抗投影变换的矢量地理数据水印算法

在矢量地理数据水印算法研究中,以往研究较多考虑增删点、裁剪、平移、旋转等攻击方式,而对抗投影变换的攻击方式研究较少。由于投影变换在GIS中具有重要意义,故本文提出了一种抗投影变换的矢量地理数据水印算法。水印嵌入前,对待嵌矢量地理数据利用道格拉斯算法进行压缩,使用四叉树分块选取特征点,提取并保存特征点及其属性信息;水印嵌入采用坐标映射和量化机制以增强水印算法的鲁棒性;检测水印时,将待检测数据与原始特征点进行属性信息匹配,匹配成功的同名点采用二元三次多项式进行最小二乘法拟合,根据拟合的多项式系数对待检测数据进行投影变换,最终实现水印信息的提取。实验结果表明,本方法能抵抗投影变换攻击、增删点、几何变换以及它们的复合攻击,具有较好的可行性和实用性。

A Watermarking Algorithm Resisting to Projection Transformation for Vector Geographic Data

Among the researches on watermarking algorithms for vector geographic data, more attentions have been paid on the attacking methods such as increasing points, deleting points, translation, rotation and so on. However, there is scarce research on watermarking algorithms resisting the attack of projection transformation. As a common data processing method in geography, projection transformation would affect and ruin the watermark embedded by normal ways in vector geographic data. Therefore this paper proposes an anti-projection watermarking algorithm for this kind of data based on the correspondence points matching. First, feature points of the data are extracted using Douglas-Peucker compression and quad-tree partition. And the points are stored with their attribute information together. Then the watermark is embedded based on coordinate mapping and quantization mechanism, which would enhance the robustness of the watermarking algorithm. Data containing watermark can now be distributed and shared. When it is needed to detect the watermark from the suspect data, the detected one will be compared and matched with the points stored according to their attribute information. If the corresponding points are matched, the project transformation of them can be realized by the bivariate cubic polynomial. When calculating the coefficients of the polynomial, the least square method and QR decomposition are used to improve its accuracy. After the calculation, the polynomial will help to transform the projection coordinate system from the detected data to the original data. And the watermark will be able to be detected after the transformation if enough accuracy is ensured. Experiments have been conducted to prove the proposed algorithm is robust against the attacks of projection transformation, adding data, deleting data, ordinary geometric transform and their compound attacks. Results of the experiments show that the algorithm has a good feasibility and can be taken into practice easily.