Oblique photography three-dimensional model watermarking method based on multi-level curvature and weighted distance

Abstract

The invention discloses a multi-level curvature and weighted distance-based oblique photography three-dimensional model watermarking method, which comprises the following steps of: firstly, calculating curvature values of vertexes in different level neighborhoods by utilizing principal component analysis, and taking the vertexes with the curvature values higher than a threshold value as selected feature points; secondly, the vertexes are grouped according to the principle that the weighted distance between the feature points and the vertexes is minimum; and finally, normalizing the height difference between the vertexes and the feature points in each group, taking the normalized height difference as a carrier for embedding the watermark information, and establishing a corresponding relation between the watermark information and an embedding position by utilizing a QIM mapping thought so as to complete embedding of the watermark information. And the robustness of the algorithm is enhanced by using a grouping embedding and normalization method. Experimental results show that the method has high robustness on common attacks such as geometry, cutting, reordering, random point addition and deletion and combination, has good invisibility, and provides a new solution for copyright protection of the oblique photography three-dimensional model.

Description

Technical Field

[0001] The invention belongs to the field of geographic space data security and relates to an oblique photography three-dimensional model watermarking method based on multi-level curvature and weighted distance. Background Art

[0002] With the rapid development of geographic information technology, cities are upgrading and transforming from "digitalization" to "intelligence", pushing society into a smarter and more connected era. In this context, traditional two-dimensional map data can no longer meet the increasingly complex economic and social needs, especially in intelligent application scenarios, the limitations of two-dimensional maps are becoming more and more obvious. In contrast, oblique photography 3D models, with their wide coverage, high precision, high resolution and clear texture structure, can more intuitively and vividly display objects, and have gradually become an important tool for all walks of life, especially in the fields of autonomous driving, smart transportation and urban planning. Due to the high precision and clear texture of oblique photography 3D models, their production cost is high and the production process is complicated. Some 3D model texture details contain sensitive information. Unauthorized dissemination or abuse will not only bring huge commercial losses, but may even endanger national security. Therefore, it is necessary to adopt effective technical means to protect their safe use and copyright.

[0003] Digital watermarking technology is an information hiding technology that embeds invisible information in data to protect its copyright and integrity. This technology can ensure that watermark information can still be extracted and verified during the dissemination and use of data, thereby preventing illegal copying and tampering. It has been widely used in copyright protection of video, audio and two-dimensional geographic data. Digital watermarking technology not only plays an important role in preventive measures for copyright protection, but also provides effective evidence support as a technical means for copyright identification afterwards. In the event of copyright disputes, the embedded digital watermark can be used to prove the originality and ownership of the work and help the creators protect their rights and interests in law. Generally, digital products that need to be protected can be embedded with watermarks by certain methods according to the use environment and robustness requirements, combined with the characteristics of the data itself, and digital watermarking technology can be used to protect data security. The existing watermarking algorithms for oblique photography 3D model data usually include methods based on scale-invariant features, vertex coordinate clustering, vertex sorting, and feature point selection based on vertex normal vector angles. These methods are not very robust to the scaling, random addition and deletion of points, combined attacks and large-scale cropping of oblique photography 3D model data.

[0004] The feature points of the oblique photography 3D model represent the basic geometric or texture features of the model. The present invention uses principal component analysis (PCA) to calculate the curvature values ​​of the local neighborhoods of vertices at different levels, and selects vertices with larger curvature values ​​as feature points according to a threshold. This method makes the feature points have better stability under cropping and random addition and deletion of points.

[0005] The method of embedding watermarks in groups is a common method to resist cropping attacks. By grouping each vertex using the weighted distance minimum principle, the grouping method can effectively resist cropping attacks.

[0006] Normalization is a key step to ensure that data remains consistent at different scales. The height differences between vertices and feature points in each group are calculated in turn, and then normalized and used as the carrier for watermark embedding, so that the watermark information still has high detectability and anti-destruction properties under scaling attacks.

[0007] The present invention comprehensively utilizes the advantages of multi-level curvature, weighted distance grouping principle and data normalization processing, and proposes an oblique photography 3D model watermarking algorithm based on multi-level curvature and weighted distance. The experiment protects the copyright of oblique photography 3D model data during storage, transmission and use. Summary of the invention

[0008] In view of this, the present invention aims at the problems that the existing watermarking algorithms for oblique photography three-dimensional models have insufficient robustness against scaling, large-scale cropping and combined attacks, and proposes a watermarking algorithm based on multi-level curvature and weighted distance. The method first uses PCA to calculate the curvature values ​​of vertices in different levels of neighborhoods, and the vertices with curvature values ​​higher than the threshold are selected as feature points. Secondly, the vertices are grouped according to the principle of minimum weighted comprehensive distance between feature points and vertices. Finally, the height difference between the vertex and the feature point in each group is normalized and used as an information carrier, and the corresponding relationship between the watermark information and the embedded position is established using the QIM mapping idea to complete the embedding of the watermark information. The robustness of the algorithm is enhanced by using the group embedding and normalization method. The watermarking method has strong robustness against common geometric attacks, cropping attacks, reordering attacks and combined attacks, and has good invisibility, providing a new solution for copyright protection of oblique photography three-dimensional models.

