第31卷第5期2006
年9月
测绘科学
Science of Surveying and M app ing
Vol 131No 15
Sep
作者简介:常晓涛(1972Ο
),男,副研究员,主要从事物理大地测量研究。E -mail:changtao@public 1bta 1net 1cn
收稿日期:2005Ο11Ο02
基金项目:国家自然科学基金(40274003,40174001);中国科学院动力大地测量学重点实验室基金项目
SGG 重力场球谐系数正则解的误差估计方法
常晓涛
①②
,章传银②,丁剑
①②
(①中国科学院动力大地测量学实验室,武汉 430077;②中国测绘科学研究院,北京 100039)
【摘 要】重力梯度为重力位的二阶导数,可以通过星载梯度仪进行观测。重力场球谐函数系数可以通过正则化
方法由重力梯度算出。本文在对正则化方法分析的基础上提出了估计球谐函数系数正则解误差的方法,为我国今后发射重力梯度卫星提供技术准备。【关键词】地球重力场;卫星重力梯度仪;重力场与稳态海洋环流探测卫星;正则化;误差估计【中图分类号】P22    【文献标识码】A     【文章编号】1009Ο2307(2006)05Ο0020Ο02
1 引言
即将于2006年发射的重力场与稳态海洋环流探测卫星(G OCE )的观测任务将持续20个月,其中包括两个为期分别为6个月的观测时段。卫星轨道高度为250Ο270k m,观测数据为由星载GPS 的高低星Ο星跟踪(SST )数据和重力梯度仪观测的卫星重力梯度(SGG )数据。这两种数据均可以各自独立地恢复地球重力场。本文仅针对利用SGG 数据恢复以球谐函数系数表示的地球重力场的算法进行研究。
SGG 数据为重力位的二阶导数,即以高频观测的重力梯度张量:XX 、YY 、ZZ 、XZ (X 轴指向卫星的瞬时速度方向,Y 轴指向瞬时轨道角动量方向,Z 轴垂直于X 、Y 轴,方向符合右手定则)。由卫星重力梯度恢复地球重力场的关键算法在于如何实现最优正则化。一般来说,正则化有两方面的作用,一方面,当我们要考虑的反演问题的解不适定时,即便是输入数据没有噪声,也必须进行正则化,否则将导致反演软件系统崩溃。输入数据没有噪声的情况下,正则化参数的选择对反演结果的影响不大,正则化参数的选择比较简单,即便最小的正则化参数也能保证反演软件的顺利运行。另一方面,正则化起到一个低通滤波的作用,它可以对反演模型中由于数据噪声造成的高频误差进行压缩。在数据含有噪声的情况下,即便反演问题的解是唯一的,正则化所起的作用及最优正则化参数的选择也都是十分重要的。正则化主要包括两方面的内容,首先是正则化方法即正则化函数或矩阵的选择,其次是正则化参数的选择。
虽然地球外部三维空间的重力场可以通过一系列的球谐函数展开进行方便的描述,但具体到实际应用,我们需要的往往是地球表面附近的重力场,因此对SGG 数据的处理还包括向下延拓的计算,而向下延拓是一个不稳定的操作,且SGG 数据还会存在彩噪声污染。
因此只有在特定的测量波段内进行SGG 观测,噪声功率谱密度的均方根才会足够小,否则噪声水平将迅速增强,会导致SGG 观测对球谐函数高阶系数的敏感性降低。
2 位系数的正则近似解
由卫星重力梯度反算地球重力场球谐函数系数将直接
导致第一类算子方程的求解,因此我们可以通过正则滤波
器来构造正则算子,对球谐函数系数正则解进行更精致的估计。尽管在实际计算中,求出第一类积分算子的奇异系统的计算量相当大,有时甚至是不可能的,但是,在广义算子逆和谱分析理论框架下,对该反问题的正则化进行分析,可以为构造新的重力场球谐函数系数恢复算法和对其解进行误差分析提供理论支持。
利用卫星重力梯度恢复重力场,SGG 与位系数呈线性
关系,即        g =A x
(1)其中g 为SGG 数据矢量,A 为设计矩阵,x 为重力位系数矢量。
设:   x α
=(A 3A +αI )-1A 3g
(2)       x αδ=(A 3A +αI )-1A 3
g δ
(3)  其中重力位系数x 的集合为X,即x ∈X,重力梯度g 的集合为G,即g ∈G,则A:X →G 为有界线性算子,A 3:
G →X 为A 的伴随算子,满足(A x,g )=(x,A 3
g );I 为X →X 的恒等算子,α为正则近似参数,δ为偏差。
则式(3)可以看作方程(1)的解A +g 的正则近似。根据正则化的原则,正则近似解的误差与解本身的光滑性等性质有关,同时也与正则算子的构造以及所选择的正则参数有关。在讨论正则近似解的误差前,我们应首先保证我们采用的算法为正则的,下面我们导出正则近似的条件。根据谱映照定理,
‖A 3A (A 3A +αI )-1‖≤1
‖(A 3A +αI )-1‖≤
1
α
  其中α为选定的参数,考虑到
‖x αδ-x α‖=‖(A 3A +αI )-1A 3
(g δ-g )‖≤
δα  所以有‖x αδ-A
+g ‖≤‖x α-A +
g ‖+δα上式中,我们称‖x α
-A +g ‖为正则误差,δα
为扰动误差。
对于参数α定义的算法,如果有
li m δ→0
α(δ)=0,li m δ→0x αδ=A
+
g   由于当g ∈D (A +
)时有‖x α(δ)
-A +
g ‖→0 (α(δ
)→0),根据上面的定义,条件:
li m
δ→0α(δ)=0,li m δ→0δ
2
α(δ)
=0
  将保证由上述参数α定义的Tikhonov 近似是正则的,既通过该正则化方法得到为重力场球谐系数的正则近似解。该条件也适用于算子存在扰动的情况,但本文仅限于对重力梯度存在扰动的情况的研究。
3 重力梯度存在扰动的误差估计
G OCE 卫星观测的SGG 数据为重力梯度张量,由重力梯度张量通过正则化方法计算重力位系数,其精度受观测数据与反演方法等多种因素的影响,起算数据重力梯度本身存在
扰动,为了方便对重力场球谐系数正则解的误差做出更精细的估计,我们定义渐进最优阶的概念,通过渐进最优阶对正则化算法的收敛速度进行分析,从而对算法的收敛性进行精细描述。下面我们对渐进最优阶进行推导,并在推导过程中,对正则化方法中正则化参数的选取进行分析。
假设M v,Δ={x =(A 3A )v z:‖z ‖≤
Δ},其中v >0,Δ≥0,z 为‖(A 3)-1
x ‖,R δ:G →X 惟一与δ>0有关的算子,我们称(R δ)δ>0关于v >0(和A )上渐进最优的。另外对于常数c v >0,如果对于任意δ>0,Δ>0,x 3∈M v,Δ及每一
个满足条件‖A x 3-g δ‖≤δ的g δ都有‖x 3
-R δg δ‖≤c v
正则化坐标(Δδ2v
