Robust high-quality three-dimensional model digital watermark embedding and extracting method

By performing topological structure analysis on the three-dimensional model and embedding watermark information on local coordinate systems, the volatile problem of watermarks under attack in the existing technology is solved, and robust watermark embedding and extraction with high capacity and low computational complexity is achieved. , improves anti-interference ability.

CN119090697AActive Publication Date: 2024-12-06ZHEJIANG UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411220182.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-02
Publication Date
2024-12-06
Estimated Expiration
2044-09-02

AI Technical Summary

Technical Problem

The existing three-dimensional model watermarking technology cannot effectively resist when facing various attacks, especially affine transformation, topology, cropping, noise and connectivity attacks, resulting in the loss or difficulty in extracting watermark information, and traditional methods maintain the quality of the model while maintaining the quality of the model. Large capacity watermark embedding and low computational complexity cannot be achieved.

Method used

By preprocessing the three-dimensional model, vertex coordinates are standardized, topological structure and vertex arrangement order are determined, vertices are classified into reference points and embedable watermark points, watermark information is embedded based on local coordinate systems, and through specific quantization step size and inverse transformation Technology ensures the robustness of the watermark.

Benefits of technology

It realizes large-capacity and low-distortion watermark embedding and extraction, which can resist attacks such as watermark quantization granularity, noise and vertex rearrangement, and improves the robustness and anti-interference ability of watermark technology.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119090697A_ABST
    Figure CN119090697A_ABST
Patent Text Reader

Abstract

The invention discloses a robust high-quality three-dimensional model digital watermark embedding and extracting method, and the watermark embedding method comprises the steps: carrying out the filtering of a vertex order initial point through the vertex in-degree of a three-dimensional grid model; a robust starting point and vertex sequence which is not influenced by vertex three-dimensional coordinate values is calculated through the sum of geodesic distances of the vertexes; dividing the vertexes into embeddable watermark points and reference points by using a proposed specific algorithm; the invention discloses a method for establishing a local space capable of embedding watermark points by using principal component analysis and embedding watermarks in the local space by using a quantization method. The invention further provides a watermark extraction method corresponding to the watermark embedding method. Compared with a traditional airspace three-dimensional grid model watermarking method, the method has the advantages that the distortion of the embedded watermark on the surface of the model is smaller, the noise attack resistance is higher, and the vertex sequence corresponding to the model is more robust.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of digital watermarking, and in particular to a robust high-quality three-dimensional model digital watermark embedding and extraction method. Background Art

[0002] The background of 3D model watermark embedding technology involves digital copyright protection and secure management of multimedia content. 3D models are a new type of media data that have been widely used in many fields such as computer-aided design, medical visualization, film and television games, and intelligent education. With the development of the Internet, streaming media, and games, it has become very easy to copy, share, and distribute multimedia content containing 3D models, making copyright protection an important issue. In order to prevent illegal copying and distribution, 3D model digital watermarking technology came into being. This technology aims to identify and protect ownership or copyright by embedding information into 3D models.

[0003] Three-dimensional model watermarking technology can be divided into fragile and robust types. Fragile watermarks are used for content verification and can detect subtle modifications, while robust watermarks ensure that the embedded information can still be extracted in the face of various attacks while maintaining the quality of the model. Watermarks can be embedded in different domains of the three-dimensional model, such as the spatial domain or the spectral domain, and are classified according to different application scenarios, including private watermarks and public watermarks, blind detection and non-blind detection watermarks, etc. Developing blind detection watermarking technology that is robust to various attacks is the focus of current research.

[0004] At present, the common attack technologies of 3D model watermarks are mainly the following:

[0005] 1. Affine transformation attack: mainly includes attacks such as rotation, translation, and scaling. Although these operations will not change the shape of the 3D model, they will change the position of the vertices, resulting in the loss of embedded watermark information.

[0006] 2. Topological attacks: mainly include mesh simplification and mesh encryption. These operations will not change the general shape of the 3D model, but will add or reduce the number of watermarks, or change the order of watermarks.

[0007] 3. Cropping attack: Cropping attack removes some vertices and faces. This type of attack will cause a large amount of watermark information attached to the model to disappear, making it difficult to restore the watermark.

