Research On Lossless Watermarking Technology For2D-vector Maps

With the rapid development of Geographic Information System (GIS),2D digital vectormaps have been widely applied in various fields to benefit the national economy. However, italso faces a serial of safety problems such as copyright protection, source tracking andauthentication. To this end, digital watermarking technology has shown great potential intackling the security problems. However, in some special occasions, the accuracy of therecovered2D vector data is highly demanded, since any disturbance to vector data induced bywatermark embedding is unbearable in many application scenarios. Under such acircumstance, the traditional watermarking technology can not address these issues. Thelossless watermarking can lossless recover the original data after watermark extraction, whichis suitable for vector maps.The research on lossless watermarking for vector maps is still in the initial stage,although some reversible watermarking and zero watermarking algorithms have beenproposed for vector maps, but these algorithms can’t fully address the security and robustnessproblems of vector maps in real-world application.The special data structure of2D vector maps has been discussed in this thesis,considering the vector map application environment, application requirements and robustnessagainst special watermark attacks, the author has studied the lossless watermarking algorithmswith high security and robustness. The contribution of this thesis can be summarized asfollows:Firstly, aiming at handling the large distortion issue in the traditional reversiblewatermarking approach, a perception-based reversible watermarking algorithm is proposed. Itis based on vertex density estimation to selects noise-insensitive area of polylines as coverdata, and embedding watermarks in DCT domain of vertex coordinates, which has improvedthe visual imperceptibility and robustness of the watermarking algorithm.Secondly, according to the data precision requirement and robustness requirementagainst special attacks of vector map, based on feature point extraction, a reversiblewatermarking over graph spectral domain is proposed. By constructing star-tree of featurepoints, the proposed algorithm selects the most optimal (Laplacian) spectral coefficients ascover data, which ensures visual imperceptions and watermark robustness, also maintains thebasic shape of vector maps with anti-simplification attack ability simultaneously.Third, aiming at addressing the capacity issue in secret transmission and embedding, aniterative embedding based reversible watermarking algorithm is proposed based on high theoretical analysis result, which ensured the proposed lossless watermarking scheme withhigh capacity and lower disturbance to original data.Then again, combined with content security and watermarking technology for copyrightprotection, a nonlinear scrambling-based reversible watermarking algorithm is proposed.Based on feature point extraction, a nonlinear scrambling scheme is proposed, it nonlinearlyscramble the feature points while maintaining its relative position, which avoids thehigh-precision vector data being illegally used by unauthorized users. The proposed algorithmreversibly embeds watermarks into coordinates of feature points, which meets the authorizedusers’ requirement for high precision data usage.Finally, A robust dual zero watermarking algorithms is proposed. This zerowatermarking scheme constructs watermarks with extracted feature information withoutdamaging the original data, which is more suitable for vector map copyright protection. Forthe extracted feature point that reflecting the outline of polyline and the non feature point thatreflecting the details of polyline, the proposed algorithm utilizes the feature point relationmodel established by BP artificial neural network and the matrix singular values to constructdouble zero watermark parameters, which can resist simplification, compression attacks andcommon geometric attacks, respectively. The experimental results show that the scheme candepict the overall characteristic of2D vector map, while being helpful to improve thewatermarking anti-attack capability, as well as achieving the overall protection purpose forthe vector map.