)1
2v =1时,我们也称相应的算法或算子为渐进最优的。同时,对于待求的重力位球谐系数,如其属于M v,Δ,则有‖A +g -R δg δ‖=O (δ2v
2v +1),我们称正则近似x δ:=R δg δ
的误差具有渐进最优阶。
重力梯度反演重力位系数的算子A 为1-1的线性紧算子,该算子的值域R (A )于G 中稠,由于重力梯度为观测
值,存在偏差,假设‖g -g δ‖≤δ<‖g δ‖,x α(δ)
δ
是待求的重力场球谐系数正则解,且满足条件‖A α(δ)
x δ-g δ‖=δ对所有δ∈(0,δ0)成立。
x α(δ)δ
极小化Tikhonov 泛函M [α,δ,x ]:=‖A x -g δ‖2
+α(δ)‖x ‖2因此有:α(δ)‖x α(δ)δ‖2+δ2
=M [α,δ,x α(δ)δ
]≤M [α,δ,x ]=‖g -g δ‖2+α(δ)‖x ‖2≤δ2+α(δ)‖x ‖2
因此‖x α(δ)δ
‖2≤‖x ‖2
对所有的δ>0成立,从而可导出估计公式
‖x α(δ)δ-x ‖2=‖x α(δ)δ‖2-2Re (x α(δ)δ
,x )+‖x ‖2≤2[‖x ‖2
-Re (x α(δ)δ,x )]=2Re (x -x α(δ)δ
,x )令x =A 3z ∈A 3
(G ),z ≤Δ,且z ∈G,其中z =‖
(A 3)-1x ‖,则有‖x α(δ)δ-x ‖2
≤2Re (g -A x α(δ)δ
,z )≤2Re (g -g δ,z )+2Re (g δ-A x α(δ)
δ
,z )≤2δ‖z ‖+2δ‖z ‖=4δ‖z ‖≤4
δΔ于是有‖x α(δ)
δ
-x ‖≤2δΔ重力梯度反演重力位系数的算子A 为1-1的,所以R
(A 3G )于x 中稠,因此Πx ∈x ,ε>0,ϖx ~
=A 3z ∈A 3
(G )使得‖x ~
-x ‖≤ε
3
,于是可算得
‖x α(δ)δ
-x ‖2≤2Re (x -x α(δ)
δ,x -x ~
)+2Re (x -x α(δ)
δ
,
A 3
z )≤2‖x α(δ)δ-x ‖·ε3
+2Re (g -A x α(δ)δ
,z )≤23
‖x α(δ)
δ
-x ‖·ε+4δ‖z ‖  上式等价于
(‖x α(δ)
δ
-x ‖-ε3)2≤ε2
9
+4δ‖z ‖如果卫星重力梯度观测的误差δ=δ(ε)>0足够小,能
满足δ‖z ‖≤112
ε2,便有(‖x α(δ)
δ
-x ‖≤ε)由以上证明可以看出,对于Tikhonov 正则化方法,根
据Mor ozov 偏差原理确定正则化参数α,则有‖x αδ-A +
g ‖
=O (δ
),从而重力位球谐系数的正则近似解具有渐进最优阶O (δ
),位系数的理论精度仅取决于重力梯度观测的精度,且只要重力梯度观测误差足够小,重力位系数的精度就足够高。同时表明在z =‖(A 3)-1x ‖≤Δ条件下,Mor ozov 偏差原则是最优的正则化方法。
4 结论
对于Tikhonov 正则化方法,根据Mor oz ov 偏差原理确定正
则化参数α,则有‖x αδ-A +
g ‖=O (δ
),从而重力位球谐系数的正则近似解具有渐进最优阶O (δ
),位系数的理论精度仅取决于重力梯度观测的精度,且只要重力梯度观测误差足够小,重力位系数的精度就足够高。同时表明在z =‖(A 3)-1x ‖≤
Δ条件下,Mor oz ov 偏差原则是最优的正则化方法。参考文献
[1] Mor ozov V A 1On Regularizati on of III ΟPosed Pr oble m s
and Selecti on of Regularizati on Para meter [J ].J 1Comp 1M ath 1Phys 1,1966,6(1):170Ο1751
[2] Engl H W ,Hanke M ,Neubauer 1Regularizati on of I n 2
verse Pr oble m s [M ].Dordrecht Klu wer,19961
[3] Xiao T Y,L i C H 1Efficient Fast A lg orith ms f or s o me Non 2
linear ConstrainedM ini m izati on A rising fr o m L inear I
nverse Pr oble ms [A ].I n:Nu merical L inear A lgebra and Op ti m i 2zati on,Pr oceedings of the 2001I nternati onal Conference on Nu mericalOp ti m izati on and Nu mericalL inearAlgebra,Sci 2ence Press [C ].Beijing China /Ne w Y ork,20011
[4] Xiao T Y 1On the A lgorith m s for S olving the Siscrepancy
Equati on in Ill ΟPosed Pr oble m s [A ].I n:1996Sy mpo 2siu m on Computati on Physics f or Chinese Overseas and at Home [C ].Beijing,19961
[5] Haber E,A scher U M ,O ldenburg D 1On Op ti m izati on
technique f or s olving nonlinear inverse p r oble m s [J ].I nverse p r oble m s ,2000,16:1236Ο12801(上接第19页)
抚仙湖试验一共采用了5枚浮标,水下收发机投放深度为100m,工作环境为淡水湖面,测试了静态定位和动态定位工作模式(图10)。
图10 抚仙湖试验结果
抚仙湖定位结果显示静态工作模式下,100m 水深静态
定位单历元解算平面离散分布为2m,垂直方面为318m (图10左)。动态定位工作模式下,首先将一个GPS 浮标挂载船尾,然后将水下收发机投放一定深度后悬挂船尾,让船拖动GPS 浮标和水下收发机一起运动,通过实时观测GPS 浮标的位置和水下收发机的位置来判断动态定位的精度(图
10右),动态跟踪精度为3m 。
6 总结
差分GPS 水下立体定位导航系统集成了GPS 技术、声呐技术、无线电通讯技术、水声通讯技术、现代大地测量技术,实现了实时水下载体立体定位、跟踪、导航等功能。为水下工程施工、水下测绘、水下资源探查等应用提供了一种获取全球坐标系的精密位置信息新手段。
