基于改进边界点排斥的激光点云模型孔洞修复

Key Points

• The aim is to develop an effective method for identifying and repairing holes in laser point clouds.
• Utilized weighted local optimal projection algorithm to optimize point cloud distribution.
• Employed normal vector analysis and local density constraints for precise hole boundary identification.
• Implemented dynamic repulsion field model to generate filling paths based on topological relations.
• Applied median filtering to ensure smoothness consistency between repaired and original point clouds.
• Achieved a hole repair rate of 98.7% in the Camel model.
• Point cloud density difference was only 4.2%.
• Repair time was significantly faster than methods like LS-SVM, Poisson, PU-Net, and SeedFormer, being only 11.7% of SeedFormer's time.

Abstract