[0008] 4. Noise attack: Noise attack changes the shape of the model by adding noise to the original vertices and faces.

[0009] 5. Connectivity attacks: such as vertex reordering.

[0010] Under the above attacks, the traditional 3D model watermarking technology has certain limitations in protecting the 3D model. It cannot effectively resist the attacks of third parties while maintaining invisibility and low distortion, and can achieve blind extraction of watermarks (that is, without the original model). At the same time, the capacity of watermark embedding needs to be as large as possible and the computational complexity as low as possible. Summary of the invention

[0011] The purpose of the present invention is to provide a robust high-quality three-dimensional model digital watermark embedding and extraction method in view of the deficiencies of the prior art.

[0012] The present invention is achieved through the following technical solutions:

[0013] A robust high-quality three-dimensional model digital watermark embedding method comprises the following steps:

[0014] S1: preprocessing the three-dimensional model to normalize the coordinates of each vertex in the three-dimensional model;

[0015] S2: determining the arrangement order of each vertex of the three-dimensional model according to the topological structure of the three-dimensional model;

[0016] S3: classify all vertices into reference points and embeddable watermark points, so that the embeddable watermark points are not adjacent to each other;

[0017] S4: Obtain the watermark sequence to be embedded, which consists of multiple watermark information arranged in order; preset the watermark parameter x to be embedded elem ,y elem 、z elem and loop num , determine the quantization step size to be embedded in the watermark according to the watermark parameters;

[0018] S5: traversing the embeddable watermark points in sequence according to the arrangement order of each vertex, for the traversed embeddable watermark points, selecting a part of the reference points as the reference points corresponding to the current embeddable watermark point, establishing a local coordinate system based on the reference points corresponding to the current embeddable watermark point, and transforming the current embeddable watermark point into the local coordinate system to obtain the local coordinates of the current embeddable watermark point;

[0019] S6: According to the arrangement order of the vertices, each embeddable watermark point corresponds to the watermark information in the watermark sequence in turn, and the corresponding watermark information is embedded into each embeddable watermark point according to the quantization step size of the watermark;

[0020] S7: transforming the local coordinates of the embeddable watermark point after the watermark is embedded back to the original coordinate system.

[0021] Furthermore, in step S1, the preprocessing is specifically as follows: calculating the center of gravity of the three-dimensional model, subtracting the center of gravity coordinates from the coordinates of all vertices to achieve the translation of the three-dimensional model; performing principal component analysis on the three-dimensional model after the current translation to obtain three linearly independent three-dimensional vectors, establishing a global coordinate system based on these three three-dimensional vectors, and converting each vertex of the three-dimensional model to the coordinate axis; uniformly scaling the coordinates of all vertices of the three-dimensional model so that the x, y, and z components of all vertex coordinates are between [-1, 1].

[0022] Furthermore, in step S2, determining the arrangement order of each vertex of the three-dimensional model according to the topological structure of the three-dimensional model specifically includes the following steps:

[0023] S2.1: Count the in-degrees of all vertices of the 3D model and combine vertices with the same in-degree into a vertex set;

[0024] S2.2: Select the vertex set with the least number of vertices. If there are multiple vertex sets with the least number of vertices, select the vertex set with the largest in-degree.

[0025] S2.3: For the selected vertex set, calculate the sum of the geodesic distances of each vertex in the set; select the vertex with the largest sum of geodesic distances in the current set as the starting point, and add the starting point to the vertex queue;

[0026] S2.4: Calculate the sum of the geodesic distances of each adjacent point of the starting point, and select the adjacent point with the largest sum of geodesic distances to add to the vertex queue;

[0027] S2.5: Based on the two vertices added to the vertex queue by L in S2.3 and S2.4, all vertices are traversed using a counterclockwise breadth-first algorithm and the traversed vertices are added to the vertex queue in turn. The final arrangement order of the vertices in the vertex queue is the determined arrangement order.