参考文献
[1] 陈永奇,李青岳1工程测量学[M ]1北京:测绘
出版社,19951[2] C A I Yan Οhui,BEI J in Οzhong,WANG Quan,etc 1Under
Οwater GPS Positi oning Syste m [A ]1I n:I nternati onal Sy m 2posiu m on GPS/G NSS 2005[C ],HongK ong,20051
[3] 蔡艳辉,章传银,王,等.差分GPS 水下立体
定位系统浮标姿态测量[J ]1测绘科学,2005,30(2)1
[4] 魏子卿,葛茂荣1GPS 相对定位的数学模型[M ]1
北京:测绘出版社,19981
1
2 第5期            常晓涛等 SGG 重力场球谐系数正则解的误差估计方法
The spectru m character isti cs of the rad i o frequency of Ch i n ese f i rst genera ti on s a tellite nav i ga ti on syste m—Be i dou I
  Abstract:Beidou I is Chinese satellite navigati on and positi oning syste m of first generati on1It covers territ ory and its surr ounding area1 The s pectru m characteristics of the radi o signal of“Beidou I”surveyed by foreign country are intr oduced in the paper1The constructi on and working frequency of"BeidouⅠ"satellite navigati on and positi on sys2 te m as well as positi on p recisi on of gr ound point t o be l ocated are dis2 cussed1A t the sa me ti m e,it gives the discussi on of the syste m’s work2 ing
p r ogra m,the frequency s pectru m of"Beidou"satellite as well as the feature map of“Beidou”satellite’s signal1
Key words:Beidou;navigati on;s pectru m
CHE N JunΟyong(State Bureau of Surveying&Mapp ing,Beijing 100830,China)
Da t a uncert a i n ti es i n geograph i c i n for ma ti on syste m
Abstract:Based on an analysis of the status quo and accomp lish2 ment of research on GI S data uncertainty,the main fra me and syste m of GI S data uncertainty is described p ri m arily1Then the core theories and the main issues of GI S data uncertainty,es pecially the positi on uncer2 tainty,the attribute uncertainty,the te mporal uncertainty,the s p read and manage ment of uncertainty,are discussed1A t length,the mathe mat2 ical methods of research on GI S data uncertainty are analyzed and con2 cluded in detail1
Key words:uncertainty;data uncertainty;GI S
WU Lun①,CHE NG J iΟcheng①,SH IW enΟzhong②(①I nstitute of RS and GI S,Peking University,Beijing100871,China;
②Depart m ent of Land Surveying&GeoΟI nfor matics,The Hong Kong Polytechnic University,Hong Kong)
Underwa ter positi on i n g syste m ba sed on D GPS
Abstract:Under water GPS positi oning syste m includes f our parts which are GPS differential reference stati on,GPS buoy subsyste m,un2 der water p inger,shi pΟbased data p r ocessing and contr ol center1GPS dif2 ferential stati on p r ovides realΟti m e differential correcti on data,and GPS buoy receives differential correcti on data t o calculate realΟti m e dyna m ic positi on and out put realΟti m e buoy’s s pace coordinate,at the sa me ti m e GPS buoy als o gets the hydr oΟphonic signal t o measure the distance t o the under water targets1The shi pΟbased data p r ocessing and contr ol cen2 ter is the most i m portant part of the syste m1A ll the data are trans m itted t o this center and the under water target’s s pace coordinates are res olved in this center1I n this paper,the syste m structure and each part’s func2 ti on are intr oduced firstly1Then,the ti m e measuring and positi oning al2 gorith m are devel oped1Some experi m ents and nu merical results are showed in the end1
