Robust blind watermarking is the important method for digital works copyright protection. A novel blind watermarking method on 3D meshes was proposed. PCA (primary component analysis) was used for 3D meshes- preprocessing to make a unique model posture. After making division of the 3D meshes based octree, octree structure could be obtained. A random scrambling was made for the RGB(red, green, blue) meaningful watermark image before embedding it into the bottom octree nodes. This method can be used directly on points-cloud, as it does nothing with the topology of the meshes. It is robust against translation, rotation, uniform scaling and random vertex reordering attacks. In addition, a method to generate ‘Zero-Watermark’ based on higher codec of octree nodes was given. The ‘Zero-watermark’ was then registered in the third-party of notarization, and it can be robust against simplification, noise addition and remeshing attacks in some extent. This method has a big volume for embedding and high popularity.