[0028] Furthermore, in step S3, the classification of all vertices into reference points and watermark embeddable points specifically includes the following steps:

[0029] S3.1: Set all vertices to embeddable watermark points;

[0030] S3.2: extracting the vertices of the three-dimensional model in order according to the arrangement order of the vertices;

[0031] S3.3: If the currently extracted vertex is a watermark embeddable point, all vertices directly adjacent to the currently extracted vertex are set as reference points, and the next vertex is extracted in the order of the vertices;

[0032] S3.4: If the currently extracted vertex is a reference point, continue to extract the next vertex according to the arrangement order of the vertices;

[0033] S3.5: Repeat S3.2-S3.4 until the last vertex.

[0034] Furthermore, the quantization step size to be embedded in the watermark is determined according to the watermark parameters, specifically:

[0035] δx=Length average *x elem

[0036] δy=Length average *y elem

[0037] δz=Length average *z elem

[0038] Among them, δx, δy and δz are the three components of the watermark quantization step length, Length average is the average side length of the 3D model.

[0039] Further, the step length between the reference points and the current embeddable watermark point is not greater than loop. num All reference points of the reference point set form a reference point set, and all reference points in the reference point set are reference points corresponding to the current embeddable watermark point; principal component analysis is performed on the reference points in the reference point set to obtain a local coordinate system.

[0040] Furthermore, in step S6, the corresponding watermark information is embedded into each embeddable watermark point according to the quantization step of the watermark, specifically:

[0041]

[0042] Among them, Q(point locx )、Q(point locy ) and Q(point locz ) is the x-component, y-component and z-component of the local coordinates of the current embeddable watermark point after embedding the watermark, point locx 、point locy and point locz are the x-component, y-component and z-component of the local coordinates of the current embeddable watermark point, and q is the watermark information corresponding to the current embeddable watermark point.

[0043] The present invention also proposes a robust high-quality three-dimensional model digital watermark extraction method, wherein the three-dimensional model is embedded with a digital watermark using the above-mentioned watermark embedding method, and the watermark extraction method comprises:

[0044] 1) Preprocessing the 3D model to normalize the coordinates of each vertex in the 3D model; determining the arrangement order of each vertex of the 3D model, and classifying all vertices into reference points and embeddable watermark points;

[0045] 2) Traversing the embeddable watermark points in sequence according to the arrangement order of each vertex, for the traversed embeddable watermark points, establishing a local coordinate system based on the reference point corresponding to the current embeddable watermark point, and transforming the current embeddable watermark point into the local coordinate system to obtain the local coordinates of the current embeddable watermark point;

[0046] 3) extracting the x component, y component and z component of the watermark information corresponding to each embeddable watermark point;

[0047] 4) Sum the three components of the watermark information. If the sum is greater than 1, the watermark information corresponding to the current embeddable watermark point is 1, otherwise it is 0; sequentially combine all watermark information to obtain the watermark sequence embedded in the three-dimensional model.

[0048] The beneficial effects of the present invention are:

[0049] It can achieve large-capacity, low-distortion watermark embedding, and the vertex order required for extracting the watermark is not affected by attacks such as watermark quantization granularity, noise, consistency changes, and vertex rearrangement. At the same time, the anti-noise ability can be improved by adjusting the configuration parameters according to one's own needs. The vertex order of the traditional spatial domain watermark method will be destroyed when the noise is large and the embedded watermark granularity is coarse, but the vertex order method based on pure topology in the present invention will not, which makes the watermark technology of the present invention have high robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a watermark embedding flow chart of the present invention;

[0051] Figure 2 It is a watermark extraction flow chart of the present invention;

[0052] Figure 3 A three-dimensional mesh model used in a test in an embodiment of the present invention;

[0053] Figure 4 This is a rendering of the three-dimensional mesh model used in the test in one embodiment of the present invention after embedding the watermark. DETAILED DESCRIPTION

[0054] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be pointed out that the embodiments described below are intended to facilitate the understanding of the present invention and do not have any limiting effect on the present invention.