[0009] To achieve the above object, the present invention provides the following solutions:

[0010] The watermarking method of oblique photography 3D model based on multi-level curvature and weighted distance includes three parts: watermark information generation, watermark embedding and watermark extraction:

[0011] The steps to generate watermark information are as follows:

[0012] S1: Read the original binary watermark image and apply the Logistic chaotic system to scramble the encrypted watermark image;

[0013] S2: Binarize the obtained watermark image to obtain a binary sequence;

[0014] The steps of watermark embedding are as follows:

[0015] S3: Read the vertices of the original oblique photography 3D model and calculate the curvatures of the vertices at different levels;

[0016] S4: Setting a threshold, setting vertices with curvature values ​​greater than the threshold as feature points;

[0017] S5: Merge feature point sequences at different levels;

[0018] S6: Calculate the weighted distance between the vertex and the feature point;

[0019] S7: Group all vertices according to the minimum distance principle;

[0020] S8: In the grouping, the height difference between the vertex and the feature point is calculated, and the difference is normalized;

[0021] S9: Calculate the watermark index using the normalized difference value, and embed the watermark into the normalized difference value;

[0022] S10: embedding watermarks into all groups of the oblique photography 3D model in sequence according to the above steps to obtain an oblique photography 3D model containing a watermark;

[0023] The steps of watermark extraction are as follows:

[0024] S11: The watermarked oblique photography 3D model obtained in step S10 is subjected to the same processing as steps S3 to S7 in the watermark embedding stage, and after being grouped, the height differences between the vertices and the feature points in each group are calculated in turn, and the height differences are normalized;

[0025] S12: Calculate the index value of the watermark using the normalized difference;

[0026] S13: Extract watermark information using QIM quantization method and voting principle;

[0027] S14: End.

[0028] The method of the invention is advanced and scientific, ensuring the effective extraction of watermark information and good robustness, and can provide a new solution for the safe use and copyright protection of oblique photography 3D models. Experiments show that the method has good robustness against attacks such as geometric transformation, large-scale cropping, reordering, random addition and deletion of points and combination of oblique photography 3D models, and has good use value. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only schematic diagrams of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0030] Figure 1 A flow chart of the method for watermarking a three-dimensional model of oblique photography based on multi-level curvature and weighted distance provided by the present invention;

[0031] Figure 2 Data used in the experiments provided by the present invention;

[0032] Figure 3 The present invention provides (a) original digital watermark image (b) watermark image after Logistic chaotic scrambling (c) restored watermark image;

[0033] Figure 4 The cropping attack and extracted watermark results provided by the present invention. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] The technical content, structural features, objectives achieved and effects achieved of the present invention are described in detail by taking an example, and are described in detail in conjunction with specific implementation methods.

[0036] 1. Watermark information generation

[0037] Step 1: Read the original binary watermark image, apply the Logistic chaotic system to scramble the watermark image, and obtain the scrambled watermark image;

[0038] Step 2: Binarize the scrambled watermark image to obtain a binary watermark sequence W = {w[j]}, j∈[0,l-1], where l represents the watermark length;

[0039] 2. Watermark Embedding

[0040] Step 3 Data preprocessing. Read the vertex coordinates P of the original oblique photography 3D model i =(x i,y i ,z i ),i∈(1,2,,L), where L represents the number of vertices of the three-dimensional model;

[0041] Step 4: Select different levels K = {K1, K2, K3}, where K1, K2, K3 represent the number of vertices in the neighborhood of a vertex;

[0042] Step 5: Use PCA to calculate the curvature values ​​of vertices in neighborhoods at different levels, obtain different curvature sequences, set a threshold ε, select vertices with curvature values ​​higher than ε as feature points, and obtain feature points Fe1, Fe2, Fe3 at different levels.

