基于显著性分析的网格模型语义特征线提取

现有的基于局部微分属性度量的网格模型特征线提取方法较难描述模型全局语义特征。为解决该问题,提出一种基于显著性分析的、符合人类视觉注意力机制的语义特征线提取算法。首先以谱图理论为工具,构建网格模型光顺三维基准面,获取网格顶点显著性重要度;利用离散拉普拉斯—贝尔特拉米算子方向属性构建网格语义特征区域,提取特征区域骨骼线,对骨骼线进行优化后得到模型语义特征线。该算法完全去除了模型局部微分几何量的度量,提取的特征线能很好地完成全局语义特征描述。实验结果证明了算法的有效性。

Semantic feature lines extraction on mesh model based on saliency analysis

The available mesh feature lines extraction methods based on the differential geometry properties cannot describe the global semantic characteristics well.To solve the problem,this paper proposed a method of extracting semantic feature lines of mesh model based on saliency analysis and human attention mechanism.Firstly,the method used the spectral theory to construct a smooth three-dimensional datum of the mesh model,obtained the salient vertices of the mesh.And then the method used the directional attributes of the discrete Laplacian-Beltrami operators to construct the mesh semantic feature areas,extracted and optimized the skeletons of the feature regions,and obtained the semantic feature lines of the model finally.The method hasn’t need to measure the local differential properties of the mesh at all,and the experimental results show that the method can accurately extract the semantic feature lines,which can describe the global meaningful features of the model well.