[0055] In one embodiment of the present invention, a robust high-quality 3D model digital watermark embedding method is provided, which aims to improve the anti-attack and concealment of the 3D model during the information embedding process, while reducing the impact on the original 3D model structure, and has a wide range of practical value in applications such as copyright protection, data verification and model authentication. Figure 1 As shown, the watermark embedding method comprises the following steps:

[0056] (1) A method for finding a starting point based on the sum of vertex in-degree and geodesic distance and a vertex rearrangement method for a three-dimensional model, the method comprising the following steps:

[0057] (1-1) Read the 3D model to be embedded with watermark, which is generally in obj format by default, and import it into the program and convert it into a half-edge model. The present invention uses a triangular mesh model, in which each face of the model is composed of three vertices to form a triangle.

[0058] (1-2) Preprocessing the three-dimensional model, the specific steps of which are to calculate the center of gravity of the three-dimensional model and translate the model so that the coordinates of the center of gravity of the model coincide with the origin of the coordinate system, perform principal component analysis on the vertex coordinates of the three-dimensional mesh model after the current translation, calculate three new linearly unrelated three-dimensional vectors, and re-establish the coordinate axis based on these three three-dimensional vectors, transform the vertex coordinates of the three-dimensional model into the coordinate axis, and uniformly scale each component of the vertex coordinate in the coordinate axis to a space between [-1,1]. Specifically, the x, y and z components of the vertex coordinate are all between [-1,1].

[0059] (1-3) Calculate the in-degree of all vertices in the model and count them, and group vertices with the same in-degree together into a set.

[0060] (1-4) Count the number of vertices contained in all sets and select the set with the least number of vertices. If there are multiple sets with the least number of vertices, select the vertex set with the largest in-degree.

[0061] (1-5) Calculate the sum of the geodesic distances of each vertex in the selected set. The method for calculating the sum of the geodesic distances of this method is as follows: the distance between two directly connected vertices is one. Based on this condition, the Dijkstra algorithm is used for the vertices to calculate the shortest distance from the current vertex to any vertex. After the calculation is completed, the shortest distance from the current vertex to all vertices in the three-dimensional model is summed up, thus completing the calculation of the sum of the geodesic distances corresponding to the vertices.

[0062] (1-6) Select the vertex with the largest geodesic distance in the current set as the starting point, add the starting point to the vertex queue, and the vertex order of the starting point is the first. Use the starting point as the initial center point and continue the following steps.

[0063] (1-7) In a three-dimensional mesh model, two directly connected vertices are each other's neighboring points. Find all the neighboring points of the center point and select one of the neighboring points to add to the vertex queue. The calculation method is to calculate the sum of the geodesic distances corresponding to all neighboring points, and select the vertex with the largest sum of geodesic distances among the neighboring points to add to the vertex queue.

[0064] (1-8) Based on the vertices that have been determined to be the first and second vertices, a breadth-first algorithm is used to traverse all vertices and add the traversed vertices to the vertex queue in turn, thereby generating a vertex order. The vertex order generation method for the adjacent points of a single vertex is as follows: All adjacent points of the vertex can be traversed and stored counterclockwise through the half-edge structure. All adjacent points are essentially a ring arranged in a counterclockwise order. The last vertex in the ring that has been determined in order is determined in a counterclockwise order. The vertex that is located after the determined vertex in the ring and whose order has not yet been determined is traversed one by one, and the traversed vertices are added to the vertex queue in turn.

[0065] (1-9) Use the breadth-first algorithm to traverse all vertices, and the order of the vertices in the vertex queue is the order in which the vertices are arranged.

[0066] (2) The low-distortion 3D watermark embedding method based on local plane is as follows:

[0067] (2-1) According to the vertex order of the current three-dimensional model, the vertices are classified into reference points and embeddable watermark points. The method is as follows: 1. All vertices are set as embeddable watermark points. 2. Vertices are taken out in order. 3. If the vertex currently taken out is an embeddable watermark point, all vertices directly adjacent to the currently taken out vertex are set as reference points, and step 2 will be continued. 4. If the currently taken out vertex is a reference point, step 2 will be continued. 5. Execute to the last vertex and end.

