Rough Registration Method for Point Cloud of Spatial Non-cooperative Target by Improving FPFH

It has become an important and difficult point to solve the problem of non-cooperative target pose measurement in space. The three-dimensional point cloud data of the target acquired by the laser radar can restore its three-dimensional morphology through point cloud registration, so as to solve the relative pose of non-cooperative target in space. In view of the fact that the traditional point cloud Iterative Closest Point (ICP) registration algorithm is prone to fall into local extremum when the initial value is not good, which leads to registration failure, a rough registration algorithm of 3D point cloud based on the combination of bounding box and FPFH descriptor using principal component analysis is proposed. In this method, PCA is used to construct a three-dimensional point cloud bounding box. The parameter threshold of FPFH describing sub-point cloud on-time matching is selected through the density of point cloud and the outer dimension of outer bounding box. The normal alignment is added to adjust the normal direction of point cloud. The experimental results show that the proposed method simplifies the complexity feature parameters selection, reduces the search range of the feature point matching, and provides the error of precise registration of about 3 degrees, to solve the spatial non-cooperative target relative position measurement technology provides technical reference.