Abstract: The existing mesh segmentation algorithms often require manual intervention or large computational load, and are generally found to be very sensitive to the model shape variations. To tackle these problems, this paper concentrates on the concave vertex investigation and presents an efficient Minima Rule Boundary Detection approach for robust mesh segmentation. First, we smooth the complex mesh model with vertex tolerance constraints to reduce the noisy impact, and then detect the Minima Rule concave vertexes by calculating the normalized convex concave signal of each vertex. Accordingly, the reasonable boundaries can be well constructed by linking the concave vertex with Skeletonizing algorithm and Dijkstra algorithm. Finally, these boundaries are smoothed and refined by means of the three-dimensional snakes, through which the semantic blocks within the mesh model can be efficiently segmented. The extensive experiments have demonstrated that the proposed algorithm is able to produce meaningful results rapidly and effectively.