Scribd
Upload
18 views15 pages

Coord Transforms Using Collocation

Uploaded by

lubna
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
18 views15 pages

Coord Transforms Using Collocation

Uploaded by

lubna
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
USING LEAST SQUARES COLLOCATION RE. DEAKIN Department of Land Information Royal Melbourne Institute of Technology P.A. COLLIER AND F. J. LEAHY Department of Surveying and Land Information University of Melbourne 6 Mach i994 i | RANSFORMATION OF COORDINATES 1 ABSTRACT ‘This paper presents three methods of transforming World Geodetic System 1984 (WGS84) coordinates to Australian Geodetic Datum 1966 (AGD66) coordinates. Traditional methods of scaling, translating and rotating are compared with the least squares collocation technique of parameter estimation, filtering and interpolation revealing some advantages of collocation, INTRODUCTION ‘The US, Department of Defence Global Positioning System (GPS) is widely used for surveying and ‘mapping applications in Australia. GPS derived coordinates are related to the WGS84 Cartesian coordinate system (x, , 2yigg whose origin is at the Earth's centre of mass, The Zaxis isin the direction of the Conventional Terrestrial Pole (CTP), as defined by the Bureau International de 'Heure (BIH) on the basis of coordinates adopted for the BIH stations around the world, the X-axis passes through the intersection of the CTP’s equator and the zero meridian plane near Greenwich as defined by the BIH and the axis isin the plane of the equator 90° east of the axis. ‘The WGS84 Cartesian coordinate origin also serves as the geomettic centre of the WGSB4 elipsoid— whose parameters, with one minor exception, are those of the Geodetic Reference System 1980 (GRS80) ellipsoid (Moritz, 1980a and Decker, 1986) ‘The minor axis of the WGS8I ellipsoid is coincident with the Zaxis and the X-Zand X-¥ Cartesian planes are coincident with the zero meridian and equatorial planes ofthe ellipsoid respectively AGD66 geodetic coordinates (latitude longitude A height A) are related to the Australian National Spheroid (ANS) whose minor axis is parallel to the direction of the CTP as defined by the BIH and whose zero meridian plane is defined as being parallel to the BIH zero meridian plane near Greenwich. The ANS is The AUSTRALIAN Surveyor an integral part of the Australian Geodetic Datum (AGD Technical Manual, 1986) and in this paper, spheroid and ellipsoid both refer to a geometric surface formed by an ellipse rotated about its minor axis. AGD66 Cartesian coordinates (x,y, 2) have their origin at the centre of the ANS; with the Z-axis in the direction of the minor axis of the ANS, the X-axis passing through the intersection of the ANS equator and zero meridian plane and the Y-axis in the plane of the ANS equator 90° east of the Xaxis, Geodetic reference ellipsoids, such as the WGS84 and ANS, are computational surfaces which approximate the whole, or portions of, an irregular ‘equipotential surface known as the geoid, where the geoid can be defined as (DMA Technical Report 1983, p10), *... that surface to which the oceans would conform over the entire earth if free to adjust to the combined effect of the earth's mass attraction and the centrifugal force of the earths rotation’. Since the WCS84 ellipsoid is a global approximation of the geoid and the ANS is only an approximation of the geoid for the Australasian region, itis known that the origins of the two ellipsoids do not coincide and translations between the AGD and WGS84 origins are approximately 6, = +133m, 5, = 148m, 5, =-MA8m (DMA Technical Report, 1980), where x19 nes *8x and similarly for yand z Furthermore, its often regarded that the axes of the two Cartesian systems are not exactly parallel and that a scale factor exists between vectors in both systems. It is for these reasons that itis necessary