Reconstruction of Freeform Objects with Arbitrary Topology Using Neural Networks and Subdivision Techniques

In reverse engineering, laser scanned data is reconstructed into a CAD model. This paper presents a new reconstruction approach that integrates neural networks with subdivision techniques. The neural network technique creates a triangular mesh that approximates the shape of an object and detects its topology, where the subdivision approach applies smooth surfaces onto this mesh. The advantage of this method is that the reconstruction can be applied on objects with arbitrary topology, and the final model can be integrated with traditional CAD systems using a NURBS representation that preserves continuity. The feasibility of the method is demonstrated on freeform objects with arbitrary topology.