鲁棒的三维模型零水印的研究

发布时间：2018-07-10 08:23

本文选题：鲁棒的三维网格模型零水印 + 积分不变量　；参考：《北京交通大学》2017年硕士论文

【摘要】：随着互联网技术的迅速发展,越来越多的数字产品进入公开网络环境,应用也越来越广泛,同时网络上的多媒体信息的版权盗用问题也显著增加。数字水印技术主要通过在多媒体内容中嵌入版权标志数据来确保内容的安全。近年来,随着三维扫描技术和计算机图形学的发展以及计算机性能的提高,3D打印技术的成熟,使得三维几何模型也面临所有权保护问题。而现有三维模型数字水印技术,多基于模型顶点、面片关系,很难抵抗3D扫描打印带来的重采样攻击,同时为了平衡三维模型水印鲁棒性和透明性之间的关系,本文提出了一种基于积分不变量的三维模型零水印算法,零水印是指利用三维模型的重要特征来构造水印信息,而不是修改三维模型的数据嵌入水印信息。该算法首先对模型利用 CPCA(Continuous Principal Component Analysis)的方法进行预处理,以实现三维模型的平移不变性、缩放不变性和旋转不变性。其次,对预处理后的模型按照Ray-Based的方法进行分组,结合上一步的操作,就可以使模型上的顶点得到一个固定顺序的分组,以便使受到攻击后的模型与原始模型构造出的水印信息可以保持一定的同步性。接着对每一分组中的数据按顺序分别构造邻域球、计算其球面积分不变量(这里本文提出两种计算方法——球冠近似法和基于球的表面积的计算方法)并归一化。最后通过比较计算出的归一化后的球面积分不变量的值与阈值(一般取1/2)的大小关系,进行水印信息的构造。水印信息构造完成后,我们需要把它结合时间戳一起注册到一个可信的第三方IPR(Intellectual Property Rights)数据库中,从而实现对三维模型的版权保护。这种技术很好地解决了不可见数字水印的透明性与鲁棒性之间的矛盾,并经实验证明,该算法可以很好地抵抗重采样攻击,对简化、细分、平滑、相似变换、顶点重排等常见攻击也具有较强的鲁棒性。本文最后针对上述提出的三维模型零水印算法,基于VS2010、C++、MFC、OpenGL以及半边数据结构,搭建了一个三维模型零水印系统。该系统主要包括三个功能模块,分别是:模型载入模块、模型交互及可视化模块、水印算法及水印信息的相关性计算模块。
[Abstract]:With the rapid development of Internet technology, more and more digital products enter the open network environment, and their applications are more and more extensive. At the same time, the problem of copyright embezzlement of multimedia information on the network also increases significantly. Digital watermarking technology mainly guarantees the security of content by embedding copyright mark data into multimedia content. In recent years, with the development of 3D scanning technology, computer graphics and the improvement of computer performance, 3D printing technology is mature, which makes 3D geometric model face the problem of ownership protection. In order to balance the relationship between robustness and transparency of 3D model watermarking, it is difficult to resist the resampling attack caused by 3D scanning and printing. This paper presents a zero watermarking algorithm for 3D model based on integral invariant. Zero watermarking is to construct watermarking information by using the important features of 3D model, not to modify the embedded watermark information of 3D model data. The algorithm preprocesses the model by CPCA (continuous Principal component Analysis) to realize the translation invariance, scaling invariance and rotation invariance of the 3D model. Secondly, the pretreated models are grouped according to Ray-Based method and combined with the previous operation, the vertices on the model can be grouped in a fixed order. In order to make the watermark information constructed by the attack model and the original model can maintain a certain synchronization. Then, the neighborhood sphere is constructed in order for the data in each group, and the spherical integral invariants are calculated (in this paper, two methods, the spherical crown approximation method and the surface area method based on the sphere), are presented and normalized. Finally, the watermark information is constructed by comparing the value of the normalized spherical integral invariant and the threshold value (generally 1 / 2). After the watermark is constructed, we need to register it with time stamp in a trusted third party IPR (intellectual property Rights) database to protect the copyright of 3D model. This technique solves the contradiction between the transparency and robustness of invisible digital watermarking, and it is proved by experiments that the algorithm can resist resampling attack very well, and simplify, subdivide, smooth, similar transform. Vertex rearrangement and other common attacks also have strong robustness. In the end of this paper, a zero watermarking system for 3D model is built based on VS2010 / MFC / OpenGL and half edge data structure. The system consists of three functional modules: model loading module, model interaction and visualization module, watermarking algorithm and the correlation calculation module of watermark information.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP309.7

【参考文献】