Advanced Reverse Engineering Methods Based on Integrated Global and Local Techniques

特别是，需要一种快速的方法来自动构建分段C2光滑的水密的NURBS曲面与任意拓扑结构的扫描云的尖锐边缘。这种方法将允许使用商业CAD系统自动处理扫描的复杂工程对象。在现有技术中，这个问题尚未解决。海法理工学院开发了一种基于神经网络（NN）技术的三维网格重建方法，以避免奇异点。柏林技术大学正在开发一种用于将曲面拟合到网格的快速细分方法。本研究建议将这些神经网络和细分算法集成到曲面重建过程中。该方法将产生符合现代工业标准的NURBS曲面。通过在柏林工业大学和Technion获得的经验，组合算法将能够解决当前逆向工程方法中的瓶颈。 该建议集中在两个研究所开发的表面重建过程。这个过程是任何RE系统的核心。RE系统以后可用于分布式劳动，并可用作远程制造系统的基础。

In particular, a fast method is needed for automatically constructing piecewise C2 -smooth watertight NURBS-surfaces with sharp edges over scan clouds of arbitrary topology. Such a method would allow the automatic processing of scanned complex engineering objects using commercial CAD-systems. In the current state of the art, this problem is unsolved. A 3D mesh reconstruction method based on the neural network (NN) technique that avoids singularities has been developed at the Technion Haifa. A fast subdivision method for fitting surfaces to a mesh is being developed at the Technical University of Berlin. The research proposes integrating these NN and subdivision algorithms into the surface reconstruction process. The approach will yield NURBS surfaces that meet modern industrial standards. Through the experience gained at the TU Berlin and the Technion, the combined algorithm will be capable of solving the bottleneck in current reverse engineering methods. The proposal concentrates on the surface reconstruction process developed by the two institutes. This process is the core of any RE system. The RE system can later be used for distributed labor and can be used as a basis for a tele-manufacturing system.