to transform WCS84 coordinates, derived from GPS measurements, into coordinates related to the AGD, ‘Three transformation models will be investigated, (i) three parameter model involving translations only, (i) a seven parameter model involving three translations, three rotations and a scale factor and (ti) least squates collocation which combines parameter estimation, filtering and interpolation, Parameters for the first two models will be derived from sets of AGD66 and WGS84 coordinates for 16 points spread across Victoria using the traditional least squares ‘approach set out in Sections 3 and 4 respectively. The ‘method used in performing the transformation by least squares collocation is detailed in Section 8. Comparing residuals at the 16 data points from the three transformation models indicates that the collocation approach may offer some advantages in determining the best estimates of transformed coordinates. RELATIONSHIPS BETWEEN CARTESIAN. AND GEODETIC COORDINATES Figure 1 below, shows the well known relationships between Cartesian coordinates (x,y2) and Geodetic The AUSTRALIAN Surveyor coordinates (4, 4, of a point Prelated to an ellipsoid whose semi-major axis is OF = a and semi- minor axis is ON b, FIGURE 1 Referring to Figure 1, the Cartesian coordinates (:sy2) of a point A¢,,,4) on an ellipsoid of semi-major axis a and flattening j may be calculated by the following formulae: x= (v+h)coseosh a) y= (v+h)cososin (22) £5 (v(loe?) +h)sing (23) where Vo HP!= eee = radius of curvature (i=e¥sin*4) in the prime vertical plane, @=fQ-f) = eccentricity squared, b= ON =a(l-f) = semi-minor axis : of ellipsoid, OH = ve" sing ‘The inverse computation of (g,4,h) from (x,y,2) can be made using the following: cosh == (24) tang = 220 sine (25) ae v (26) where rave ey) March 994 7 A derivation of these classical formulae above, may be found in Torge (1980, pp.A7-52), Note: In equation (2.5), functions of the latitude appeat on both sides of the equation which requires an iterative solution for g. A first approximation for the latitude may be obtained from ang = 2, Convergence wil be rapid since «, THREE PARAMETER TRANSFORMATION MODEL Figure 2 shows the 3-parameter transformation model where the two parallel Cartesian coordinate systems; XYZ with origin O; and UVW with origin O;, are linked by the vector A whose components are the three translations 8, 5y 8 For 1 points common to both systems, the vector equation for point Pyis: ajebj+Asy, @) where a and b are position vectors, Aisa vector of translations and vis a vector of residuals. ‘The least squares estimate of the three parameters in A, if all 3n coordinate paits are considered to be of equal precision is Seo 8.) ) 2” ] [4 fh 2O- Lely, | 8, (3.2) where fue and Gano ate position vectors of the centroid in both systems. ‘The 3-parameter transformation model assumes that the XYZand UVW coordinate axes are parallel and no scale factor between vectors in both systems. This model will be used as the basis for systemati¢ trend removal in the collocation process discussed in later sections. SEVEN PARAMETER TRANSFORMATION MODEL Figure 3 shows the 7-parameter (Bursa-Wolf) transformation model (Krakiwsky and Thomson, 1974) FIGURE 2 FIGURE 3 ‘where the UVW system is scaled, translated and rotated with respect to the XYZ system. Small rotations (o,g, x) around the (U,1,¥7) axes respectively, are considered positive anti-clockwise ‘when viewed from the positive end of the axis looking towards the origin, The product of three consecutive rotations around the axes can be expressed in rotation matrices (Harvey, 1986) as: foorp 0 -singT1 00 0 1 0 [0 corm sao tine 0 core fo -snw cose (41) Since the rotations are always small, R can be approximated by Ral-x 1 @f-lo 1 ofsl« 9 of-reaR (4.2) cote Ina similar way to the 3-parameter transformation, but including a scale factor 4 and the rotation