[0043] Step 6: Merge the feature points extracted at different levels to obtain the final feature point sequence Fe;

[0044] Step 7 calculates the weighted comprehensive distance between all vertices P and Fe in the 3D model based on the geometric space distance and feature space distance. The calculation formula is as follows:

[0045] d(Fe j ,P i )=α×d enclidean (Fe j ,P i )+β×d featire (Fe j ,P i )

[0046]

[0047] d feature (Fe j ,i)=‖Fe j -P i ‖

[0048] Among them, α and β are equilibrium parameters, d enclidean Represents the geometric space distance; d feature represents the feature space distance, P represents the vertex, and Fe represents the feature point;

[0049] Step 8: Group the vertices P according to the principle of minimum weighted comprehensive distance;

[0050] Step 9 calculates the difference D between the Z coordinates of all vertices in each group and the Z coordinates of the feature points in turn, and normalizes the difference to obtain Dˊ;

[0051] Step 10: Enlarge the normalized difference Dˊ by 10 u times, and calculate the index value index. The index value calculation formula is as follows:

[0052] index=[Dˊ×10 u ]mod l

[0053] In the formula, the symbol [] indicates rounding;

[0054] Step 11 Apply the QIM method to expand Dˊ by 10 n The embedding calculation formula is as follows:

[0055]

[0056] Step 12: Reduce the difference after embedding the watermark by 10 n times and perform anti-normalization;

[0057] Step 13 calculates the Z coordinate after embedding the watermark based on the Z coordinate of the feature point in each group and the denormalized difference;

[0058] Step 14 replaces the original Z coordinate with the Zˊ coordinate embedded with the watermark to obtain the oblique photography 3D model Pˊ containing the watermark.

[0059] 3. Watermark extraction

[0060] Step 15: The watermarked oblique photography 3D model obtained in Step 14 is subjected to the same processing as Steps 3 to 8 in the watermark embedding stage, and after being grouped, the height differences between the vertices and the feature points in each group are calculated in turn and normalized;

[0061] Step 16 uses the normalized difference to calculate the index value index;

[0062] Step 17 Since there is a one-to-many relationship between the watermark and the embedded position of the watermark, the QIM quantization method and voting principle are used to extract the watermark in the watermark extraction process, and the extraction result Wˊ is recorded. The quantization value R is the same as that during embedding;

[0063] Step 18 converts the extracted watermark information Wˊ into a watermark image;

[0064] Step 19 is finished.

[0065] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for watermarking a three-dimensional model of oblique photography based on multi-level curvature and weighted distance, characterized in that: It includes three parts: watermark information generation, watermark embedding and watermark extraction: The steps to generate watermark information are as follows: S1: Read the original binary watermark image and apply the Logistic chaotic system to scramble the copyright image; S2: Binarize the scrambled watermark image to obtain a binary watermark sequence W = (w[j]}, j∈[0, l-1]); The steps of watermark embedding are as follows: S3: read the vertex coordinates of the original oblique photography 3D model, and select different levels K = {K1, K2, K3}, use PCA to calculate the curvature values ​​of the vertices in the neighborhoods of different levels, obtain different curvature sequences, set a threshold ε, and take the vertices with curvature values ​​higher than ε as feature points, and obtain feature points Fe1, Fe2, Fe3 at different levels; S4: Merge feature points at different levels to obtain the final feature point sequence F e ; S5: Calculate all vertices P and F in the 3D model e The weighted distance between them is calculated by grouping the vertices P according to the principle of minimum weighted distance. The calculation formula is as follows: d(Fe j ,P i )=α×d enclidean (Fe j ,P i )+β×d feature (Fe j ,P i ) Among them, α and β are equilibrium parameters, d enclidean Represents the geometric space distance; d feature represents the feature space distance, P represents the vertex, and Fe represents the feature point; S6: Calculate the difference D of the Z coordinates of the vertices and feature points in each group in turn, normalize the difference to get D′, and enlarge D′ by 10 u times, and calculate the index value index. The index value calculation formula is as follows: index=[D′×10 u ]mod l In the formula, the symbol [] indicates rounding; S7: Apply the QIM method to expand D′ by 10 n After the watermark is embedded, the embedding calculation formula is as follows: S8: Reduce the difference D″ after embedding the watermark by 10 n times and perform anti-normalization; S9: Calculate the Z′ coordinate after embedding the watermark according to the Z coordinate of the feature point in each group and the denormalized value; S10: Replace the Z coordinate of the original vertex with the Z′ coordinate after embedding the watermark, and obtain the oblique photography 3D model P′ containing the watermark; The steps of watermark extraction are as follows: S11: The watermarked oblique photography 3D model obtained in step S10 is subjected to the same processing as steps 3 to 5 in the watermark embedding stage, and after being grouped, the height differences between the vertices and the feature points in each group are calculated in turn, and the height differences are normalized; S12: Calculate the index value using the normalized difference, and extract the watermark information using the QIM quantization method and voting principle; S13: End.

2. The method for watermarking a 3D model of oblique photography based on multi-level curvature and weighted distance according to claim 1, characterized in that: In step S7, the watermark embedding operation is completed using the watermark information generated from S1 to S2.

3. According to claim 1 or claim 2, the method for watermarking a 3D model of oblique photography based on multi-level curvature and weighted distance, the application scenario also includes copyright protection of 3D mesh data, characterized in that: This watermark method is used to authenticate the copyright of oblique photography 3D models.

Images