Point cloud feature edge point curve fitting

Structured light edge point clouds suffer from scattering noise and non-uniform sampling, rendering traditional discrete fitting inadequate. We propose a one-stage parametric reconstruction framework that bypasses intermediate discrete representations. By integrating mask-driven dynamic control point regression with a differentiable geometric constraint solver, our model directly predicts topologically correct primitives (Bézier curves, lines, arcs) from scattered points. Demonstrating high robustness against scanning noise, it achieves a Chamfer Distance of 0.0038, and a Hausdorff Distance of 0.0764 on the ABC dataset. Preliminary self-consistency tests on real-world sensors demonstrate the method's robustness, offering a potential 3D edge parameterization solution for reverse engineering.