[0068] (2 - 2) Calculate the average side length of the three-dimensional model average , and determine the four parameters required for embedding the watermark, which are x elem (X-axis quantization granularity), y elem (Y-axis quantization granularity), z elem (Z-axis quantization granularity), loop num (Refer to step size). The quantization step size of the watermark is determined by the above parameters, and the specific formula is as follows:

[0069] δx=Length average *x elem

[0070] δy=Length average *yelem

[0071] δz=Length average *z elem

[0072] Among them, δx, δy and δz are the three components of the watermark quantization step.

[0073] (2-3) Take out the vertices in the order determined in step 1. If they are watermark points that can be embedded, first follow the loop num Determine the surrounding reference points by traversing and obtaining the vertices outside the embeddable watermark point, and select the vertices whose step length with the embeddable watermark point is less than or equal to loop num All vertices of are a set, and all embeddable watermark points in the set are removed. The remaining vertices in the set are the reference points corresponding to the vertices, forming a reference point set. The step length is the length of the shortest path between two vertices, and the length of the shortest path between two directly connected vertices is 1.

[0074] (2-4) Using principal component analysis on the reference points in the reference point set, a local center of gravity coordinate center can be obtained loc And three local coordinate axes locx , Axis locy , Axis locz , subtract the center value from the coordinates of all reference points in the reference point set and the current embeddable watermark point loc , and convert it to the local coordinate axis to obtain the local coordinate point of the current embedded watermark point loc .

[0075] (2-5) Obtain the watermark sequence to be embedded. The watermark sequence consists of multiple watermark information (0 or 1). Each embeddable vertex corresponds to the watermark information in the watermark sequence in the order of vertex arrangement. Since the number of model vertices is generally much larger than the number of watermark sequence bits, the watermark sequence is first repeatedly arranged into a long sequence, and the watermark information in the long sequence corresponds to each vertex arranged in order. Take out the watermark information q corresponding to the embeddable watermark point and use the quantization method to repeatedly embed it into point loc The quantization formula of the x component is as follows, and the quantization method of the y component and the z component is the same as the quantization method of the x component.

[0076]

[0077] Where Q(point locx ) is the x component of the local coordinate after embedding the watermark, point locx The reference point locThe x-component of the coordinates of Represents the floor operator.

[0078] (2-6) Perform the inverse transformation corresponding to (2-4) on the local coordinates after embedding the watermark and convert them into global coordinates.

[0079] like Figure 2 As shown, in another embodiment of the present invention, a matching three-dimensional model digital watermark extraction method is also provided. After a model embedded with a watermark is obtained according to the above watermark embedding method, the watermark is extracted according to the following steps.

[0080] (3-1) Similar to step (1) in the previous embodiment, the model with the watermark is preprocessed and the vertex arrangement order of the three-dimensional model is obtained.

[0081] (3-2) Similar to step (2-1) in the previous embodiment, the vertices of the watermarked model are classified into reference points and watermark embeddable points.

[0082] (3-3) is the same as steps (2-2) to (2-5) in the previous embodiment. Each vertex is taken out in the order of vertex arrangement, and the reference point set of the embeddable watermark point is obtained. The local coordinate system is calculated through the reference point and the coordinates of the embeddable watermark point are transformed into the local coordinate system.

[0083] (3-4) Extract the watermark, where the operation on the x component is as follows, and the operations on the y component and the z component are the same as the operation on the x component.

[0084]

[0085] Among them, point x is the x-component of the coordinates of the embedded watermark point in the local coordinate system, point' x is the x component of the coordinates of the embedded watermark point after the watermark is extracted, q x is the x component of the watermark information corresponding to the current embeddable watermark point. Repeat the above operation on the y and z components to obtain the y and z components of the watermark information.

[0086] (3-5) The sum of the three components of the watermark is greater than 1, and the watermark information embedded in the vertex is determined to be 1, otherwise it is determined to be 0. After traversing all the points that can be embedded with watermarks in the order of vertex arrangement, the watermark information corresponding to each point that can be embedded with watermarks is arranged in order, and the watermark sequence is obtained.

[0087] (3-6) Figure 3Four different common 3D mesh models are shown in Figure 1, namely human portrait (upper left), rabbit (upper right), horse (lower left), and armadillo (lower right). Table 1 shows the accuracy of the model watermark under vertex rearrangement attack, affine transformation attack, and different degrees of noise attack under the corresponding configuration parameters of these four 3D mesh models.

