Algorithm 1 Feature matching for point clouds based on PCV.

Require: The source point cloud ${\mathcal{P}}^s$ and the target point cloud ${\mathcal{P}}^t$. Ensure: Consistent feature correspondences ${\mathcal{C}}_{inlier}$ between ${\mathcal{P}}^s$ and ${\mathcal{P}}^t$.  1: Generate ${\mathcal{C}}_{initial}$ by matching local keypoint descriptors;  2: Initialize the iteration index $i=0$ and ${\mathcal{C}}_{inlier}=Ø;$  3: Define the initial voting set ${\mathcal{C}}_{voting}^0$ based on $s_f({\mathbf{p}}^s, {\mathbf{p}}^t)$;  4: while $i<N_{iter}$ do  5:    Calculate the compatibility score between $c\in {\mathcal{C}}_{initial}$ and $c_j\in {\mathcal{C}}_{voting}^{i-1}$ using Equation (5);  6:    Calculate the voting score for $c\in {\mathcal{C}}_{initial}$ using Equation (7);  7:    Update ${\mathcal{C}}_{voting}^i$ using Equation (8);  8:    $i=i+1$;  9: end while  10: Sort ${\mathcal{C}}_{initial}$ based on the scoring result at the last iteration;  11: Push top-scored correspondences into ${\mathcal{C}}_{inlier}$;  12: Return ${\mathcal{C}}_{inlier}$.