面向LINS点云的迭代特征扩展椭球滤波算法OA

To address the challenges of severe noise aliasing in single-photon point clouds acquired by the laser integrated navigation sensor(LINS)in the Tianwen-2 mission,this paper proposes an iterative feature extension ellipsoid filtering algorithm based on manifold evolution.By leveraging the geometric structural differences between signal points(local 2D manifolds)and background noise(3D random distribution),the algorithm constructs an adaptive ellipsoidal neighborhood evolution model with volume constraints.This mechanism drives the filtering neighborhood to evolve from isotropic to anisotropic,enabling precise separation of background noise adhering to the terrain surface.Validation using LINS ground test data demonstrates that the proposed algorithm achieves a root mean square error(RMSE)as low as 0.061 m in scenarios with low SNR and drastic photon density variations.Compared with the statistical outlier removal(SOR)and density-based spatial clustering of applications with noise(DBSCAN)algorithms,the proposed algonithm achieves Hausdorff distance(HD)reductions of 67.2%and 70.1%,respectively,effectively eliminating diffuse outlier noise within the field of view.The results verify the algorithm's capability for terrain recovery under strong background light interference,providing reliable data preprocessing support for high-precision relative navigation in asteroid exploration missions.