

If f _mAXfor the gray level maximal value of image slices vegetarian refreshments, Y-PSNR PSNR computing formula is

$\mathrm{PSEN}=10{\log }_{10}\left(\frac{F_{\mathrm{MAX}}^2}{\mathrm{MSE}}\right)=10{\log }_{10}\left(\frac{F_{\mathrm{MAX}}^2}{\frac{1}{M\·N}\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{\left|\right[f(i,j)-f^{*}(i,j)\left]\right|}^2}\right)$

The unit db of Y-PSNR calculates, and the larger picture quality of value of signal to noise ratio (S/N ratio) is better, and when signal noise ratio (snr) of image is greater than 30db, image there will not be and obviously degrades, and can reach good visual quality.

Normalized-cross-correlation function is the important indicator weighed original watermark and extract correlativity between watermark, if original watermark is W, is W from containing the watermark of extracting watermarking images ^*, the calculating formula of normalized correlation coefficient NC is

$\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}W(i,j)W^{*}(i,j)}{\sqrt{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}W{(i,j)}^2\·\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{W}^{*}{(i,j)}^2}}$

Normalized-cross-correlation function more levels off to 1, then former watermark is more relevant to extraction watermark.

If original trademark image as shown in Figure 3, selected original watermark image as shown in Figure 4, adopt Digital Watermarking Embedded Algorithm of the present invention, namely select discrete cosine transform (DCT) and wavelet transform (DWT) hybrid digital watermarking algorithm, its anti-false trademark image generated as shown in Figure 5.Not under fire time extract watermarking images as shown in Figure 6, its performance evaluation parameters PSNR=29.6285, NC=1; Trademark image be subject to 1/4 Partial Shear attack after, as shown in Figure 7, the watermarking images of extraction as shown in Figure 8, now, its performance evaluation parameters PSNR=7.5120, NC=0.9985.Difference obviously.

In the application process of digital watermark anti-fake trade mark, in order to detect its true and false, suitable detecting instrument is extremely important.The digital watermark anti-fake trade mark detector composition structure formed based on imageing sensor, High Performance DSP and display of the present invention as shown in Figure 9.

Detecting instrument is primarily of cmos image sensor, DSP (DM6437), DDR2 SDRAM, the compositions such as FLASH, LCD display, power circuit, cmos image sensor, DDR2 SDRAM, FLASH is connected with dsp chip respectively with LCD display, and power circuit is powered.The course of work of detecting instrument is: after instrument powers on, and program loads and brings into operation from FLASH.When detecting digital watermark anti-fake trade mark, start anti-false trademark by button to detect, first DSP controls cmos image sensor and gathers trademark image, and the Video processing front end (VPFE) passing through the inner integrated video processing subsystem assembly (VPSS) of DSP obtains cmos image sensor collection trademark image data, and pre-service is carried out to the digital picture gathered, result is deposited in SDRAM, then DSP carries out gray processing to view data again, except making an uproar, the process such as correction, watermark information is extracted again by watermark extraction algorithm, finally do degree of correlation computing, judge the true and false of trade mark, testing result can show on an lcd display.

In the composition of digital watermark anti-fake trade mark detector, the key property of main modular is described below.

(1) cmos image sensor

The OV7660 carrying plastic cement camera lens that cmos image sensor adopts Omni vision company of the U.S. to produce.OV7660 is the complete programmable graphics acquisition module chip of a low-voltage, high integration, 300,000 pixel images (640 × 480), and digital video output terminal provides the data output format of 8.OV7660 can reach 30 frames speed per second, and can be controlled the transmission of picture quality, form and output data completely by user.All image processing functions (comprising exposure control, gamma correction, white level balance, color saturation, tone control etc.) can pass through SCCB (Serial Camera Control Bus, serial camera control bus) and use programmed control.The video timing sequence generating circuit of OV7660 can produce the multiple internal clock signals such as multiple synchronization signal and pixel clock such as row is synchronous, field synchronization, mixed video are synchronous.Peripheral control unit is by I ²c bus interface carries out the duty, the working method and data output format etc. that arrange or read OV7660.In addition, utilize the sensor technology of Omni Vision company advanced person, more clear, complete, stable coloured image can also be produced by the common light/electric resources (as fixed pattern noise, smearing, flash of light etc.) in minimizing or removal of images flaw station.

(2)DSP(DM6437)

DSP is the core component of whole instrument system, the digital video processor DM6437 of special 32 fixed DSP Leonardo da Vinci (DaVinci) technology for high-performance, the exploitation of low cost Video Applications selecting TI company to release.The dominant frequency of DM6437 is up to 600MHz, comprises 8 processing units, and under the clock frequency of 600MHz, when in sheet, 8 processing units run simultaneously, its arithmetic speed can reach 4800MIPS, extremely meets the system requirements that image procossing is complicated like this.It provides higher handling property with lower price.

The feature of Davinci technology dsp chip is, chip is integrated with a video processing subsystem VPSS, is exclusively used in video acquisition and display, has great convenience for the user to carry out video acquisition and video display.

