倾斜摄影三维模型数字水印算法

倾斜摄影三维模型作为一种地理数据,其数据安全问题不容忽视,数字水印技术可以为其版权保护提供有效的解决方案。本文基于倾斜摄影三维模型数据特征,根据其垂直方向上的稳定性,提出了一种基于距离映射机制的三维点云数据盲水印算法。首先,提取倾斜摄影三维模型中的点云并对其按照高度值排序;其次,将点云间相邻点的高度差作为信息载体,利用映射思想建立水印信息与水印嵌入位置的对应;最后,通过修改坐标值的奇偶性完成水印信息的嵌入。检测水印时通过顶点提取、排序、水平距离与水印位的映射和统计相邻点水印嵌入位置的奇偶性实现水印信息的提取。实验表明,该算法具有良好的不可感知性,并对平移、裁剪和旋转等常见攻击具有较强的鲁棒性。

The Digital Watermarking Algorithm for 3D Models of Oblique Photography

The importance of data security cannot be ignored in 3D model data of oblique photography, which is an important kind of geographic data. Digital watermarking technique can provide powerful protection for oblique photography 3D models. In the past, among studies of 3D model watermarking algorithms, oblique photography 3D models were mentioned rarely and those algorithms cannot meet the demand for the data security of oblique photography 3D models. If a watermarking algorithm is designed in combination with the characteristics of oblique photography 3D model data, obviously, it will be applicable for corresponding 3D models. In this paper, a blind watermarking algorithm is designed based on the characteristic and vertical stability of oblique photography 3D model data. Before embedding watermark in data, some preprocessing steps must be completed. The 3D points will be extracted from oblique photography 3D model data, and sorted by z-coordinate from smallest to largest. Then, using height difference between neighboring points as the watermark information carrier, watermark information is embedded by modifying the parity of the height difference. In order to build a correspondence between embedding location and watermark information, the horizontal distance of neighboring points is used to complete mapping. In the process of watermark testing, points extracting and sorting should be finished firstly as in the process of embedding. Then information can be extracted by distinguishing the parity of the height difference. In this process, mapping is also used. In order to analyze the robustness of this algorithm quantitatively, some experiments were completed. The results show that the algorithm is robust against some usual attacks such as data cutting, translation and rotation.