基于 RBF 神经网络的点云孔洞修复算法研究

针对逆向工程中三维点云模型产生的孔洞问题, 文章提出一种基于径向基函数(radial basis function, RBF)神经网络映射的点云孔洞修复优化算法。首先对点云数据进行预处理, 检测孔洞边界点和拓展边界点, 并对边界点进行平滑处理; 然后将孔洞边界点投影到不同坐标轴平面上, 选择投影面积最大的孔洞坐标系作为映射坐标系, 将孔洞映射到坐标系中, 使用支持向量机(support vector machine, SVM)进行初步修复; 最后使用 RBF 神经网络映射点云, 优化孔洞修复算法进行孔洞填充。实验结果表明, 该算法能快速有效地填充孔洞, 填充效果优于其他算法, 且在填充孔洞的同时保证其拓扑性。

Aiming at the problem of holes generated by three-dimensional point cloud model in reverse engineering, this paper proposes an optimized point cloud hole repair algorithm based on radial basis function (RBF) neural network mapping. Firstly, the point cloud data is preprocessed, the hole boundary points and the extended boundary points are detected, and the boundary points are smoothed. After that, the hole boundary points are projected onto different coordinate planes, and the hole coordinate system with the largest projection area is selected as the mapping coordinate system. The hole is mapped into the coordinate system, and the support vector machine (SVM) is used for preliminary repair. Finally, the RBF neural network is used to map the point cloud to fill the hole. Experimental results show that the algorithm can quickly and effectively fill holes, outperforming other algorithms in filling effect while preserving topology throughout the filling process.