[0088] Table 1

[0089]

[0090]

[0091] From the experimental results, we can see that the watermarking method is completely robust to affine transformation and vertex rearrangement attacks, and also works well against noise attacks, especially on models with a small number of vertices. Figure 4 The rendering of the model after watermarking is given under the configuration parameters in Table 1. It can be seen that this watermark embedding method has little effect on the quality of the 3D mesh model itself.

[0092] The embodiments described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A robust high-quality 3D model digital watermark embedding method, characterized in that: The following steps are involved: S1: preprocessing the three-dimensional model to normalize the coordinates of each vertex in the three-dimensional model; S2: determining the arrangement order of each vertex of the three-dimensional model according to the topological structure of the three-dimensional model; S3: classify all vertices into reference points and embeddable watermark points, so that the embeddable watermark points are not adjacent to each other; S4: obtaining a watermark sequence to be embedded with a watermark, the watermark sequence consisting of a plurality of watermark information arranged in an orderly manner; presetting watermark parameters to be embedded with a watermark, the watermark parameters including an X-axis quantization granularity, a Y-axis quantization granularity, a Z-axis quantization granularity and a reference step size, and determining a quantization step size to be embedded with a watermark according to the watermark parameters; S5: traversing the embeddable watermark points in sequence according to the arrangement order of each vertex, for the traversed embeddable watermark points, selecting a part of the reference points as the reference points corresponding to the current embeddable watermark point, establishing a local coordinate system based on the reference points corresponding to the current embeddable watermark point, and transforming the current embeddable watermark point into the local coordinate system to obtain the local coordinates of the current embeddable watermark point; S6: According to the arrangement order of the vertices, each embeddable watermark point corresponds to the watermark information in the watermark sequence in turn, and the corresponding watermark information is embedded into each embeddable watermark point according to the quantization step size of the watermark; S7: transforming the local coordinates of the embeddable watermark point after the watermark is embedded back to the original coordinate system.

2. The robust high-quality 3D model digital watermark embedding method according to claim 1, characterized in that: In step S1, the preprocessing is specifically: Calculate the center of gravity of the three-dimensional model, subtract the center of gravity coordinates from the coordinates of all vertices to achieve the translation of the three-dimensional model; perform principal component analysis on the three-dimensional model after the current translation to obtain three linearly independent three-dimensional vectors, establish a global coordinate system based on these three-dimensional vectors, and transform each vertex of the three-dimensional model to the coordinate axis; uniformly scale the coordinates of all vertices of the three-dimensional model so that the x, y, and z components of all vertex coordinates are between [-1,1].

3. The robust high-quality 3D model digital watermark embedding method according to claim 1, characterized in that: In step S2, the arrangement order of the vertices of the three-dimensional model is determined according to the topological structure of the three-dimensional model, which specifically includes the following steps: S2.1: Count the in-degrees of all vertices of the 3D model and combine vertices with the same in-degree into a vertex set; S2.2: Select the vertex set with the least number of vertices. If there are multiple vertex sets with the least number of vertices, select the vertex set with the largest in-degree. S2.3: For the selected vertex set, calculate the sum of the geodesic distances of each vertex in the set; select the vertex with the largest sum of geodesic distances in the current set as the starting point, and add the starting point to the vertex queue; S2.4: Calculate the sum of the geodesic distances of each adjacent point of the starting point, and select the adjacent point with the largest sum of geodesic distances to add to the vertex queue; S2.5: Based on the two vertices that have been added to the vertex queue in S2.3 and S2.4, use the counterclockwise breadth-first algorithm to traverse all vertices and add the traversed vertices to the vertex queue in turn. The final arrangement order of the vertices in the vertex queue is the determined arrangement order.

4. The robust high-quality 3D model digital watermark embedding method according to claim 3, characterized in that: The specific calculation method of the sum of the geodesic distances of the vertices is: define the distance between two directly connected vertices to be one, calculate the shortest distance from the vertex to other vertices by Dijkstra algorithm, and sum the shortest distance from the current vertex to each other vertex to obtain the sum of the geodesic distances of the vertex.

