基于局部加权线性回归的电离层层析重构

Ionospheric Tomography Reconstruction Based on Local Weighted Linear Regression

电离层层析重构的精度和穿过电离层的观测射线的分布和数量息息相关,由于地面观测站与卫星数量和分布的局限性,电离层大量格网内没有射线通过,进而使得没有射线穿过的格网迭代收敛后的像素值与初始值相同;然而实际反演过程中,初始值通常是由精度不高的经验电离层模型给出,从而使得电离层层析重构的精度不高。围绕这一问题,将机器学习领域的局部加权线性回归算法拓展应用到电离层层析重构技术中,依据电离层格网空间内像素的电子密度空间平滑分布特性,为没有射线穿过的格网通过局部加权线性回归算法计算格网像素值,解决了没有观测射线穿过的格网对初值的过度依赖问题,进而提高电离层层析重构的精度。数值模拟实验和实测数据的仿真结果表明新算法相比于传统层析算法更加有效、可靠和优越。

The accuracy of ionospheric tomography reconstruction is closely related to the distribution and quantity of observational rays passing through the ionosphere.Due to the limited number of terrestrial stations’s and satellites’s uneven distribution,a large number of grids in the ionosphere do not pass the ray,and the grid pixel values without rays pass through after the iteration with the same values as the initial values.In the actual inversion process,the initial value is usually given by the empirical ionospheric model with poor accuracy,so that the accuracy of ionospheric tomography reconstructions is not high.Focusing on this issue,the locally weighted linear regression algorithm in machine learning was extended to the ionospheric tomography reconstruction.According to the spatial distribution of the electron density in the ionosphere grid pixels,the pixel value is calculated by the local weighted linear regression algorithm for the grid without rays passing through.It solves the problem of over-reliance on the initial value of the grid through which no observed rays pass,thus improving the accuracy of ionospheric tomography reconstruction.The simulation results of numerical simulation and measured data show that the new algorithm is more effective,reliable and superior to traditional tomography algorithms.