本文選題：缺失數(shù)據(jù)　切入點(diǎn)：插補(bǔ)　出處：《測(cè)繪科學(xué)》2017年09期 　論文類型：期刊論文

【摘要】：針對(duì)傳統(tǒng)基于空間插值和時(shí)間序列上的插值補(bǔ)全形變?nèi)笔?shù)據(jù)的方法在空間點(diǎn)位分布稀疏、觀測(cè)值連續(xù)缺失以及含有粗差的情況下插補(bǔ)效果不佳的問(wèn)題,提出了一種基于抗差Kriged Kalman Filter的形變?nèi)笔?shù)據(jù)插補(bǔ)方法。該方法是一種時(shí)空插值的算法,在空間點(diǎn)位分布稀疏時(shí)考慮時(shí)間上的相關(guān)性,在時(shí)間上出現(xiàn)連續(xù)缺失時(shí)考慮其他點(diǎn)位對(duì)插補(bǔ)點(diǎn)的影響,以提高插補(bǔ)缺失數(shù)據(jù)的精度。又將抗差估計(jì)融合到Kriged Kalman Filter中以抵抗形變數(shù)據(jù)中粗差對(duì)插補(bǔ)精度的影響。利用模擬數(shù)據(jù)及天津GPS地面沉降數(shù)據(jù)進(jìn)行了實(shí)驗(yàn)分析。結(jié)果表明:由于該法考慮了監(jiān)測(cè)點(diǎn)的時(shí)空相關(guān)性以及具有抗差性能,使得插補(bǔ)結(jié)果在空間點(diǎn)位稀疏、連續(xù)缺失或具有粗差的情況下都具有較高的插補(bǔ)精度。
[Abstract]:The traditional interpolation method based on spatial interpolation and time series is used to solve the problem of sparse distribution of spatial points, continuous absence of observation values and poor interpolation effect in the case of gross error. A deformation-missing data interpolation method based on robust Kriged Kalman Filter is proposed, which is a spatio-temporal interpolation algorithm, which considers the temporal correlation when the spatial point distribution is sparse. Considering the effect of other points on the interpolation point when there is a continuous loss in time, In order to improve the accuracy of interpolation missing data, the robust estimation is fused into Kriged Kalman Filter to resist the influence of gross error in deformation data on interpolation accuracy. The simulation data and Tianjin GPS ground subsidence data are used for experimental analysis. The results show that the method takes into account the spatio-temporal correlation of monitoring points and its robust performance. The interpolation results have higher interpolation accuracy when the space points are sparse, continuously missing or with gross errors.
【作者單位】： 中南大學(xué)測(cè)繪與遙感科學(xué)系;湖南省精密工程測(cè)量與形變?yōu)暮ΡO(jiān)測(cè)重點(diǎn)實(shí)驗(yàn)室;
【基金】：國(guó)家“973”項(xiàng)目(2013CB733303) 中南大學(xué)教師研究基金項(xiàng)目(2014JSJJ003)
【分類號(hào)】：P207
，

本文編號(hào)：1580062