matrix R, the vector equation for P; for n points common to 8 Mach WOd The AUSTRALIAN Surveyor | both systems is a= AR, + A+ Wp (43) Letting the scale factor 4 = 1 + 84 and R=1+8R, equation (4.3) becomes apm (1+ 52)(+ ARIbj+ A+ (4a) and since 6) and aR are small and the product 61.8R = 0, equation (44) becomes y= 6Rby + 8b, + A+ by + Vy (45) Rearranging (45), each common point gives rise to an equation of the following form: vy] [0 -w vy 10 ofe) fx-w v|e]w 0 -w v 01 of ojaly-» & 8 8, 8, (48) For the m points common to both systems, equations (4.6) can be represented as conor, FIGURE 4 v+Ba=f (47) and the least squares estimate of the seven parameters in A, if all 3n coordinate pairs are considered to be of equal precision, is (Mikhail, 1976, section 7.3) A=(BTBY BY =[0 ox 528,5,5,]" (48) ‘The 7-parameter transformation model is the commonly accepted standard for transforming GPS derived WGS84 coordinates to AGD66 coordinates. THE TRANSFORMATION DATA. ‘The transformation data were derived from a high precision GPS network covering Victoria and New South Wales, jointly conducted by the Land Information Centre, Bathurst, NSW. and Survey & Mapping Victoria. The part of the network covering Victoria is shown in Figure 4 and consists of 33 stations connected by 146 GPS vectors. A 3 Dimensional unconstrained adjustment of the data relating to these 33 stations (using program GeoLah™) yielded WGS84 coordinates of 32 stations; Kosciusko being held fixed at known WGS84 values. Sixteen of these stations, indicated by a A on the diagram, also have known AGD66 values of latitude agg 20, some The AUSTRALIAN Surveyor Mach 1994 9 (8), longitude (2) and Australian Height Datum (AHD) heights (= height above the geoid). Ceoid-spheroid separations (A) for the WGS84 ellipsoid were computed at all stations in the network using the Ohio State University Spherical Harmonic Gravity Field Model OSU9LA (Rapp et al, 1991) and transformed to ANS N-values using the three parameter transformation mode! mentioned in Section 1 (DMA Technical Report, 1987), These values were then used to calculate spheroidal heights, hays = Hyp + Naat the 16 common stations. A summary of the relevant data is contained in Appendix A. THREE PARAMETER TRANSFORMATION RESULTS For the 16 common stations in the GPS network, the data given in Appendix A were converted to Cartesian coordinates using equations 2.1 to 2.3 and the translation vector 4 calculated from the position vectors of the centroid in both systems as given by equations (3.2). eta si8) [4432500 |_| [Link]|=| sarase -[Link]| [147.204 Residuals v at the common stations, calculated from, equation (6.1), are listed in Table 1. Shi» ; -arnsiosa 8-15, |= cou Rear =| 2897331022 s, 203912566, TABLET Residuals at common stations from 3-Parameter Transformation Model Name cn ee) Arthurs Seat 0.444 0.231 0128 Atkinson 0.362 0.190 0146 Bambadin (PM 3) 70520 ~1016 0.589 Bellarine (GPS Ecc) 0575 0,108 0.199 Benambra South Base) 03770538 0.208, Cann 02791239681 Chapple “0516-0403 00085 Gredgwin Silo (Bec A). 0.261.125. 0.487 Holey Hill 04850220279 Ka 05-0306 -0348 Jung 0084-0365 -0646 Kosclusko (Pillar) 157-0391 0.969 Matlock “0110 O86 Samaria 04050224 0.096 Talgarno 06900528028 Weejort 0.073 0375 ‘The WGS84 Cartesian coordinates of the other 17 stations in the network were transformed to AGD66 Cartesian coordinates according to equation (6.2) ¥ 2 8, vy) =p] +18, (62) Flav L# ware LS, and converted to Australian Map Grid (AMG66) east and north coordinates (AGD Technical Manual, 1986). These values are shown in Table 2 together with spheroidal heights of each station. TABLE2 AMG66 coordinates ond spheroidal heights of transformed stations (3-Parameter Transformation Model) Name “Zone East) North) im) Barham Reservoir 34 240373799 6053714397 108790 Brumby 38 602843182 5988323.494 421.695 Bullanginya 55 369005,090 6037359153, 