Abstract: In order to improve the problems of over-segmentation,obvious zigzag segmentation lines,and too much human intervention in existing 3D mesh model segmentation methods,a segmentation method based on energy optimization and distinction was proposed.Firstly,it used two more robust features,i.e.,energy and distinction,to improve the accuracy of the segmentation boundaries.Secondly,based on the energy,distinction,and concavity,the segmentation points were found;by using the adjacency of the points,the segmentation points sets were obtained;and the segmentation lines were obtained by refining the segmentation points sets based on the corrosion algorithm.Finally,closed segmentation lines were constructed based on the breadth-first search algorithm and the minimum energy principle.In addition,a Dijkstra type optimization method was provided to optimize the shape and position of the segmentation lines.Experiments on the Princeton segmentation benchmark were carried out,and comparisons with seven general segmentation methods under the Princeton Shape Benchmark were done.The most important index,called the Rand index,is 0.21 higher than the seven other methods in average,shows that the proposed method can effectively get more meaningful segmentation results.