|
| Author | Input model | Method | Limitation |
|
| Zhang et al. [11] | Mesh model | Using Gaussian curvature to mark vertices and detect the boundaries of model, then segmented by region growing algorithm | The boundaries detected may not be correct, because only vertices are considered |
| Lavoué et al. [12] | Mesh model | Using the standard k-means algorithm to classify vertex principal curvatures, then segmented based on the region growing and merging of the labeled mesh vertex | It is difficult to get the reasonable curvature tensor on the sparse region, and the output depends heavily on the choice of the user defined threshold |
| Mangan and Whitaker [13] | Mesh model | Partitioning a mesh model by the total curvature of the surface as an indication of region boundaries, segmented along high curvature regions | The boundaries between components are often jagged |
| Sun et al. [14] | Mesh model | Detecting the edge based on the eigen analysis of the surface normal vector in a geodesic window, then used to segment the surface mesh using a watershed algorithm | The boundaries detected may not be correct, because only vertices are considered |
| Brown et al. [16] | Mesh model | Treating the mesh model as a weighted graph, then achieving segmentation by minimum graph cut | It needs user to interact where he wants the partition |
| Golovinskiy and Funkhouser [17] | Mesh model | Dual graph is first built, then segmentation is achieved by clustering similar meshes and obtaining minimum graph cuts | This method is not usual in CAD models |
| Mejia et al. [18] | Mesh model | Treating the mesh model as a connected and undirected graph and relabeling the vertices of the graph based on the Fiedler vector; then, segmentation will be defined by cutting thresholds on it | Geometric information is not considered and requiring mesh need to be homogeneous |
| Miandarhoie et al. [3] | Mesh model | Partitioning the point clouds of mesh model into individual regions can be approximated using extruded surface | It lacks semantic meaning |
| Xiao et al. [29] | Mesh model | Dividing the triangles into sparse and dense regions by the agglomerative hierarchical clustering method, using gauss sphere and mean shift method to segment sparse and dense regions into primary surfaces | Segmentation information loss occurs when the method is used for CAD models [30] |
| Shlafman et al. [25] | CAD model | Using a k-means clustering algorithm to segment model, and the meaningful components of the objects are found | The boundaries between the patches are often jagged and not always correct |
| Lavoué et al. [31] | Mesh model | Classify the vertices according to the absolute values of their principal curvatures into k clusters using the k-means algorithm | The parameter k needs to be defined manually |
| Wang et al. [27] | CAD model | Partitioning by clustering based on multiobjective optimization | The result is affected by partitioning the face selected and partitioning the order |
| Wang et al. [15] | CAD model | Represented by the attributed adjacent graph and segmented by ant clustering algorithm | Parameter settings are subjective to be changed according to different types of models |
| Ma [28] | CAD model | Employed a surface region segmentation method for CAD models | Segmented region is impacted by the order of the merging and lack of semantic meaning |
|