168129 Cobbin (P) 55 42062282 5966103829 1273199 Eden Breakvater(P) 55 758401659 '5892746.459 10807 Euston Reservoir 5 659873584 6172243.460 83584 Lake Littra a 500133561 6245492.764 33092 Lianiduck (RM S) 54 672296474 6097026170 96728 Loka 8 505922898 6030858722 era729 Major (RM3 Brass) 55 382912768, 5974833898, 383.445 Mootong (P) 55 527619385 e1i4oais72 308793 MT Gambier (7022) 54 ATSALLEOL 53811703.065 193824 Thiele (SA) 54 490133395, 6206448631 sa7ld ‘Tower Hil (1862) 54 618740431 5757483.088 103278 Cobram (TS 72313) 8 377914.275 6024014079 150402 Wentworth Lock 54 583237532 6225091:830 35237 Yalta (S8M) 54 59259271 6225118823, 59137 10° Mach i994 The AUSTRALIAN Surveyor ‘SEVEN PARAMETER TRANSFORMATION RESULTS For the 16 common stations in the GPS network, the data given in Appendix A were converted to Cartesian coordinates using equations 2.1 to 2.3. The scale factor 1, elements of the rotation matrix R and the vector of translations A were computed using equation 4.8, comega(a) = 47811343 B+ radians phi(g) = -2461240 E-6 radians = -0.508 seconds| appa(e) = +2073098E-7 radians = +0,043 seconds scale(t) = 01999007194 (-281 ppm) = +0083 seconds atx) = +129728m a) = #57423m a) = 166014 m Re-ordering equation (4.5), residuals are calculated by (7.1) and are tabulated in Table 3. = 40.61 seconds ] ¥ ¥ ¥ 8, sly] -[y] arly] -aaly] 18, Flreou Lhe Leben Lebron 8s (7) TABLE 3 Residuals ot common stations from 7-Parometer “Transformation Mo ‘Name va) vy) vz) Arthurs Seat 0008-0199 017 Atkinson 0107-0083 0126 Bambadin (PM 3) 0376-0076 = (O372 Bellarine (GPS Ecc) 0162 0001-0222 Benambra (South Base) 0021 0018. -0.114 Cann 0024 IT +0043, Chapple 241-0068 0201 Gredgwin Silo (Bcc A) -0.019 510-0009 Holey Hill 0328-0167 +0169 Ida 0225 © 0178-0208, Jung 00490069 -0.027 Kosciusko Pilar) 0476-0250 0525 Matlock “014-0091 0108 Samaria 01420059 -0.009 ‘Talgamo 0006 =O -0.052 Weejort 0459-0114 -00033 ‘The WGS84 Cartesian coordinates of the other 17 stations in the network were transformed to AGD66 Cartesian coordinates according to equation (7.2) x ¥ 6, ly} =ARLy} 418, (72) Fhe — LElvarn L8y and converted to. AMG66 east and north coordinates. ‘These coordinates together with spheroidal heights of cach station are shown in Table 4, c, 0 ce: and, ° Gy, THE LEAST SQUARES COLLOCATION MODEL Mikhail (1976, p:394) describes collocation as “a general least squares technique combining classical adjustment with interpolation and filtering ..”, where interpolation is the estimation of quantities at locations where no observational data are given and filtering isthe estimation of these quantities taking into account the random measuring errors assumed to have occurred at the data points. Mikhail (1976) and Krakivosky (1975) both demonstrate that least squares parameter estimation (adjustment) and least squares interpolation and filtering are special cases of collocation and Moritz (1980b, p.132) provides ample mathematical proof that, as is the case with traditional least squares... collocation is optimal in the sense that it gives the most accurate results that are obtainable on the basis of the available data”. ‘To develop the collocation equations, consider equation (4.7) where the transformation model for n observations of w parameters can be represented as Vian + BonwAtecy = finn an and suppose that the residuals v are decomposed into a correlated signal component $ and a random noise component n. The noise in the mode! is simply the random measuring errors and the signal can be described as that component of the model which reflects the inability of the w selected parameters A to accurately describe the physical relationships. Furthermore, the signal component can be subdivided into signals at the n observation points t and signals at the m computation points u. These three random. vectors can be combined as oft] we oft] one “fi @n and so equation (4.7) becomes Meu [am Too Fao) tan |*Baodan fan (82) Av+Banf (83) Assuming that no correlation exists between signal and noise, the a-priori variance-covariance matrix associated with the random quantities in equation 63)is en Cy 0 — fe a | 0 0G, (8.4) The AUSTRALIAN Surveyor March 199411 TABLE AMG6é coordinates and spheroidal heights of transformed stations (7-Parameter Transformation Model) Name Zone East(m) North(m) ‘h{m) Barham Reservoir 4 240373,766 6053713.718, 108.592, Brumby 55 (602842.267 5988323631, 421.214 Bullanginya 55 369204.732 6037358,750, 167.826 Cobbin (P) 55 642061277 5966104.098 1272707 Eden Breakwater(P) 55 758400480, 5892747.118- 10297 Euston Reservoir 54 (659873,637 6172242,226 83,324 Lake Littra 5H 500133918 6245491080 32871 Tianiduck (@M3 S) 54 672296630 6097025168 96579 Loka 55, 505922175 6030858.581 674,287 Major (RM3 Brass) 55 382912484 5974833,694 383,231, Moorong (P) 55 527618631 6114041239 303.187 MT Gambier (7022) 54 478412920 5811702516 194.317 ‘Thiele (SA) 54 490134,354 6206446.984 52.568 ‘Tower Hill (1862) 5h 618741.355 5757482951 103.712 Cobram (TS 72313) 55 377913916 6024013.729 150.112, ‘Wentworth Lock 54 583237,698 6225090.310 34.967 ‘Yelta (SSM) 54 592592915 6223117.325 58.861 where solution, The underlying difference between the two Cay is the (mm) variance-covatiance matrix for the approaches is the incorporation of the signal term s signals at the computation points, and the components of its variance-covariance matrix Gq ‘isthe (on) vrtanc-evadanee mates for the in the solution equations. Infact, iti this matrix Souths chenetin poke Ce. which represents the central point in collocation Cup is the (m,n) covariance matrix between the signals, at the computation and observation points, and Cyn is the (o,n) variance-covariance matrix of the notse Applying the least squares principle to equation (83) in the manner of Mikhail (1976, pp19-420) leads to the best estimates for the parameters, signals and noise as qa = (B°D"BY BD) (85) Wing = CD Uf - BA) (86) tint) = CD" - BA) 87) na) = CanD"M - BA) (88) where Dian) = (Cir + Cn) (89) ‘The mathematical model described above (equations 8.1 to 8.4) and its solution (equations 85 to 88) is known as Collocation with Parameters and allows the simultaneous estimation of: (i) the vector of parameters 4; (i) the vector of signals u at the computation points, (known as interpolation ii) the vector of signals t at the observation points, (known as filtering); as well as (iv) the vector of random noise nat the observation points. Those familiar with the traditional least squares approach will notice the strong resemblance to the collocation model and its and allows quantities, which are normally linked by ‘mathematical relationships, to be described in a statistical manner, ie, by the use of variance- covariance matrices. In describing an application of collocation Cross (1992, p.142) uses the example of predicting the ‘unknown height of a point surrounded by a number of other points of known height.’The location of all points are known and itis assumed that a distance dependent function is known which is capable of computing the covariance of the heights of any two points in the region. This covariance function enables all the elements of a height variance-covariance ‘matrix to be computed which describes the variation of height in the region in a statistical manner. In flat areas, the heights of neighbouring points would be highly correlated (large covariances), but in highly undulating areas, points would be weakly correlated (small covariances). Solving the collocation equations, with the appropriate variance-covariance matrices, allows unknown heights of points within the region to be interpolated (or predicted). This approach contrasts with the usual method of fitting a surface to the known points and using the parameters of that surface