摘 "要:針對(duì)附參數(shù)約束等式方程的函數(shù)模型解算問題,提出一種對(duì)約束方程賦權(quán)值再聯(lián)合觀測(cè)方程統(tǒng)一解算的約束方程賦權(quán)統(tǒng)一解法。為驗(yàn)證此方法的可靠性和效率,基于矩陣分析法推導(dǎo)了適用于附參數(shù)約束方程解算的正交三角分解法和廣義奇異值分解法。通過算例對(duì)3種解算法進(jìn)行比較分析,結(jié)果表明:當(dāng)給約束方程附加合適范圍內(nèi)的權(quán)值時(shí),約束方程賦權(quán)統(tǒng)一解法獲得的參數(shù)估計(jì)值與矩陣分解算法的參數(shù)估計(jì)結(jié)果在要求的精度范圍內(nèi)一致;賦權(quán)統(tǒng)一解法與正交三角分解法的計(jì)算效率接近,廣義奇異值分解法計(jì)算效率較低;約束方程賦權(quán)統(tǒng)一解法更符合參數(shù)模型統(tǒng)一表達(dá)和平差準(zhǔn)則設(shè)定,有利于平差算法設(shè)計(jì)和誤差特性分析和研究。
關(guān)鍵詞:參數(shù)估計(jì);約束方程賦權(quán),統(tǒng)一解法;正交三角分解法;廣義奇異值分解法
中圖分類號(hào):P228 " " " " " " " " 文獻(xiàn)標(biāo)志碼:A " " " " " " " " "文章編號(hào):1008-0562(2024)03-0379-06
Research on a unified resolution method for the observation model applying weight to its parameter constraint equation
LU Jianbing1, YANG Rengui2
(1. Science and Technology Division, Lanzhou Resources amp; Environment Voc-Tech University, Lanzhou 730021,China; 2. Yellow River Basin Ecotope Integration of Industry and Education Research Institute, Lanzhou Resources amp; Environment Voc-Tech University, Lanzhou 730021, China)
Abstract: Aiming at the problem of function model solution with parameter constraint equation, a unified solution of constraint equation weighting is suggested, which combines the weighting value of constraint equation with the unified solution of observation equation. In order to verify the reliability and efficiency of this method, the orthogonal triangular decomposition formula and the generalized singular value decomposition formula suitable for the solution of the parameter constraint equation are deduced based on the matrix analysis method, and the three solution algorithms are compared and analyzed through examples. The results show that when a suitable weight value is given to the constraint equation, the parameter estimation values obtained through the unified solution of the constraint equation are consistent with the parameter estimation results of the matrix decomposition algorithm within the required accuracy range. The computational efficiency of the weighted unified solution is close to that of the orthogonal triangular decomposition method, and the computational efficiency of the generalized singular value decomposition method is low. The weighted unified solution of the constraint equation is more in line with the unified expression of the parameter model and the adjustment criterion, which is conducive to the design of the adjustment algorithm and the analysis and research of the error characteristics.
Key words: parameter estimation; constrain equation weighting; unified solution; orthogonal trigonometric decomposition method; generalized singular value decomposition method
0 "引言