DM6437 video processing subsystem VPSS is made up of 2 parts: one is Video processing front end (VPFE), for defeated people's digital of digital video data, for the digital video input of multiple standards provides interface, and it is the pre-service of digital of digital video data work necessity of defeated people; Two is Video processing rear end (VPBE), for output video data, with driving display display video image.

Utilize DM6437 Video processing front end (VPFE) directly can be connected with cmos image sensor, complete the acquisition function of trademark image.

(3) DDR2 SDRAM and FLASH memory expansion

To in the acquisition and processing process of trademark image, need a large amount of storage of storage space in order to supported data and the loading of program, and DM647 internal storage space has been not enough to this function, therefore need the memory span being improved system by the mode of outside extended memory.Outer extension memory comprises two parts: one is the DDR2 storer for storage system program and view data, and two is the FLASH memory for cure system program code and algorithm configuration information.

The DDR2 SDRAM of JESD79D-2A standard supported by the DDR2 controller of DM6437, and extending out max cap. is 256M.External memory interface (EMIF) in DM6437 supports the interface of multiple memorizers or external unit, comprises SRAM, NOR FLASH and NAND FLASH etc.

After system electrification, program starts from FLASH, and runs in loading system program to DDR2, and the view data gathered in system operation is all stored in DDR2 storer, and the algorithm configuration information that user downloads need keep power-off not lose, can stored in FLASH memory.

(4) clock, reset, configuration and JTAG debugging interface

Clock, reset, configuration and JTAG debugging interface are that DM6437 normally works the peripheral circuit that must configure.

Needed for DM6437, clock is produced by external crystal-controlled oscillation, is accessed by CLK1N pin, and nominal frequency can be selected to be the external crystal-controlled oscillation of 27MHz.DM6437 has an inner PLL, and can do frequency multiplication to produce internal clocking to input clock, PLL frequency multiplier is by software design patterns.

In order to ensure that DSP can enter normal duty, reset operation is absolutely necessary part, by building external hand reset circuit, can produce the reset signal that the reset pin (RST) that meets DM6437 requires, DM6437 is resetted.

Can the mode of operation of control DSP by configuration circuit, namely configure the boot pattern needed when DSP starts to perform.

DM6437 chip, with the jtag interface meeting IEEEI149.1 standard, can connect DSP emulator by this interface, debugging emulation instrument system.

(5) button, LCD display and communication interface

Can the trademark image collection of start detection instrument and identifying by button, and by the LCD display display of testing result by expansion.In addition, under some applied environment, also by the communication interface (serial communication, network service) of expansion, testing result is sent to remote control computer display, preservation etc.

(6) power circuit

Power supply provides energy for whole detecting instrument, is the basis of whole instrument works fine.Particularly using DM6437 as the instrument system of primary processor, need for DSP provides high precision, the power supply of high stability.

In this detecting instrument, consider that DM6437 needs 1.2V (for DSP core), 1.8V (for DDR2 interface and internal memory), 3.3V (the I/O buffer interface for DSP) three kinds of supply voltages, OV7660 needs two kinds of operating voltage (analog operation voltage 2.5V, digital signal voltage 1.8V), at the power requirement in conjunction with other interface chips, detecting instrument adopts 5V Power supply, then produces 1.2V, 1.8V, 2.5V, 3.3V tetra-kinds of supply voltages by power circuit.

The characteristic part that detecting instrument hardware of the present invention is formed is: through many-sided selection, and final choice is core forms instrument based on the DSP of Davinci technology, reduces the complicacy of hardware composition.Meanwhile, make full use of the pretreater in video front input (VPFF) interface of Leonardo da Vinci DSP, reduce the complexity of software, ensure that the image information of high speed, high-quality acquisition trade mark to be checked from hardware view.

In detecting instrument, hardware is basis, and detection method is crucial, and it embodies the functional realiey of whole instrument system.Detection method of the present invention is embodied as concrete software program, and software adopts modular design, mainly comprises trademark image acquisition module, pre-processing image data module, digital watermark detecting module and trade mark true and false judge module etc.The workflow of detecting instrument as shown in Figure 10.Key point in software simulating is: the functional configuration of DSP, watermarking algorithm and trade mark authenticating method.

Leonardo da Vinci's chip is the advanced of TMS320C6000 series, carries out the configuration that software development relates to operating system, inner function module, administering digital video component etc., programming process relative complex to it.In order to ensure the steady operation of detecting instrument, the CCD controller (CCDC), DDR2 controller, EMIF port etc. of reasonably configuration DM6437 Video processing front end (VPFE) are just extremely important.And course of work design of the present invention fully ensure that this point, obtain cmos image sensor by the Video processing front end VPFE of the inner integrated video processing subsystem assembly VPSS of dsp chip and gather trademark image data, and by video processing subsystem assembly VPSS, pre-service is carried out to the digital picture gathered, comprise rotation and Slant Rectify, pre-processed results is deposited in SDRAM, then dsp chip carrying out gray processing to view data again, except making an uproar and colour correction process, then extracting watermark information by watermark extraction algorithm.

