密集三维散乱点数据的拓扑矩形网格自组织压缩重建

探讨了曲面密集三维散乱点数据的拓扑矩形网格自组织压缩重建方法。建立了基于自组织特征映射神经网络的三维散乱点数据的拓扑矩形网格自组织压缩重建模型。该模型利用神经元对曲面散乱点的学习和训练来模拟曲面上的点与点之间的内在关系，结点连接权矢量集作为对散乱点集的工程近似化并重构曲面样本点的内在拓扑关系，实现曲面密集三维散乱点数据的自组织压缩。按矩形阵列侧抑制邻区训练调整网络神经元权重矢量，使网络输出层结点呈矩形阵列分布，可生成测量点集压缩后的拓扑矩形网格，可用于NURBS曲面重构。计算机仿真实验表明，所建模型可以实现三维密集散乱点数据自组织压缩，生成期望疏密程度和精度的双有序点列，重建矩形拓扑网格。

Self-organizing Extraction Reconstruction of Topologic Rectangular Mesh for Dense 3-D Scattered Data

Based on the self-organizing feature map(SOFM) neural network, an approach is developed to extract the dense 3-D scattered data and to produce the topologic rectangular mesh. The inherent topologic relations between the scattered points on the curved surface are reconstructed by the weight vectors of the neurons on the output layer of the neural network. The weight vectors of the neurons on the output layer of the neural network are used to approximate the dense 3-D scattered points, so the dense scattered points can be reduced to the reasonable scale, while the topologic feature of the whole scattered points remained. The region of the lateral inhibition is rectangle within which the neuron weight vectors are adjusted according to the SOFM training algorithm. The neurons on the output layer are distributed in the array of rectangle after training, thereby the topologic rectangular mesh in a way of high approximation is produced which can be used to reconstruct the surface with NURBS method. The computer simulation results show that this approach is satisfactory.