基于模糊聚类的Kohonen神经网络三角网格曲面重建

Triangle mesh surface reconstruction based on Kohonen neural network and fuzzy clustering technique

针对密集点云的三角网格曲面重建,提出一种用于数据精简和分块的神经网络算法:将模糊聚类方法与Kohonen神经网络算法结合.该算法具有按不同曲率进行曲面点云分块重建的能力,而且提高了自组织神经网络的效率.并应用该算法进行了仿真试验,建立了三角拓扑网格曲面,验证了算法的有效性.

To resolve the triangle mesh surface reconstruction problem pertaining to densely-distributed point clouds, a novel neural network algorithm is postulated for data extraction and classification. Accordingly, this approach integrates both Kohonen neural network and fuzzy clustering technique. It is found that this algorithm can not only reconstruct the point clouds regarding diverse curvature, but also improve the efficiency of Kohonen neural network. Finally, this approach has been proven effective by simulating a topological triangle mesh surface.