太空非合作遥感卫星三维重建点云的平面滤波方法

受太空环境光照条件弱、卫星高速运动、卫星表面纹理匮乏等因素的影响，太空非合作遥感卫星的三维重建点云中往往会存在大量噪声，给太空非合作目标的准确探测带来了非常不利的影响。文章以非合作遥感卫星的形状特征作为先验知识，提出一种三维重建点云的平面滤波方法。针对点云噪声的不同特点，首先利用统计分析滤波方法消除离群离散点噪声；其次，利用改进的区域生长聚类方法，分割出点云模型中的平面结构与离群点簇噪声；然后，分别用RANSAC算法和最小二乘法对平面结构进行滤波，并将平面结构附近的噪声点垂直投影至平面内，去除模型表面噪声。实验结果表明，该优化方法可以有效去除非合作遥感卫星三维重建点云中的大部分噪声，得到卫星平面结构的光滑点云数据。

Plane Filtering of 3D Reconstruction Point Cloud for Space Non-cooperative Remote Sensing Satellites

Due to the weak illumination conditions in space environment, the high speed motion of satellite and the lack of surface texture of satellite, there is usually a lot of noise in the 3D reconstruction point cloud of space uncooperative remote sensing satellite, which has a very adverse effect on the accurate detection of space uncooperative target. In this paper, a plane filteringmethod for 3D reconstruction of point cloud is proposed based on the shape characteristics of non-cooperative remote sensing satellites as prior knowledge. According to the different characteristics of point cloud noise, the outlier discrete point noise is eliminated by statistical analysis filtering method. Secondly, the plane structure and outlier cluster noise in the point cloud model are separated by the improved region growth clustering method. Then, RANSAC algorithm and least square method were used to filter the plane structure respectively, and noise points near the plane structure were vertically projected into the plane to remove the surface noise of the model. The experimental results show that the proposed method can effectively remove most of the noise in the point cloud reconstructed by the cooperative remote sensing satellite and obtain the smooth point cloud data of the satellite plane structure.