Space target reconstruction method based on accelerating Gaussian mixture model

Abstract

A space target reconstruction method based on an accelerating Gaussian mixture model comprises the steps that point cloud data are read in; according to the number K of components of the given Gaussian mixture model, K points are screened out from point clouds to serve as initial clustering centers by using a K-means++ algorithm; the point clouds are clustered into K clusters by using a K-means algorithm; according to the situations of points included in each cluster, the weight, the mean value and a covariance matrix are initialized to serve as initial values of parameters of the Gaussian mixture model; and the parameters are iteratively updated by using the Gaussian mixture model to obtain final estimation values. According to the space target reconstruction method, the target three-dimensional structure reconstruction problem in close-distance perception of a spacecraft is researched, the target point clouds include a large number of points, and the space target reconstruction method is realized by combining algorithm procedure design and CPU-GPU heterogeneity; and in the reconstruction process, the calculating speed is high, and robustness is high.

Description

technical field [0001] The invention relates to the field of short-distance perception of spacecraft, in particular to a space target reconstruction method based on an accelerated Gaussian mixture model. Background technique [0002] Point cloud can be used to intuitively describe the shape of the target, but for tasks such as further approach or capture, this form of description is far from meeting the task requirements, and it is necessary to further obtain an analytical description of the target shape on this basis. [0003] Gaussian Mixture Model (GMM), as a powerful parametric probability model, the core idea is to model the target as a weighted sum of multiple Gaussian distributions, which has been widely used in background modeling, image segmentation, super It has achieved fruitful results in various fields such as pixel generation and speech recognition. [0004] Jian et al first used GMM to model the point cloud in the process of two point cloud registration, and ...

AI Technical Summary

Problems solved by technology

This reconstruction method studies the problem of 3D structure reconstruction of the target in the short-distance perception of spacecraft, and solves the problems of large amount of calculation and weak robustness existing in the current method.

Images