3D segmentation has been a hot spot in numerical geometry processing. But the accuracy of the segmentation methods can be easily affected by the types of models because their sensitivity to different models is uneven. To address this problem, we propose a semi-automatic 3D mesh segmentation algorithm based on harmonic field. Firstly, our algorithm utilizes the strokes of users as constraints on the harmonic field of the mesh surface. Secondly, a smooth harmonic field based on Laplacian operator and Poisson equation is introduced. Through the generated harmonic field, the correct weights are selected to further fit the geometric characteristics of the grid. Afterwards, we find a set of most suitable isolines on the harmonic field as the segmentation path. Finally, a mesh density enhancement method is designed, which optimizes sub-graphs after segmentation. Experimental results demonstrate that the effectiveness of our proposed algorithm. Moreover, the semi-automatic 3D mesh segmentation algorithm can better understand the intention of users.

We propose a semi-automatic 3D mesh model segmentation algorithm based on harmonic field. Firstly, our algorithm ultilizes the strokes of users as constraints on the harmonic field of the mesh surface. Secondly, A Laplace operator and Poisson equation based smooth harmonic field is introduced, and the generated harmonic field is made to further fit the geometric features of the mesh itself by choosing the correct weight values. A set of isolines are sought on the mesh surface harmonic field with the previous-mentioned constraints. The most appropriate isoline is selected from them as the segmentation path. Finally, after selecting the optimal segmentation isolines, a mesh density enhancement method, which optimizes sub-graphs after this segmentation by mesh density enhancement, is introduced. Experimental results demonstrate the effectiveness of the proposed segmentation method. Moreover, the semi-automatic mesh segmentation method can better understand the intention of users, and achieve a better segmentation effect.