Key words:under water positi oning syste m;DGPS;GPS buoy;hy2 perbola
WANG Quan①,CHE NG PengΟfei①,ZHANG ChuanΟyin①, WANG ZeΟm in②,CA I YanΟhui①,D I N
G J ian①(①Chinese Acade my of Surveying and Mapp ing,100039;②No1715Research I nstitute of Shi p2 building I ndustry Cor p1of China,310012)
Regul ar i za ti on error esti m a ti on of spher i ca l har m on i c coeff i c i en ts fro m SGG observa ti on
  Abstract:Gravity gradient is the second order derivative of gravita2 ti onal potential,which can be observed by satellite gradi ometer1The s pherical har monic coefficients of the gravity field can be obtained fr om gravity gradients thr ough regularizati on sche me1Based on the analysis of the regularizati on sche me,a method t o esti m ate the err or of regularizati on s oluti on is p resented in this paper1The results of this paper will p r ovide technical materials for Chinese future satellite gravity m issi on1 Key words:earth gravity field;satellite gravity gradi ometer;
G OCE;regularizati on;err or esti m ati on
CHANG XiaoΟtao①②,Z HANG ChuanΟyin②,D I N G jian①②(①La2 borat ory of Dyna m ical Geodesy,Chinese Academy of Sciences,W uhan,China430077;②Chinese acade my of surveying and mapp ing,Beijing 100039,China)
Study on the an tiΟsoluti on of l a titude functi on
Abstract:The ele ments of earthΟelli p s oid such as radiusM of cur2 vature in meridian are the functi on of latitude1According t o the defini2 ti on of functi on and antiΟfuncti on,the authors demonstrate the existence of antiΟs oluti on and deduce the antiΟs oluti on f or mula of ele mentary func2 ti on1They give the antiΟs oluti on computing p r ogra m of transcendental e2 quati on for C ASI O fxΟ4800p calculat or1The correctness of all antiΟs olu2 ti on p r ogra m have been p r oofed by exa mp le computati on。
Key words:functi on of latitude;ele mentary functi on;transcenden2 tal equati on;antiΟs oluti on p r ogra m
Z HONG YeΟxun①②,HUANG JunΟhua①③,L I ZhanΟyuan①②(①Guangxi Regi onal Bureau of Surveying and Mapp ing1Nanning,530023, China;②School of Res ource&Envir on ment,Guangxi Teachers’Col2 lege,Nanning530001,China;③Depart m ent of Civil Engineering, Guilin University of Technol ogy,Guilin541004,China)
Auto ma ted cycleΟsli p detecti on and correcti on for GPS tr i pleΟfre2 quency und i fferenced observables
  Abstract:I n the p r ocess of GPS tri p le frequency observables,the r m s of the integer ambiguity isn’t