to compute heights at other points. Determining the appropriate covariance function to compute the elements of Cis the central issue in collocation, 12 March 994 The AUSTRALIAN Surveyor THE COLLOCATION MODEL APPLIED TO. 3-D CARTESIAN COORDINATE TRANSFORMATIONS In the collocation model described above, the selection of the uw parameters in A depends on whether it is assumed the coordinate systems are related by translations only, or by scale, rotations and translations, In the former, A =[8,5y5,\? and u=3, whilst in the latter A = [0 9 « 82 8, 8y: Syd" and u=7. In the case of three parameters the coefficient matrix B, for each of the 1 points common to both systems will be equal to the identity matrix I and in the case of seven parameters By will take the form given in equation (4.6), In both cases, the vector f will contain humeric terms which are coordinate differences (5% Aaanes~ (% Zvess4 In this paper, for simplicity, the collocation model will assume A= (5,5, 84]? and w=3, THE COVARIANCE FUNCTION FOR THE MODEL ‘The covariance function, necessary for computing the elements of the signal vatiance-covariance matrices Cy and Cys must be determined empirically from the observational data and as a prerequisite, the systematic component of the data, modelled by Ba, must be removed. This is known as ¢rend removal and Ba is referred to as the trend surface, Inspection of equations (8.6 to 88) shows that u, t and n are determined from the observational data after trend removal, expressed in the equations as (f - BA), Since a collocation model with A = [5, 8,2!” and w=3is, assumed, then the residuals arising from the 3- parameter transformation (given in Table 1) are in fact the observational data with the trend removed and are represented by ¥ in equation (81), It is commonplace to assume that any correlation between these “observed” quantities I=[4, b, &... hl? is distance dependent and variances and covariances are calculated (Mikhail, 1976, p.405) as Go= xe (toa) Variance: Conarinces — Ci,)= PU, (102) Equation (10.1) shows thatthe vance computed by summing the squares of all data values fp and dividing by the number of data points 1. Equation (10.2) shows that a covariance may be computed from calculated for each distance class. Table 5 shows covariances (cm2) in x,y2 directions for particular distance classes and the number of data products. used to calculate each covariance. TABLES Spatial Distonce Class Covariances Dist Class x-eov yrov —zeov prods (km) (cm squared) 0.00 0.00) 0.00) 0 5-50 231787 2m79 NT 2 50-75 160557 43802-18679 1 75-100 12432 80160 96615 4 100-125 258268 156998 70776 5 125-150 788 157830 12362016 150-175-4584 26470 17494 8 ¥75-200 — -106290 53603 55658 6 200-225 119043 36077 69131 6 205-250-5706 M4396 673392 250-275 9.06 7655 (25188 6 275-300 -109210 -aa17— -40829 9 300-325 -1082.27 68721-62739 6 325-350 -103265 3837-25717 4 Narignces calculated from, the data are of= 2341.20 em, of= 2758.08 cn and 02 = 1849.36 cné, Figures 5,6 and 7 show graphs of covariance versus spatial distance (solid line) where the covariance is plotted at the mid-point of the particular data class. Superimposed over these graphs isa dotted line representing a distance dependent covariance function of a Gaussian form represented by the equation ld) = ce-oray ao3) where C(d) is the covariance between two points distance d apart, ¢ — isaconstant equal to the variance and, @ — isaconstant. ‘The constants ¢ and a for these three covariance functions were derived from a weighted least squares “best fit” solution of the positive covariances shown in ‘Table 5, where weights were assigned according to the number of data pairs used to calculate the covariance ‘The three covariance functions shown as dotted lines in Figures 5, 6 and 7 are the mg data pairs (or products) falling within a eld) =2a8e 6) io.) particular distance class interval