5. The robust high-quality 3D model digital watermark embedding method according to claim 1, characterized in that: In step S3, the classification of all vertices into reference points and watermark embeddable points specifically includes the following steps: S3.1: Set all vertices to embeddable watermark points; S3.2: extracting the vertices of the three-dimensional model in order according to the arrangement order of the vertices; S3.3: If the currently extracted vertex is a watermark embeddable point, all vertices directly adjacent to the currently extracted vertex are set as reference points, and the next vertex is extracted in the order of the vertices; S3.4: If the currently extracted vertex is a reference point, continue to extract the next vertex according to the arrangement order of the vertices; S3.5: Repeat S3.2-S3.4 until the last vertex.

6. The robust high-quality 3D model digital watermark embedding method according to claim 1, characterized in that: In step S4, the quantization step size to be embedded in the watermark is determined according to the watermark parameters, specifically: δx=Length average *x elem δy=Length average *y elem δz=Length average *z elem Among them, δx, δy and δz are the three components of the watermark quantization step length, Length average is the average side length of the three-dimensional model, x elem The x-axis is the quantization granularity, and the y elem The Y-axis quantifies the granularity, z elem Quantify the granularity for the Z-axis.

7. The robust high-quality 3D model digital watermark embedding method according to claim 1, characterized in that: In step S5, a part of the reference points are selected as the reference points corresponding to the currently embeddable watermark points, and a local coordinate system is established based on the reference points corresponding to the currently embeddable watermark points, specifically: All reference points whose step length with the current embeddable watermark point is not greater than the reference step length are selected to form a reference point set. All reference points in the reference point set are reference points corresponding to the current embeddable watermark point. Principal component analysis is performed on the reference points in the reference point set to obtain a local coordinate system.

8. The robust high-quality 3D model digital watermark embedding method according to claim 6, characterized in that: In step S6, the corresponding watermark information is embedded into each embeddable watermark point according to the quantization step size of the watermark, specifically: Among them, Q(point locx )、Q(point locy ) and Q(point locz ) is the x-component, y-component and z-component of the local coordinates of the current embeddable watermark point after embedding the watermark, point locx 、point locy and point locz are the x-component, y-component and z-component of the local coordinates of the current embeddable watermark point, and q is the watermark information corresponding to the current embeddable watermark point.

9. The robust high-quality 3D model digital watermark embedding method according to claim 1, characterized in that: The transformation in step S7 is the inverse transformation of the transformation in step S5.

10. A robust high-quality three-dimensional model digital watermark extraction method, characterized in that: The three-dimensional model is embedded with a digital watermark using the watermark embedding method described in claim 1, and the watermark extraction method comprises: 1) Preprocessing the 3D model to normalize the coordinates of each vertex in the 3D model; determining the arrangement order of each vertex of the 3D model, and classifying all vertices into reference points and embeddable watermark points; 2) Traversing the embeddable watermark points in sequence according to the arrangement order of each vertex, for the traversed embeddable watermark points, establishing a local coordinate system based on the reference point corresponding to the current embeddable watermark point, and transforming the current embeddable watermark point into the local coordinate system to obtain the local coordinates of the current embeddable watermark point; 3) Extract the x component, y component and z component of the watermark information corresponding to each embeddable watermark point, wherein the x component is extracted in the following formula, and the y component and z component are extracted in the same way as the x component: Among them, point x is the x component of the local coordinates of the embedded watermark point, point' x is the x component of the local coordinates of the watermark point that can be embedded after extracting the watermark, q x is the x component of the watermark information corresponding to the current embeddable watermark point; 4) Sum the three components of the watermark information. If the sum is greater than 1, the watermark information corresponding to the current embeddable watermark point is 1, otherwise it is 0; sequentially combine all watermark information to obtain the watermark sequence embedded in the three-dimensional model.

Citation Information

Patent Citations

  • Vertex weight value-based OBJ (object)-format three-dimensional model digital-watermarking method

    CN102314667A

  • Method for watermarking a three dimensional object and method for obtaining a payload from a three dimensional object

    US20160063661A1