CN108921769A - A kind of 3D grid blind watermatking generation method based on layering pseudomorphism analysis - Google Patents

Abstract

A kind of 3D grid blind watermatking generation method based on layering pseudomorphism analysis, including step：Using 3D polygonal mesh M={ V, E } as input covering archetype, and axis aligned pretreatment is carried out to archetype；Watermark recovery is generated using watermark keys and pseudo-random generator；Spread-spectrum watermark algorithm is obtained by the location information based on vertex discrete in circular cylindrical coordinate, and watermark insertion is carried out to pretreated archetype using watermark recovery；Model Reconstruction is carried out to obtain watermarking model to the archetype of watermark insertion.The present invention is based on robustness and not visible property that the watermark that the 3D grid blind watermatking generation method of layering pseudomorphism analysis is realized has reply adverse circumstances; it protects 3D model content when digital field is shared, but also can also be protected when 3D digital content is converted to analog content by 3D printing.

Description

3D grid blind watermark generation method based on layered artifact analysis

Technical Field

The invention relates to the technical field of 3D printing, in particular to a 3D grid blind watermark generation method based on layered artifact analysis.

Background

In the technical field of 3D printing, one technical difficulty is the copyright protection problem of 3D model contents. There is an urgent need for a powerful and efficient technique to avoid unauthorized copying, tampering, and illegal distribution, thereby providing protection to intellectual property rights. Currently, 3D model providers try to protect 3D model content through fingerprinting, Digital Rights Management (DRM), access control, encryption and digital watermarking techniques, with the problem that encryption and DRM based 3D model content protection is ineffective because the 3D printing process disables these protections. Therefore, the 3D model contents can be illegally copied not only on the internet but also in the market and redistributed online. For this reason, research and development of a digital watermarking technology for protecting 3D printing are required to ensure prosperity of the 3D content industry.

Digital watermarking technology is a process of hiding digital information in a signal such as multimedia data, and when copyright disputes occur, a watermark can be used to determine the identity of an author, and when a prototype is revealed, it can be used as a fingerprint to trace a distribution path. In addition, digital watermarking can also be used as an active component of an automatic system for managing unauthorized users in a content sharing environment. Therefore, the digital watermark of the 3D model should be embedded secretly into the 3D model content before distribution, and furthermore, the embedded watermark must resist attacks that may violate copyright.

At present, various watermarking methods have been proposed to protect a 3D mesh model for 3D printing, however, the current three-dimensional model watermarking technology is applied only to medical images, geographic information, and Computer Aided Design (CAD) data, etc. Most of these efforts are focused on conventional attacks during normal operation of the digital domain, while the offline distribution environment of 3D content is rarely considered, since 3D printing processes, there are situations of distribution and handling both online and offline, and 3D printed objects have new requirements for 3D model watermarking in terms of robustness.

Disclosure of Invention

In order to solve the problems in the prior art, the invention provides a 3D grid blind watermark generation method based on layered artifact analysis, so as to solve the technical problem that copyright is violated because the existing 3D model content is illegally copied not only on the Internet, but also illegally copied and redistributed in the online market.

In order to achieve the above object, the present invention provides a 3D mesh blind watermark generation method based on layered artifact analysis, comprising the following steps:

s1, covering the original model by taking the 3D polygonal grid M ═ { V, E } as input, and carrying out axis alignment pretreatment on the original model;

s2, generating a watermark pattern by adopting a watermark key and a pseudo-random generator;

s3, acquiring a spread spectrum watermarking algorithm based on the position information of the discrete vertex in the cylindrical coordinate, and embedding the watermark into the preprocessed original model by adopting a watermarking pattern;

and S4, carrying out model reconstruction on the original model of the embedded watermark to obtain a watermark model.