dj. The sixteen (16) {a data points common to both systems give rise to G(d)=1792 is) (10.5) (n2~ n)/2 = 120 data products which can be grouped 4a) into spatial distance classes d and a covariance Ga) = 1047 e (0s) The AUSTRALIAN Surveyor March 99413 FIGURE 5, FIGURE 6 Y-conrancev Sena ost FIGURE 7 14 March 1994 The AUSTRALIAN Surveyor ‘These covariance functions were used to compute the Loka 0586 0209 elements of matrices Cy and Cyr. ‘Major (RM3 Brass) 0317 -0.093 ‘The noise min the model is assumed to be random Mga 70am) ha poe and furthermore itis assumed that no correlation Thiele SA) ane pa exists between signals s and noise n. Mikhail (1976, ‘Tower Hill (0862) choos “biog p.395-399) shows that the variances of the noise can Cobram (TS 72313) 0251 “ont be calculated from the relationship ‘Wentworth Lock 0012 “0207 C40) = C40) + C40) oz) Yetta (SSM) aly eval! where TABLE7 Cr0,Cy ave covariances of obseriations, signals _Sigals athe 16 observation commen) points and nolse respectively, Name SOSCSC«S)SSCSY 0) is the covariance between two points ~ distance d= 0 apart; or in other words, Arthurs Seat “0499 un 0080 thewmiaoce Atkinson 0424-0221 0155 . Bambadin (PM 3) 0.533 0.678 0448 Using this relationship, the elements of the Bellarine (GPS Ecc) 0.549 0231 0.096 diagonal noise variance matrix Cyp were calculated Benambra (South Base) 0376 0540 0404 from the observations and the covariance functions as Cann 0282 oT? aba 7 2 Chapple 0531-0312, -0083 oF = [2341-2438] = 97 om? ©, =£0.098m — Gredgwin Silo (Ecc A) 0212. -007T-—-049 of =2758~1792= 966 en? 0.311m Holey Hill Sek ee caer 2 - 2 - Jung 0.107 “0517 0516 Saleem ce 0, = 40.283 m Kosciusko (Pillar) 1066 = 0496. Matlock 0.131 0122 0,093 COLLOCATION RESULTS ame asst nine ones For the 16 common stations in the GPS network the Weejore toe 0AM O60 parameters A were computed using equation (85) as 8.) [-+132.622" TABLES me lstlslpe 7 ‘Noise n at the 16 observation (common) ella a Name ni) ni) 8.1 [147.208 Arthurs Seat 023 0130 The small ferences between these values and those AUS ie 3B obtained from the 3-parameter model, which assumes Balin (GPS Bec) 088 ae “measurements? of equal precision, are due to the fact Benambra South Base) 00ST “035 that the variance-covariance matrix D is included in Can 11005 oz the collocation solution. The matrix D reflects the Chapple ~0.017 0139 varying precision between the “measurements” in the Gredgwin Silo (Eee A). IT -0067 collocation model. Hey pos oan ‘The signals u and tand the noise n were computed 5 a using equations (8.6), (87) and (88) and are given in Haske lla nee Hea Tables 6, 7 and 8. Matlock -00uL 0.005 Samaria 0037 045 TABLE 6 Talgamno 0,000 -0136 ‘Signals u at the 17 computation points Weejort 0.019 053. Name a ) Barham Reservoir 0198 0.064 ~~ -0362 ‘The WGS84 Cartesian coordinates of the other 17 Brumby 1082 0482-443 stations in the network were transformed to AGD66 Bullanginya 0.206. 0169 0.134 Cartesian coordinates according to equation (11.1) Cobbin (P) 1082 0499 ¢ s 1.) (8, Eden Breakwater(P) 0315, 0.459 * * Euston Reservoir 0.043 ~0289 | sly] +f, fefe, (1) Lake Littra 0.016 0159 Lianiduck (RM3 S) 0.120 ~04l1 Fro L? lyase U4) LBs. The AUSTRAIIAN Surveyor Woah io4 15 and converted to AMG66 east and north coordinates, ‘These values are shown in Table 9 together with the spheroidal heights of each station. TABLE [AMG66 coordinates and spheroidal heights of transformed stations (Collocation Transformation Model) Name’ Zone Bast(m) Northim) i) Barham Reservoir 54 240373.618 6053714034 108.865 Brumby 55 602842.225 5988323.462 420840 Bullanginya 55 369204.813 66037359.004 168.111 Cobbin (P) 55 (642061256 5966103 858, 1272330 Eden Breakwater(P) 55 758400999 5892746.861 10503 Euston Reservoir 5A 659873564 6172243198 83.664 Lake Littra 54 500133641 (6245492607 33,094 Lianiduck (RM3 S) 54 672296449 6097025754 96805, Loka 55 