fit for detecting and correcting cycle sli p s for the p seudoΟrange noise1The noise will be reduced with a si m p le running average filter or a p r ocess of combinati on for the ra w observa2 bles1It is recommended that selecting observables after a running aver2 age filter or combinati on as a detecti on observable can detect and correct cycle sli p s f or single frequency1Four p rinci p les are p r oposed f or con2 structing detecti on observables and three geometryΟfree linear combina2 ti ons are chosen according t o these p rinci p les1Cycle sli p s can be detec2 ted and corrected for tri p leΟfrequency undifferenced observables with these selected detecti on observables1
Key words:cycle sli p;a mbiguity res oluti on;detecti on observa2 bles;geometryΟfree phase combinati on;MW combinati on1
F AN J ianΟjun,WAN
G FeiΟxue,G UO GuiΟr ong(Satellite Naviga2 ti on R&D Center,School of Electri onic Science and Engingeering,Na2 ti onal University of Defence Technol ogy,Changsha Hunan China, 410073)
Robust Ka l m an f ilter i n g for the mar i n e sound i n g da t a
Abstract:After analyzing the pattern of s ounding,the models based on the algorith m s of Kal m an filte
ring were p resented in this paper t o filter the surveyΟboat positi ons and dep ths al ong a s ounding line1 However,the existing methods could not re move the alias s oundings in fact1So,a filtering method with a truncated weight functi on is designed t o remove outliers,such as gr oss err ors,false echo1The method has been tested using observed data1The results show that the r obust Kal m an filtering with the truncated functi on is really r obust in resisting the outli2 ers in s oundings1
Key words:marine s ounding;a s ounding line;r obust Kal m an fil2 tering
L IM ingΟsan①②③,L I U YanΟchun①②,BAO J ingΟyang①②,LV Zhi Οp ing③,X I A O FuΟm in①②(①Depart m ent of Hydr ography&Cart ogra2 phy,Dalian Naval Academy,Dalian116018,China;②the Geomatics and App licati ons Laborat ory,L iaoning Technical University,Fuxin 123000,China;③I nstitute of Surveying and M app ing,I nf or mati on En2 gineering University,Zhenzhou450052,China)
The prec isi on ana lysis of s a tellite stereophotogramm etry usi n g phys2 i ca l m ethod to deter m i n e exter i or or i en t a ti on elem en ts
  Abstract:This paper analyses the p recisi on of satellite stereophot o2 gra mmetry in the case of using physical method t o deter m ine exteri or ori2 entati on ele ments1It infers theoretical p recisi on calcula
ting for mula of satellite stereophot ogra mmetry and calculates the theoretical p recisi on of SP OT5HRS1Mean while it makes a test with few GCPs1The experi m ent
·
3
·
ABST RACTS OF THE PRESE NT I SS UE

版权声明:本站内容均来自互联网,仅供演示用,请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198,我们将在24小时内删除。