Primitive-based Model Reduction for Huge 3D Meshes of Industrial Plants

We address the model reduction for huge 3D meshes of industrial plants, which typically consist of millions of triangles. The proposed method converts an input plant mesh into a set of primitive shapes, each of which is represented by an implicit representation. To this end, we initially partition the input mesh using geometric features, and then, an implicit function is fitted to each part such that the surface deviation is minimized. In this way, the proposed method achieves to reduce the model size to less than 10% of the original one. By sorting the primitives using a bounding volume hierarchy (BVH), our system can also visualize the model-reduced shape format approximately in real-time.