关键词: 逆向工程, 三角形网格, 神经网络, 边缘误差, 散乱点云

Abstract: An approach based on the Self-Organizing Feature Map（SOFM） neural network has been developed to reconstruct triangle mesh for the unorganized digitized point cloud.However the approach suffers from boundary problems.A three-steps training method is proposed in order to reduce the boundary error.All the neurons of the mesh model are trained directly over the unroganized digitized point-cloud.Only the boundary neurons of the mesh model is undergo trained by the boundary points of the digitized point-cloud.Only the winner neurons with respect to the corner points are trained by the corner points of the boundary points.As a result of applying the proposed training method，the boundary error is greatly reduced and the mesh is drawn toward the sampled object with higher precision compared with the original SOFM training algorithm.The feasibility of the developed training method is demonstrated on two examples.

Key words: reverse engineering, triangle mesh, neural network, boundary error, unorganized point cloud