An integrated framework for surface reconstruction from incomplete point clouds based on constraint feedforward migration.

Abstract

3D point cloud surface reconstruction is crucial for intelligent transportation and industrial applications. Despite advances, external disturbances induce local point cloud gaps where existing methods fail to guarantee reconstruction accuracy, particularly when coexisting with intrinsic structural holes. To address this issue, we propose an integrated framework based on constraint feedforward migration, which reassigns global shape regulation from implicit reconstruction to point-domain support generation and couples point completion with surface realization. Furthermore, we introduce a gradient-constrained surface reconstruction approach, where non-zero level sets in the unsigned distance function (UDF) guide zero-level set optimization to enhance surface continuity and geometric accuracy. Extensive experiments on public datasets, real-world electrical equipment datasets, and large-scale autonomous driving scenarios (KITTI dataset) demonstrate the superior performance of our method. The results validate its robustness and potential for real-world applications in smart electrical systems, intelligent transportation, and related fields.