505922291, 6030858716 674312 ‘Major (RM3 Brass) 55 382912456 5974832718 383317 ‘Mootong (P) 55 527619217 6114041.660 303564 MT Gambier (7022) 54 478411981 5811702257 193800 ‘Thiele (SA) 54 490134054 6206448422 52722 ‘Tower Hil (1862) 54 618740763 5757482845 103.238, Cobram (TS 72313) 55 377913971 6024013927 150344 Wentworth Lock 54 583237592 6225091638 35275 Yelta (SSM) 54 592592826 6223118632 5979 COMPARISON OF TRANSFORMATION RESULTS Table 10 shows differences between published ‘AMG66 and AHD values and the three transformation models; S-parameter, 7-parameter and collocation. ‘The reader may draw conclusions as to the relative worth of the three transformation models based on the differences in Table 10 (published-transformed) but.a statistical analysis of the residuals v for the 3- and 7-parameter models and the noise n in the collocation model may be useful in quantifying the precision of the three transformation methods. From Tables 1, 3 and 8, the standard deviations of the residuals v and the noise n are o.] [20.50 ‘20.24 40.03 o,|=|3054| —,|40.21) |] 4018) metres .} (20.44, pus (30-20), pera L427 ‘These standard deviations indicate that the collocation model is of comparable precision to the 7 parameter model and far better than the 3-parameter model. ‘The means and standard deviations of the differences in Table 10 are tabulated below ‘These values indicate that collocation is able to predict coordinates and heights to accuracies comparable with the 7-parameter transformation and generally better than the 3-parameter model. DISCUSSION ‘The theoretical foundations of collocation are described in texts such as Morita (1980b), Mikhail (1976) and Krakiwsky (1975) but little mention made of applications to coordinate transformations. ‘Moritz (1972, pp.5I-66) and Cross (1992, pp.150-151) describe how collocation can be applied to coordinate transformations and Ruffhead (1987) uses the technique to transform the grid coordinates of 15 Ordnance Survey Stations in England and Wales from the OSGB36 system to the OSGB70(SN) system. Rufthead describes his results as remarkable when compared to other two-dimensional transformation models but Vincenty (1987), who re-worked Mr Ruffhead’s data, comments that collocation is, complicated and that a 5th, order polynomial transformation gives better results. Vincenty, in his final comments on the plane transformations he investigated, notes that no consideration had been given to internal distortions existing within the data. ‘This is a crucial point and it should be emphasised that collocation does provide a means by which distortions can be modelled statistically, provided that sufficient data is available to determine the necessary covariance functions, Determining the correct oats026 16 Mach 1994 “The AUSTRALIAN Surveyor TABLE TO Differences belwoon published AMG66 and AHD values and transformation vols, Name Zone East(n) Northim) HIAHD) Brumby 58 602842432 5988923.738 a2 G-parameter) 0750 0244 007 (parameter) 0163 0107 055 (collocation) 0207 092 Bullanginya 56 369204524 16221 ~0.568 oat ~0.208 O51 0289 Yi 023 Cobbin (P) 55 642061369 5966104178 126239 ~0893 0349 “O21 092 0.080 028 -0.464 0.068 012 Eden Breakwater(P) 55 1758400535 5892746.929 520 “1124 ‘0470 “019 “0189 032 0.068 012 Lianiduck (RMS S) 54 61229871 66097025,081 9136 ~0697 =1089 057 ~0853 -0087 o72 0672 -0673 050 Loka 55 505921.984 6030858580 66786 “0914 “0142 ? “0191 00001 2 0307 -0316 ? Major (RMB Brass) 55 382912181 5974833706 37645 ~0587 “0192 008 0303 0012 029 0275 0012 020 Mootong (P) 55 527618889 114041.055 29772 01666 0517 a2 0258 1184 102 0328 605 065 ‘Tower Hill (1862) 5A 18740369 515T182.756 955 0.438 0332 03 0486 0195 01 0106 0.089 04 CCobram (TS 72313) 55 377913.656 6024014011 4434 0619 -0.068 024 01260 0282 053 0315 0084 030 Yelta (SSM) EN 595592332 6223117.652 2 (G-parameter) ~0439

You might also like