Nucleation distance fuzzy clustering orthogonal spectral imaging pose sensor calibration method

Abstract

The invention relates to space pose measurement and related sensor technologies, and aims to solve the problem of a dense matrix occurring when an orthogonal spectral imaging pose sensor adopts a large amount of scattered data for interpolation calibration and solve the problem of calibration of the orthogonal spectral imaging pose sensor. Therefore, the technical scheme adopted by the invention is as follows: regional division is carried out by adopting a fuzzy clustering algorithm based on a nucleation distance; a local expression of each region is fitted by adopting a global radial basis function interpolation algorithm; a weight function is created by adopting an inverse distance weighted interpolation Shepard method, and calculated by adopting a local tight support radial basis function in the Shepard method; the global expression is generated by calculation of the weight function and local expressions of all regions, and finally calibration of the orthogonal spectral pose sensoris achieved. The invention is mainly applied to space pose measurement occasions.

Description

technical field [0001] The invention relates to space pose measurement and related sensor technology, in particular to a method for calibrating a pose sensor for kernelized distance fuzzy clustering orthogonal spectroscopic imaging. Background technique [0002] The orthogonal spectroscopic pose sensor adopts dual linear array CCD to simulate a single area array CDD to achieve large-scale and high-precision spatial pose measurement. In order to cooperate with the image acquisition of the dual linear array CCD, the optical structure of the pose sensor is relatively complicated, and the assembly structure is complicated. Therefore, the pose sensor has various distortions, such as radial distortion caused by a large field of view, one-way distortion caused by a cylindrical mirror, and nonlinear distortion caused by assembly. In order to achieve high-precision measurement, distortion correction must be performed before calibration. Yang Qian et al. used a polynomial method to s...

AI Technical Summary

Problems solved by technology

In 2001, Beatson et al. proposed to use the alternate projection method for region division and weight function construction to improve the efficiency of radial basis interpolation, but the process of region division and weight function construction is complicated

In 2005, Yutaka Ohtake and others proposed that the area division adopts the idea of ​​octree, the local expression adopts piecewise polynomial function, and the weight function adopts B-spline curve construction to solve the problem of expressing local information and apply it to 3D image reconstruction, but the weight function More complicated, not easy to implement

In 2007, Michael Macri proposed the OctPUM method for the first time. This method is based on the idea of ​​quadtree and octree for two-dimensional and three-dimensional area division respectively, and uses local compactly supported radial basis functions to construct the leaf nodes of quadtree and octree. , to solve the problem of analog reconstruction of local information of 3D images, but the storage structure of quadtree and octree is complex, occupying a large memory space, and the operation efficiency is low

In 2010, Rio Yokota and others proposed to use parallel algorithm to realize radial basis interpolation to improve operation efficiency, but the weight function structure is still complicated

In 2018, Michal Smolik et al. proposed a divide-and-conquer algorithm for region division. The divided subregions can be of any shape. The local expression is represented by a locally compactly supported radial basis function. The weight function is constructed by bilinear interpolation. However, when the subspace When no data points or all points are included, the subspace needs to be re-divided, and the steps are complicated

Existing research methods do not consider the distribution of scattered data points when dividing regions, so it is impossible to ensure that the divided subregions contain an appropriate number of points to express local features

The weight function construction methods in some studies are more complicated and not easy to implement

Images