測(cè)量平差問題大多屬于基于線性模型進(jìn)行參數(shù)估計(jì)的反演問題,即由現(xiàn)象推求機(jī)理[1]。反演過程有時(shí)因觀測(cè)量信息不足或者其他機(jī)理原因出現(xiàn)不適應(yīng)問題,例如秩虧或病態(tài),這是參數(shù)估計(jì)中要解決的問題。針對(duì)參數(shù)方程不適定問題,研究人員提出了一系列算法,例如預(yù)設(shè)基準(zhǔn)法[2]、虛擬觀測(cè)值法[3-6]、嶺估計(jì)法[7]、正則化解法[8-9]及其他顧及方差分量估計(jì)或隨機(jī)誤差的推估算法[10-14]等。這些算法都可以表示為測(cè)量平差中不適定問題解的統(tǒng)一表達(dá)法-選權(quán)擬合法[15]、帶準(zhǔn)則參數(shù)的平差準(zhǔn)則及其統(tǒng)一與解算[16-17]。依矩陣分析理論,參數(shù)估計(jì)值也可采用正交三角分解法[18-19](QR算法)和廣義奇異值分解法[20-21](GSVD算法)直接計(jì)算得到。對(duì)一些更復(fù)雜的非線性問題,有關(guān)學(xué)者探討了基于遺傳算法的廣義非線性最小二乘測(cè)量平差方案[22]。
無論是參數(shù)選權(quán)擬合法還是帶準(zhǔn)則參數(shù)的平差準(zhǔn)則,都是表示在數(shù)學(xué)函數(shù)模型中,除了含觀測(cè)值的觀測(cè)方程外,還含有參數(shù)約束方程。參數(shù)約束方程有時(shí)是數(shù)學(xué)物理模型中隱含的先驗(yàn)信息組成的參數(shù)間不等式方程或具有較為明確含義的參數(shù)間等式方程。上述不適定問題解算算法,適用于參數(shù)不等式約束方程,即弱約束方程。而等式參數(shù)約束方程是強(qiáng)約束,如果只是簡(jiǎn)單套用上述解算算法,一定程度上會(huì)弱化等式參數(shù)方程數(shù)學(xué)物理含義,降低參數(shù)估計(jì)的精度。
針對(duì)附參數(shù)約束等式方程的解算問題,參考借鑒上述不適定問題的解算方法和觀測(cè)方程加權(quán)最小二乘準(zhǔn)則,本文提出一種對(duì)約束方程賦權(quán)值的統(tǒng)一解算方法,簡(jiǎn)稱約束方程賦權(quán)統(tǒng)一解法(WULS算法)。主要思路是對(duì)等式約束參數(shù)方程施加強(qiáng)約束,即附加較大的權(quán)值,然后與觀測(cè)方程統(tǒng)一表達(dá)與解算,類似于虛擬觀測(cè)解算方法,不同的是虛擬觀測(cè)值的權(quán)值要遠(yuǎn)遠(yuǎn)大于實(shí)際觀測(cè)值的權(quán)值。為了驗(yàn)證該方法的可靠性和計(jì)算效率,本文基于參數(shù)等式約束方程和觀測(cè)方程的聯(lián)合方程式,應(yīng)用矩陣分析的方法,推導(dǎo)了QR算法和GSVD算法的參數(shù)估計(jì)表達(dá)公式,并設(shè)計(jì)數(shù)值算例,與約束方程賦權(quán)統(tǒng)一解法進(jìn)行比較分析研究。
等式約束方程的權(quán)值λ對(duì)于法矩陣條件數(shù)和參數(shù)方程解的影響(與GSVD解做差比較)見表2、圖1、圖2。由表2和圖1、圖2可見,當(dāng)λ取值范圍為1.0×104~1.0×1011時(shí),采用WULS算法計(jì)算的 與廣義奇異值分解法計(jì)算的 之差很小,而當(dāng)λ取值在1.0×104~1.0×1011之外時(shí),兩種算法均顯著地影響未知參數(shù)的估計(jì)精度。
3.2 "算例2
為了驗(yàn)證上述解法的計(jì)算效率,設(shè)計(jì)一個(gè)未知參數(shù)個(gè)數(shù)為10,觀測(cè)方程個(gè)數(shù)為70,且等式約束方程個(gè)數(shù)為70的GNSS載波相位觀測(cè)的數(shù)學(xué)函數(shù)估計(jì)模型。其中模糊度參數(shù)個(gè)數(shù)為7,位置參數(shù)為3。當(dāng)模糊度參數(shù)固定為整數(shù)值后,再附加這個(gè)模糊度整數(shù)參數(shù)值為函數(shù)等式約束方程,與原觀測(cè)方程統(tǒng)一解算。
同上,取λ為1.0×104,分別采用WULS算法和QR分解算法、GSVD分解算法計(jì)算各未知參數(shù),結(jié)果見表3。
為比較3種參數(shù)估計(jì)計(jì)算方法的計(jì)算效率,對(duì)2個(gè)算例分別采用3種算法循環(huán)計(jì)算1萬次,將計(jì)算時(shí)間進(jìn)行比較,結(jié)果見圖3和圖4。
由表2、圖3、圖4可見,采用QR分解算法求解未知參數(shù)的計(jì)算時(shí)間最短,效率最高,WULS算法次之,GSVD分解算法計(jì)算時(shí)間最長(zhǎng),效率最低。而參數(shù)計(jì)算結(jié)果顯示,GSVD分解算法和QR分解算法計(jì)算得到的估計(jì)參數(shù)值精度完全一致。WULS算法的計(jì)算精度則受λ的影響,但當(dāng)λ在合適的范圍內(nèi),也能完全滿足精度要求。
4 "結(jié)論
(1)參考虛擬觀測(cè)算法、正則化解法、不適定問題統(tǒng)一表達(dá)及解算等研究成果,提出了一種約束方程賦權(quán)統(tǒng)一解法(WULS算法),該算法先對(duì)約束方程賦權(quán)值,再對(duì)觀測(cè)方程形成統(tǒng)一表達(dá)式并求解。
(2)推導(dǎo)了解決此類數(shù)學(xué)模型的QR分解法和GSVD分解法的參數(shù)估計(jì)公式,并用數(shù)值算例的方法對(duì)三種算法進(jìn)行效果分析,驗(yàn)證WULS法的可靠性和計(jì)算效率。
(3)算例分析表明,當(dāng)λ在合適范圍內(nèi)時(shí),WULS算法的計(jì)算效率較高,參數(shù)估計(jì)精度能滿足精度要求。但WULS算法中如何確定λ的最優(yōu)值及其與高維方程解算效率的關(guān)系,還需后續(xù)深入分析研究。
參考文獻(xiàn)(References):
[1] WOLF P R,GHILANI C D.Adjustment Computations: Statistics and Least Squares in Surveying and GIS[M].3rd ed.New York:John Wiley amp; Sons,1997:375-376.