In trade mark true and false testing process, except utilizing the parameters such as normalized-cross-correlation function to the correlativity judging original watermark and extract between watermark, judge outside the true and false of trade mark with this, be replicated to prevent digital watermark information, the present invention proposes a kind of chromatogram Hydrochloric acid, namely by the digitized processing to trade mark chromatogram, calculate the relation with contents of the chromatogram equivalent of trade mark to be detected and the chromatogram equivalent of standard label sample, with this judge whether be copy after trade mark, thus Certification Mark detect credibility.

The invention provides a kind of portable detector for trademark anti-counterfeit, prior art is reported for work there are no relevant, the solution of the present invention coupling apparatus and method joint research, through many-sided selection, coordinate and debugging efforts, finally obtain applicability of the present invention wide, structure is simple, be easy to carry the trademark anti-counterfeit detecting instrument used, except can be used as medicine, food, tobacco and wine, beverage, the detecting instrument of the digital watermark anti-fake trade mark of the commodity such as cosmetics is with outward, also can be used for athletic meeting, the digital watermark anti-fake of the admission tickets such as park detects, and certificate, the digital watermark anti-fake of bill detects, application prospect is very wide, there is important social value and economic worth.

Claims (4)

1. a digital watermark anti-fake trade mark implementation method, it is characterized in that the method that described embedding algorithm employing wavelet transform combines with discrete cosine transform, process is:

1) first, one-level wavelet transformation is carried out to original trademark image, be 4 subgraphs picture breakdown, respectively: low frequency ll channel LL, level detail subgraph HL, vertical detail subgraph LH and diagonal line details subgraph HH, its medium and low frequency ll channel is the low frequency part of original trademark image, level detail subgraph HL, vertical detail subgraph LH and diagonal line details subgraph HH maintains the horizontal edge details of original trademark image respectively, vertical edge details and diagonal angle edge details, it is the HFS of original trademark image, after wavelet transformation, obtain the low frequency ll channel coefficient F of image _l,

2) embed watermark information in low frequency ll channel coefficient, adopt the method for superposition by watermark embedment in ll channel coefficient FL, concrete formula is F ' _l=F _l+ α W, in formula, W is watermark information W (j) that will embed, and α is watermark embed strength, and its span is 0 to 0.5;

3) overall discrete cosine transform and Zigzag sequence are carried out to the image of embed watermark information, obtain one group of coefficient by the arrangement of frequency ascending order, to extract wherein L low frequency coefficient as the proper vector U (j) comprising image feature information;

4) use Hash function, generate key K (j), i.e. K (j)=U (j) ⊕ W (j) according to proper vector U (j) and watermark information W (j);

5) according to proper vector U (j), to step 3) discrete cosine transform after newly-generated image carry out inverse discrete cosine transform, obtain the image of transform domain containing watermark, namely new LL subgraph, again discrete wavelet inverse transformation is done to it, obtain the trademark image containing watermark information.

2. based on the implementation method of the detector of digital watermark anti-fake trade mark described in claim 1, it is characterized in that the hardware of detector comprises cmos image sensor, dsp chip, DDR2SDRAM storer, FLASH memory, LCD display and power circuit, cmos image sensor, DDR2SDRAM storer, FLASH memory are connected with dsp chip respectively with LCD display, power circuit is powered, and wherein dsp chip adopts Davinci technology dsp chip;

Dsp chip carries out the process of following extraction watermark to the trademark image of cmos image sensor collection:

First wavelet transform is carried out to the trademark image containing watermark information, again discrete cosine transform is carried out to the low frequency ll channel coefficient that wavelet transform obtains, obtain the proper vector U ' (j) of trademark image to be measured, then according to the key K (j) generated in watermark embed process and the proper vector U ' (j) asked for, Hash function is utilized to obtain watermark information W ' (j), i.e. W ' (j)=K (j) ⊕ U ' (j), normalized correlation coefficient NC is calculated according to original watermark information W (j) and watermark information W ' (j) calculated, and whether judge in tested trademark image containing original watermark information according to NC, realize anti-false trademark to detect.

3. the implementation method of the detector of digital watermark anti-fake trade mark according to claim 2, chromatogram Hydrochloric acid is adopted to be replicated to prevent digital watermark information when it is characterized in that the true and false judging trade mark, be specially: by the digitized processing to trade mark chromatogram, calculate the relation with contents of the chromatogram equivalent of trade mark to be detected and the chromatogram equivalent of standard label sample, with this judge whether be copy after trade mark, thus Certification Mark detect credibility.

4. the implementation method of the detector of the digital watermark anti-fake trade mark according to Claims 2 or 3, it is characterized in that dsp chip adopts the DM6437 of TI company, obtain cmos image sensor by the Video processing front end VPFE of the inner integrated video processing subsystem assembly VPSS of dsp chip and gather trademark image data, and by video processing subsystem assembly VPSS, pre-service is carried out to the digital picture gathered, comprise rotation and Slant Rectify, pre-processed results is deposited in SDRAM, then dsp chip carries out gray processing to view data again, except making an uproar and colour correction process, watermark information is extracted again by watermark extraction algorithm.