Constraint driven optimization of surface features from point cloud in reverse engineering

In order to capture the original design intent and improve the whole quality of reconstructed models, a practical technique is presented for geometric constraint driven optimization of surface features from point cloud in reverse engineering, including constraint decomposition and numerical solution. In the stage of constraint decomposition, coupled constraints of complex surface features are eliminated by Design Structure Matrix partitioning algorithm. A new clustering method based on multi-scale surface feature analysis is proposed to reduce the geometric constraint system and decompose it into constraint subset. In the stage of numerical solution, mathematical models of optimization are built with exponential penalty. The BFGS method is studied for stable numerical solution of surface feature model optimization. Approximate error and constraint satisfaction error is analyzed. And the results show that the proposed method can achieve a global optimization result by increasing the approximate error at low level and decreasing the constraint satisfaction error.