[2] 魯鐵定,張立亭,周世健,等.自由網(wǎng)平差的直接解算[J].西安科技大學(xué)學(xué)報(bào),2004,24(4):447-450,455.
LU Tieding,ZHANG Liting,ZHOU Shijian,et al.A direct algorithm of rank-deficient adjustment[J].Journal of Xi'an University of Science and Technology,2004,24(4):447-450, 455.
[3] 王樂洋,于冬冬.病態(tài)總體最小二乘問題的虛擬觀測(cè)解法[J].測(cè)繪學(xué)報(bào),2014,43(6):575-581.
WANG Leyang,YU Dongdong.Virtual observation method to ill-posed total least squares problem[J].Acta Geodaetica et Cartographica Sinica,2014,43(6):575-581.
[4] 余俊鵬.基于先驗(yàn)求權(quán)的虛擬觀測(cè)值法及其在線陣影像定向中的應(yīng)用[J].測(cè)繪科學(xué),2009,34(3):124-126.
YU Junpeng.The virtual observation method with weight prior-determined and its application in orientation of the linear array imagery[J].Science of Surveying and Mapping,2009,34(3):124-126.
[5] 趙海濤,郭廣禮,查劍鋒.附有限制條件間接平差的虛擬觀測(cè)算法[J].測(cè)繪科學(xué),2008,33(2):33-34,37.
ZHAO Haitao,GUO Guangli,ZHA Jianfeng.The virtual observation algorithm with restriction condition parameter adjustment[J].Science of Surveying and Mapping,2008,33(2):33-34,37.
[6] 馮光財(cái),戴吾蛟,朱建軍,等.基于虛擬觀測(cè)的病態(tài)問題解法[J].測(cè)繪科學(xué),2007,32(2):38-39,66,177.
FENG Guangcai,DAI Wujiao,ZHU Jianjun,et al.A virtual observational approach to ill-conditioned problem[J].Science of Surveying and Mapping,2007,32(2):38-39,66,177.
[7] 翁燁,邵德盛.病態(tài)加權(quán)總體最小二乘的廣義嶺估計(jì)解法[J].全球定位系統(tǒng),2021,46(6):84-89.
WENG Ye,SHAO Desheng.Generalized ridge estimation method for morbid weighted total least squares[J].GNSS World of China,2021,46(6):84-89.
[8] 沈云中,劉大杰.正則化解的單位權(quán)方差無偏估計(jì)公式[J].武漢大學(xué)學(xué)報(bào)(信息科學(xué)版),2002,27(6):604-606,610.
SHEN Yunzhong,LIU Dajie.Unbiased estimation formula of unit weight mean square error in regularization solution[J].Editoral Board of Geomatics and Information Science of Wuhan University,2002,27(6):604-
606,610.
[9] 毛云蕾,鄒時(shí)林,吳星.病態(tài)模型改進(jìn)正則化解的單位權(quán)中誤差無偏估計(jì)[J].大地測(cè)量與地球動(dòng)力學(xué),2020,40(4):411-416.
MAO Yunlei,ZOU Shilin,WU Xing.Truncated regularized solution to ill-posed model and its unbiased estimation of unit weighted variance[J].Journal of Geodesy and Geodynamics,2020,40(4):411-416.
[10] 孫海迅,羅健欣,潘志松,等.基于約束總體最小二乘的單應(yīng)性矩陣求解方法[J].計(jì)算機(jī)技術(shù)與發(fā)展,2022,32(12):50-56.
SUN Haixun,LUO Jianxin,PAN Zhisong,et al.A method for solving homography matrix based on constrained total least squares[J].Computer
Technology and Development,2022,32(12):50-56.
[11] 楊元喜,張菊清,張亮.基于方差分量估計(jì)的擬合推估及其在GIS誤差糾正的應(yīng)用[J].測(cè)繪學(xué)報(bào),2008,37(2):152-157.
YANG Yuanxi,ZHANG Juqing,ZHANG Liang.Variance component estimation based collocation and its application in GIS error fitting[J].Acta Geodaetica et Cartographica Sinica,2008,37(2):152-157.
[12] 姚宜斌,孔建.顧及設(shè)計(jì)矩陣隨機(jī)誤差的最小二乘組合新解法[J].武漢大學(xué)學(xué)報(bào)(信息科學(xué)版),2014,39(9):1028-1032.