As a further preferable embodiment of the present invention, in step S1, the preprocessing of overlaying the original model with the 3D polygonal mesh M ═ { V, E } as input, and performing axis alignment on the original model specifically includes:

s11, determining the z-axis of watermark embedding, and adjusting the direction of the grid M, wherein the adjustment calculation formula is as follows:

wherein v is_iE.g. V andis a rotation matrix and defines the following equation:

wherein, the ratio of theta,is the relative degree of the predefined printing direction, the z-axis being the orthogonal direction determined to be the 3D printed layer;

s12, mixing v_i＝(x_i，y_i，z_i) Classifying as vertexes L different bins having a range equal to the z-axis value, calculating in advance the maximum and minimum values of the z-axis, and dividing the values by z_max，z_minDenotes the L-th layer L_lCalculated by the following formula:

L_l＝{v_i|z_min+δ_z·l<z_i<z_min+δ_z·(l+1)} (3)

wherein L is more than or equal to 1 and less than or equal to L, i is more than or equal to 1 and less than or equal to V, j is more than or equal to 1 and less than or equal to VI, and V_lIndicates to belong to L_lUsing the symbol v (l, j) representing the jth vertex of the ith bin instead of v_i∈L₁Thickness delta_zByDefining the number of bins in the table as a constant selected by a user;

s13, calculating the central position of each LThe calculation formula is as follows:

through the center position c_lFor its vertex v (L, j) ∈ L_lThe position of (2) is normalized as follows:

wherein,denotes the L-th layer L_lIs the p, q, r order moment of volume, each point v (L, j) is epsilon to L_lTo overlay the sliced layers of the model M, the overlay model is sliced into L layers, each layer centered on the z-axis.

In a further preferred embodiment of the present invention, the watermark pattern generated in the step S2 has a length l_wThe binary watermark pattern of (a).

As a further preferable technical solution of the present invention, the watermark signal Ψ of the spread spectrum watermark algorithm in step S3 is synthesized into a spread spectrum signal without changing the spectral coefficient, and the watermark signal Ψ is embedded in the radius ρ of each cylinder vertex, and its formula is as follows:

wherein the signal Ψ is determined by the azimuth of the input vertexIs generated by a function of the watermark signal ΨIs defined and added to the radial distance of the cylindrical coordinateIn (1).

As a further preferable technical solution of the present invention, in the step S3, watermark embedding is performed on the preprocessed original model by using a watermark pattern, and the specific implementation steps are as follows:

step S31, converting the vector v in the Cartesian coordinates_i＝(x_i,y_i,z_i) Conversion to cylindrical coordinatesThe coordinate function is as follows:

wherein,

step S32, watermark embedding is carried out by adopting the watermark pattern, and the watermark embedding form is as follows:

where 1 ≦ i ≦ V, α (·) is a function of the visual masking described;

step S33, synthesizing the spread spectrum signal psi as a sine signal into a frequency domain [ f_s+1，f_s+l_w]The synthesis formula is as follows:

wherein f is_sIs the value of the minimum frequency band, and_i,lis an imperceptible phase parameter, yielding V '═ V'_iV as a set of watermark vertices, where V is 1,2

As a further preferable embodiment of the present invention, the step S3 further includes the steps of:

and adopting a visual masking method based on surface roughness to improve the invisibility of the embedded watermark.

As a further preferable technical solution of the present invention, the using of the visual masking method based on the surface roughness to improve invisibility of the embedded watermark specifically includes:

covering the rough area of the model to hide some geometric distortions, using the covering model M and the vertices v_iE.v, function of watermark strength α (V)_i) The calculation is as follows:

α(v_i)＝s·Υ_i·β_i(10)

wherein, s, γ_iAnd β_iStrength factor, roughness measurement and smoothing parameters, respectively;

using curvature analysis based on local windows on the 3D grid M, to measure the value denoted R (v)_i) Roughness, strength factor gamma_iDetermined by the average roughness value of each local area i, which is calculated by the formula:

wherein R is_iIs the roughness of the region i, R_meanAnd R_minThe average and minimum roughness of the entire grid is indicated.

As a further preferable technical solution of the present invention, in step S4, the model reconstructing the original model with the embedded watermark to obtain the watermark model specifically includes:

coordinates v 'of watermark vertex'_iE.g. V ', converted into Cartesian coordinates (x'_i,y′_i,z′_i) The conversion formula is as follows:

v'_iClassified into L different bins in the same manner as in equation (3) above, the position function of vertices x and y is as follows:

wherein,andfor the input position calculated in equation (5), a watermark model M '═ { V', E } is finally obtained through calculation, where E is the set of connectivity information of the original model.

