Branch and Bound Point Cloud Algorithm Registration Based on Neighborhood Curvature Features

Point cloud registration plays an important role in points-drive computer graphics as it affects modling quality directly.Aiming at the registration of multi-view point cloud data, It proposes a new registration method based on Branch and Bound(BNB)point cloud algorithm and neighborhood curvature features.The method introduces a new Zero -mean Normalized Cross-correlation Coefficient (ZNCC) to measure curvature similarity of the neighborhood of a point. The initial matching points is built.The transformation parameters of point cloud registration are obtained by the least square model, and the initial registration parameters are obtained.The global optimum solution is obtained by branch and bound method.The results show that the proposed algorithm can converge quickly to point clouds with apparent curvature features, and get the global optimum solution.