YAO Yibin,KONG Jian.A new combined LS method considering random errors of design matrix[J].Geomatics and Information Science of Wuhan University,2014,39(9):1028-1032.
[13] 歐陽(yáng)文森,朱建軍.經(jīng)典平差模型的擴(kuò)展[J].測(cè)繪學(xué)報(bào),2009,38(1):12-15.
OUYANG Wensen,ZHU Jianjun.Expanding of classical surveying adjustment model[J].Acta Geodaetica et Cartographica Sinica,2009,38(1):
12-15.
[14] 國(guó)強(qiáng),劉雪萌,周凱.一種改進(jìn)粒子濾波算法實(shí)現(xiàn)的多徑參數(shù)估計(jì)[J].西安電子科技大學(xué)學(xué)報(bào),2022,49(3):120-128.
GUO Qiang,LIU Xuemeng,ZHOU Kai.Multipath parameter estimation realized by an improved particle filter algorithm[J].Journal of Xidian University,2022,49(3):120-128.
[15] 歐吉坤.測(cè)量平差中不適定問題解的統(tǒng)一表達(dá)與選權(quán)擬合法[J].測(cè)繪學(xué)報(bào),2004,33(4):283-288.
OU Jikun.Uniform expression of solutions of ill-posed problems in surveying adjustment and the fitting method by selection of the parameter weights[J].Acta Geodaetica et Cartographic Sinica,2004,33(4):283-288.
[16] 朱建軍,田玉淼,陶肖靜.帶準(zhǔn)則參數(shù)的平差準(zhǔn)則及其統(tǒng)一與解算[J].測(cè)繪學(xué)報(bào),2012,41(1):8-13.
ZHU Jianjun,TIAN Yumiao,TAO Xiaojing.United expression and solution of adjustment criteria with parameters[J].Acta Geodaetica et Cartographica Sinica,2012,41(1):8-13.
[17] 宋迎春,謝雪梅,陳曉林.不確定性平差模型的平差準(zhǔn)則與解算方法[J].測(cè)繪學(xué)報(bào),2015,44(2):135-141.
SONG Yingchun,XIE Xuemei,CHEN Xiaolin.Adjustment criterion and algorithm in adjustment model with uncertainty[J].Acta Geodaetica et Cartographica Sinica,2015,44(2):135-141.
[18] 劉麗麗,劉建華,左繼紅,等.基于最小二乘QR算法的質(zhì)心頻率下移法應(yīng)用研究[J].數(shù)學(xué)的實(shí)踐與認(rèn)識(shí),2019,49(5):217-224.
LIU Lili,LIU Jianhua,ZUO Jihong,et al.Study on application of centroid frequency down-shift method based on least-squares QR algorithm[J].Mathematics in Practice and Theory,2019,49(5):217-224.
[19] 趙學(xué)智,葉邦彥,陳統(tǒng)堅(jiān).大型矩陣奇異值分解的多次分割雙向收縮QR算法[J].華南理工大學(xué)學(xué)報(bào)(自然科學(xué)版),2010,38(1):1-8.
ZHAO Xuezhi,YE Bangyan,CHEN Tongjian.Multi-partition and double-direction shrink QR algorithm for singular value decomposition of large-scale matrix[J].Journal of South China University of Technology(Natural Science Edition),2010,38(1):1-8.
[20] 張顥月,師俊朋,史姝赟,等.基于GSVD的分布式MIMO雷達(dá)測(cè)向算法[J].信息對(duì)抗技術(shù),2024, 3(1): 57-69.
ZHANG Haoyue,SHI Junpeng,SHI Shuyun,et al.GSVD-based distributed MIMO radar direction finding algorithm[J]. Information Countermeasure Technology,2024,3(1):57-69.
[21] 劉圣恩,肖霖,楊鼎成,等.MIMO雙向中繼GSVD線性預(yù)編碼的功率優(yōu)化算法[J].西安電子科技大學(xué)學(xué)報(bào),2015,42(5):161-167.
LIU Sheng'en,XIAO Lin,YANG Dingcheng,et al.Power allocation under GSVD linear precoding in the MIMO two-way relay network[J].Journal of Xidian University,2015,42(5):161-167.
[22] 寧偉,張永常陳志榮,等.基于遺傳算法的廣義非線性最小二乘測(cè)量平差[J].遼寧工程技術(shù)大學(xué)學(xué)報(bào)(自然科學(xué)版),2006,25(S1):87-89.
NING Wei,ZHANG Yongchang,CHEN Zhirong,et al.Surveying and mapping adjustment of generalized least squares based on genetic algorithm[J].Journal of Liaoning Technical University (Natural Science),2006,25(S1):87-89.