The 3D grid blind watermark generation method based on the hierarchical artifact analysis can achieve the following beneficial effects:

according to the 3D grid blind watermark generation method based on the layered artifact analysis, by adopting the technical scheme, the watermark realized by the method has robustness and invisibility for severe environment, not only is the 3D model content protected when shared in the digital field, but also the 3D digital content can be protected when converted into analog content through 3D printing, in the scheme, the watermark model obtained by model reconstruction is used as a printing artifact, and the model which is used for printing and scanning by using the analysis result of the printing artifact as a template for providing orientation information for a watermark detector is synchronous with the original direction; experimental results confirm that the method does not lose the embedded pattern after 3D print scanning, and is particularly practical for low cost printers.

Drawings

The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.

Fig. 1 is a flowchart of an example method provided by the 3D mesh blind watermark generation method based on layered artifact analysis according to the present invention;

FIG. 2 is a sample of an example sampling model and its surface vertices;

fig. 3 is a graph of the effect of no visual masking (MSDM2 ═ 0.522);

fig. 4 is a graph of the effect of visual masking (MSDM2 ═ 0.355);

fig. 5 is a distortion diagram of fig. 4.

The objects, features and advantages of the present invention will be further explained with reference to the accompanying drawings.

Detailed Description

The invention will be further described with reference to the accompanying drawings and specific embodiments. In the preferred embodiments, the terms "upper", "lower", "left", "right", "middle" and "a" are used for clarity of description only, and are not used to limit the scope of the invention, and the relative relationship between the terms and the terms is not changed or modified substantially without changing the technical content of the invention.

In the technical field of 3D printing, one existing technical difficulty is the problem of copyright protection of 3D model contents. There is a need for a powerful and efficient technique to avoid unauthorized copying, tampering and illegal distribution to provide protection for intellectual property rights. Currently, 3D model providers try to protect 3D model content through fingerprinting, Digital Rights Management (DRM), access control, encryption and digital watermarking techniques, with the problem that encryption and DRM based 3D model content protection is ineffective because the 3D printing process disables these protections. Therefore, the 3D model contents can be illegally copied not only on the internet but also in the market and redistributed online. For this reason, research and development of a digital watermarking technology for protecting 3D printing are required to ensure prosperity of the 3D content industry.

In the current 3D printing field, 3D model content can not only be illegally copied over the internet, but also be illegally copied and redistributed in the market on line, thereby infringing copyright, a powerful and effective technique is needed to avoid unauthorized copying, tampering and illegal distribution, thereby providing protection for intellectual property rights.

The invention provides a 3D grid blind watermark generation method based on layered artifact analysis, which realizes that a watermark has robustness and invisibility for severe environment, not only ensures that 3D model content is protected when shared in the digital field, but also can be protected when 3D digital content is converted into analog content through 3D printing, and as shown in figure 1, the 3D grid blind watermark generation method based on layered artifact analysis comprises the following steps:

s1, covering the original model by taking the 3D polygonal grid M ═ { V, E } as input, and carrying out axis alignment pretreatment on the original model;

s2, generating a watermark pattern by adopting a watermark key and a pseudo-random generator;

s3, acquiring a spread spectrum watermarking algorithm based on the position information of the discrete vertex in the cylindrical coordinate, and embedding the watermark into the preprocessed original model by adopting a watermarking pattern;

and S4, carrying out model reconstruction on the original model of the embedded watermark to obtain a watermark model.

In a specific implementation, in step S1, the pre-processing that covers the original model by using the 3D polygon mesh M ═ { V, E } as an input, and performs axis alignment on the original model specifically includes:

s11, determining the z-axis of watermark embedding, and adjusting the direction of the grid M, wherein the adjustment calculation formula is as follows:

wherein v is_iE.g. V andis a rotation matrix and defines the following equation:

wherein, the ratio of theta,is the relative degree of the predefined printing direction, the z-axis is the orthogonal direction determined to be the 3D printed layer, in practice, in order to distribute the content of the 3D printed model, the provider should print the watermarked model along the z-axis, and in addition, some of the model vertices have a very biased distribution, as shown in fig. 2, which is a sample of the model and its surface vertices, in this example, the model is cut into 30 layers, the center position of each layer is aligned with the z-axis, and the model vertices are presented by small dots. In this case, the input original model is uniformly re-partitioned, so that not only can the irregularity of the vertex density be removed, but also the robustness of the re-partitioning attack generated by the 3D scanning can be provided;

s12, after the shaft aligning step, connecting v_i＝(x_i，y_i，z_i) Classifying as vertexes L different bins having a range equal to the z-axis value, calculating in advance the maximum and minimum values of the z-axis, and dividing the values by z_max，z_minDenotes the L-th layer L_lCalculated by the following formula:

L_l＝{v_i|z_min+δ_z·l<z_i<z_min+δ_z·(l+1)} (3)

wherein L is more than or equal to 1 and less than or equal to L, i is more than or equal to 1 and less than or equal to V, j is more than or equal to 1 and less than or equal to VI, and V_lIndicates to belong to L_lUsing the symbol v (l, j) representing the jth vertex of the ith bin instead of v_i∈L₁Thickness delta_zByDefining the number of bins in the table as a constant selected by a user;

s13, calculating the central position of each LThe calculation formula is as follows:

through the center position c_lFor its vertex v (L, j) ∈ L_lThe position of (2) is normalized as follows:

wherein,denotes the L-th layer L_lIs the p, q, r order moment of volume, each point v (L, j) is epsilon to L_lTo overlay the sliced layers of the model M, the overlay model is sliced into L layers, each layer centered on the z-axis.

In a specific implementation, the watermark pattern generated in the step S2 has a length l_wThe binary watermark pattern of (a).

In a specific implementation, the watermark signal Ψ of the spread spectrum watermark algorithm in step S3 is synthesized into a spread spectrum signal without changing the spectral coefficients, and the watermark signal Ψ is embedded into the radius ρ of each cylinder vertex, and its formula is as follows:

wherein the signal Ψ is determined by the azimuth of the input vertexIs generated by a function of the watermark signal ΨIs defined and added to the radial distance of the cylindrical coordinateIn (1).

The spread spectrum watermarking algorithm has the following advantages:

1) robustness to layered slices: the watermark signal Ψ is not affected by z-axis directional variations, and therefore, the signal Ψ is invariant to variations along the z-axis, making the embedded signal Ψ robust to the hierarchical slicing process;

2) blind watermark synchronization: the original euler angles (theta, psi,) Estimating the printing axis to restore the synchronization of the Euler angles theta and psi by analyzing the layering effect using an estimator, the embedded watermark corresponding to the Euler angles during the detectionHas invariance because the embedded signal is decomposed into discrete fourier transformed amplitudes;

3) robustness against connectivity and recombination attacks: the composite signal Ψ is in continuous coordinatesHas a sinusoidal structure on the vertex radius, and therefore watermark embedding is not affected by the spatial regularity of the vertices, and furthermore, the watermark signal Ψ is hardly affected by the uniformity of the vertex sampling by the uniform resampling and interpolation processing in the detection step.

In a specific implementation, the step S3 of embedding the watermark into the preprocessed original model by using the watermark pattern specifically includes the following steps:

step S31, converting the vector v in the Cartesian coordinates_i＝(x_i,y_i,z_i) Conversion to cylindrical coordinatesThe coordinate function is as follows:

wherein,in this step, the arctangent arctan must be properly defined, while the correct quadrant of (x, y) must be considered;

step S32, watermark embedding is carried out by adopting the watermark pattern, and the watermark embedding form is as follows:

where 1 ≦ i ≦ V, α (·) is a function of the visual masking described;

step S33, synthesizing the spread spectrum signal psi as a sine signal into a frequency domain [ f_s+1，f_s+l_w]The synthesis formula is as follows:

wherein f is_sIs the value of the minimum frequency band, and_i,lis an imperceptible phase parameter, yielding V '═ V'_iV as a set of watermark vertices, where V is 1,2

In a specific implementation, the step S3 further includes the following steps:

and adopting a visual masking method based on surface roughness to improve the invisibility of the embedded watermark.

In a specific implementation, the visual masking method based on surface roughness is adopted to improve the invisibility of the embedded watermark, and the concept of the visual masking is that a rough area of an overlay model can hide some geometric distortions if the frequencies of the rough area are very similar, and the implementation method specifically comprises the following steps:

covering the rough area of the model to hide some geometric distortions, using the covering model M and the vertices v_iE.v, function of watermark strengthNumber α (v)_i) The calculation is as follows:

α(v_i)＝s·Υ_i·β_i(10)

wherein, s, γ_iAnd β_iStrength factor, roughness measurement and smoothing parameters, respectively;

using curvature analysis based on local windows on the 3D grid M, to measure the value denoted R (v)_i) Roughness, strength factor gamma_iDetermined by the average roughness value of each local area i, which is calculated by the formula:

wherein R is_iIs the roughness of the region i, R_meanAnd R_minThe average and minimum roughness of the entire grid is indicated.

In an embodiment of the visual masking method based on surface roughness, as shown in fig. 3-5, wherein fig. 3 and 4 are both samples of watermark models with high intensity factor (s-10 e)^-4) The difference is that fig. 3 shows no visual masking (MSDM2 is 0.522), fig. 4 shows visual masking (MSDM2 is 0.355), fig. 3 is rougher than fig. 4, and fig. 5 shows the distortion diagram of fig. 4. Both models use high intensity factor (s 10 e)^-4) The watermark is performed to show the visual shape of the watermark, the geometric distortion is detected using a mesh structure distortion measure (MSDM2) with a standard configuration (number 3, symmetric), the embedded watermark appears as a strongly represented stripe pattern on certain areas of the model surface, which is hidden from the embedded watermark in fig. 4 compared to fig. 3.

In a specific implementation, in step S4, the performing model reconstruction on the original model in which the watermark is embedded to obtain the watermark model specifically includes:

coordinates v 'of watermark vertex'_iE.g. V', turnConversion to Cartesian coordinates (x'_i,y′_i,z′_i) The conversion formula is as follows:

v'_iClassified into L different bins in the same manner as in equation (3) above, the position function of vertices x and y is as follows:

wherein,andfor the input position calculated in equation (5), a watermark model M '═ { V', E } is finally obtained through calculation, where E is the set of connectivity information of the original model.

By adopting the technical scheme, the watermark realized by the method has robustness and invisibility for severe environment, the 3D model content is protected when shared in the digital field, and the 3D digital content can also be protected when converted into analog content through 3D printing; experimental results confirm that the method does not lose the embedded pattern after 3D print scanning, and is particularly practical for low cost printers. The method of the present invention can estimate parameters of low cost 3D printing methods such as Laminate Object Manufacturing (LOM) and Fused Deposition Modeling (FDM), and the proposed method can cover today's most dominant 3D printing solutions such as RepRap, Prusa, MakerBot, Ultimaker.

Although specific embodiments of the present invention have been described above, it will be appreciated by those skilled in the art that these are merely examples and that many variations or modifications may be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims.

Claims (8)

1. A3D grid blind watermark generation method based on layered artifact analysis is characterized by comprising the following steps:

s1, covering the original model by taking the 3D polygonal grid M ═ { V, E } as input, and carrying out axis alignment pretreatment on the original model;

s2, generating a watermark pattern by adopting a watermark key and a pseudo-random generator;

s3, acquiring a spread spectrum watermarking algorithm based on the position information of the discrete vertex in the cylindrical coordinate, and embedding the watermark into the preprocessed original model by adopting a watermarking pattern;

and S4, carrying out model reconstruction on the original model of the embedded watermark to obtain a watermark model.

2. The method for generating a 3D mesh blind watermark based on layered artifact analysis as claimed in claim 1, wherein said step S1, overlaying the original model with the 3D polygonal mesh M ═ { V, E } as input, and performing pre-processing of axis alignment on the original model specifically comprises:

s11, determining the z-axis of watermark embedding, and adjusting the direction of the grid M, wherein the adjustment calculation formula is as follows:

wherein v is_iE.g. V andis a rotation matrix and defines the following equation:

wherein, the ratio of theta,is the relative degree of the predefined printing direction, the z-axis being the orthogonal direction determined to be the 3D printed layer;

s12, mixing v_i＝(x_i，y_i，z_i) Classifying as vertexes L different bins having a range equal to the z-axis value, calculating in advance the maximum and minimum values of the z-axis, and dividing the values by z_max，z_minDenotes the L-th layer L_lCalculated by the following formula:

L_l＝{v_i|z_min+δ_z·l<z_i<z_min+δ_z·(l+1)} (3)

wherein L is more than or equal to 1 and less than or equal to L, i is more than or equal to 1 and less than or equal to V, j is more than or equal to 1 and less than or equal to VI, and V_lIndicates to belong to L_lUsing the symbol v (l, j) representing the jth vertex of the ith bin instead of v_i∈L₁Thickness delta_zByDefining the number of bins in the table as a constant selected by a user;

s13, calculating the central position of each LThe calculation formula is as follows:

through the center position c_lFor its vertex v (L, j) ∈ L_lThe position of (2) is normalized as follows:

wherein,denotes the L-th layer L_lIs the p, q, r order moment of volume, each point v (L, j) is epsilon to L_lTo overlay the sliced layers of the model M, the overlay model is sliced into L layers, each layer centered on the z-axis.

3. The layered artifact analysis-based 3D mesh blind watermark generation method according to claim 2, wherein the watermark pattern generated in the step S2 is a length l_wThe binary watermark pattern of (a).

4. The layered artifact analysis-based 3D mesh blind watermark generation method according to claim 3, wherein the watermark signal Ψ of the spread spectrum watermark algorithm in the step S3 is synthesized as a spread spectrum signal without changing the spectral coefficients, and the watermark signal Ψ is embedded in the radius ρ of each cylinder vertex, and its formula is as follows:

wherein the signal Ψ is determined by the azimuth of the input vertexIs generated by a function of the watermark signal ΨIs defined and added to the radial distance of the cylindrical coordinateIn (1).

5. The layered artifact analysis-based 3D mesh blind watermark generation method according to claim 4, wherein in the step S3, the watermark embedding is performed on the preprocessed original model by using the watermark pattern, and the specific implementation steps are as follows:

step S31, converting the vector v in the Cartesian coordinates_i＝(x_i,y_i,z_i) Conversion to cylindrical coordinatesThe coordinate function is as follows:

wherein,

step S32, watermark embedding is carried out by adopting the watermark pattern, and the watermark embedding form is as follows:

where 1 ≦ i ≦ V, α (·) is a function of the visual masking described;

step S33, synthesizing the spread spectrum signal psi as a sine signal into a frequency domain [ f_s+1，f_s+l_w]The synthesis formula is as follows:

wherein f is_sIs the value of the minimum frequency band, and_i,lis an imperceptible phase parameter, yielding V '═ V'_iV as a set of watermark vertices, where V is 1,2

6. The layered artifact analysis-based 3D mesh blind watermark generation method according to claim 5, wherein the step S3 further comprises the following steps:

and adopting a visual masking method based on surface roughness to improve the invisibility of the embedded watermark.

7. The method for generating a 3D mesh blind watermark based on layered artifact analysis as claimed in claim 6, wherein said employing a visual masking method based on surface roughness to improve invisibility of the embedded watermark specifically comprises:

covering the rough area of the model to hide some geometric distortions, using the covering model M and the vertices v_iE.v, function of watermark strength α (V)_i) The calculation is as follows:

α(v_i)＝s·Υ_i·β_i(10)

wherein, s, γ_iAnd β_iStrength factor, roughness measurement and smoothing parameters, respectively;

using curvature analysis based on local windows on the 3D grid M, to measure the value denoted R (v)_i) Roughness, strength factor gamma_iDetermined by the average roughness value of each local area i, which is calculated by the formula:

wherein R is_iIs the roughness of the region i, R_meanAnd R_minThe average and minimum roughness of the entire grid is indicated.

8. The layered artifact analysis-based 3D mesh blind watermark generation method according to claim 7, wherein the step S4 of performing model reconstruction on the original watermark embedded model to obtain the watermark model specifically comprises:

coordinates v 'of watermark vertex'_iE.g. V ', converted into Cartesian coordinates (x'_i,y′_i,z′_i) The conversion formula is as follows:

v'_iClassified into L different bins in the same manner as in equation (3) above, the position function of vertices x and y is as follows:

wherein,andfor the input position calculated in equation (5), a watermark model M '═ { V', E } is finally obtained through calculation, where E is connectivity information of the original modelAnd (4) collecting.