This International Standard specifies operations on SRF coordinates and, in the case of 3D object-spaces, on SRF spatial directions. Underlying these operations is the similarity transformation associated with two ORMs. Similarity transformations are treated first in 10.3. Then the general case of changing positional data from one SRF to another is specified in 10.4, followed by important special cases. The specification of a spatial direction in the context of an SRF is defined, and the general case of changing a SRF spatial direction to another SRF is specified (10.5).
Euclidean distance in 2D and 3D object-space is specified in 10.6. Distance and azimuth on the surface of an oblate ellipsoid (or sphere) are specified in 10.7. Vertical offset is defined in 9.3.
An important category of spatial operations is changing spatial information represented in one SRF to spatial information represented in a second SRF. For this category of operations, the adjective “source” shall be used to refer to the first SRF, and the adjective “target” shall be used to refer to the second SRF.
The notation in Table 10.1 is used throughout this clause.
Table 10.1 — Notation
Notation |
Description |
ORMS |
Source ORM |
ORMT |
Target ORM |
ORMR |
Reference ORM for a given spatial object |
HSR |
Reference transformation from ORMS to the reference ORMR |
HTR |
Reference transformation from ORMT to the reference ORMR |
HST |
Similarity transformation from the embedding of ORMS to ORMT |
SRFS |
Source SRF based on ORMS |
SRFT |
Target SRF based on ORMT |
SRFL |
The local tangent frame SRF at a coordinate (See 10.5.2) |
CSS |
CS of SRFS |
CST |
CS of SRFT |
|
Generating function of CSS |
|
Inverse generating function of CST |
|
Coordinate of a spatial position in SRFS |
|
Coordinate of a spatial position in SRFT |
|
Direction vector in SRFS (See 10.5.2) |
|
Direction vector in SRFT (See 10.5.2) |
The similarity transformations HST between source/target pairs ORMS and ORMT underlie the coordinate operations in 10.4. Given a set of n ORMs there are n(n-1) such source and target ORM pairs. Instead of specifying the full set of similarity transformations, this International Standard reduces the requirement to specifying the reference transformation HSR from each object-fixed source ORMS to the reference ORMR for a given object. This subclause specifies the methods of expressing a similarity transformation HST in terms of the reference transformations for the source and target ORMs. The cases of ORMs for a single object are treated in 10.3.2. The more general cases in which ORMS and ORMT are ORMs for different objects are treated in 10.3.3.
If ORMS is an object-fixed ORM, its reference transformation HSR may be specified as a seven-parameter transformation in the 3D case (see 7.3.2) and a by four-parameter transformation in the 2D case (see 7.3.3). The general form of HSR in the 3D case is given by Equation (7). The form in the 2D case is similar. As vector operations, they are in the form of a scaled invertible matrix multiplication followed by a vector addition. This form of vector operation is an invertible affine transformation. In the 3D case using the notation of Equation (5):
NOTE The processes by which ORMs for the Earth are established are based on physical measurements. These measurements are subject to error and therefore introduce various types of relative distortions between ORMs. Equation (7) is based on the assumption that positions in object-space are error free and the equation includes no compensation for these distortions.
The reference transformation HTR from ORMT to ORMR is similarly specified. An important operation is the similarity transformation HST from ORMS to ORMT, when neither the source nor the target is necessarily the reference ORM. The HST transformation may be expressed as the composition of with (the inverse of HTR) as in Equation (8) (see Figure 10.1):
The inverse operation is also an affine transformation:
Because the matrix is a rotation matrix, its transpose is also its inverse. Its inverse is also the matrix corresponding to the reverse rotations of ORMT with respect to ORMR. In particular:
and
.
The composite operation reduces to:
where:
If the rotation parameters are equal, then is the identity matrix, and if, HST simplifies to a translation of the origin:
.
Equation (8) and Figure 10.1 also apply to the 2D case.
If the source ORMS is a time-dependent ORM for a spatial object, ORMS(t) shall denote the ORMS at time t, and shall denote the similarity transformation from the embedding of ORMS(t) to the embedding of the object-fixed reference ORMR. If the similarity transformation can be determined, it is a time-dependent affine transformation. For a fixed value of time t0, Equation (8) and Figure 10.1 generalize to . The generalizations to a time-dependent target ORMT(t) are andfor the ORMS static and time-dependent cases, respectively.
EXAMPLE ORMS(t) is the ORM EARTH_INERTIAL_J2000r0 at time t. ORMR is the Earth reference ORM WGS_1984. Because ORMS(t) and ORMR share the same embedding origin, the transformation is a (rotation) matrix multiplication operation (without vector addition). The matrix coefficients for selected values of t account for polar motion, Earth rotation, nutation, and precession. Predicted values for these coefficients are computed and updated weekly by the International Earth Rotation Service (IERS) [IERS] (see 7.5.2). See Annex B for other examples of dynamic ORM reference transformations.
If a spatial object S exists in the space of another spatial object T, and if ORMR is the reference ORM for object T, and if the two objects are fixed with respect to each other, then HSR shall denote a similarity transformation from the embedding of ORMS to the embedding of ORMR. HSR is an affine transformation. If ORMT is an object-fixed ORM for the object T then HST is given by Equation (8). The time dependent generalizations of Equation (8), defined in 10.3.2, are also applicable to this case.
EXAMPLE ORMS is an ORM for the space shuttle (as a spatial object). ORMR is the Earth reference ORM WGS_1984. When in orbit at time t, transforms positions with respect to ORMS to positions with respect to ORM WGS_1984.
If the object-space of S and the object-space of T do not share locations or are otherwise unrelated, a similarity transformation between ORMs for the respective object-spaces is not defined. An abstract object S and a physical object T is an important instance of this case (see 10.4.6). However, if HSR is an invertible affine transformation between ORMS and the reference ORM for T, then, given an object-fixed ORM for object T, ORMT, Equation (8) may be used to define an invertible affine transformation HST, from ORMS to ORMT.
Given a coordinate in a source SRF, SRFS, and a target SRF, SRFT, the change coordinate SRF operation22 computes the corresponding coordinate in SRFT. The general case of changing the spatial coordinate of a location from SRFS to SRFT is presented in formulations in 10.4.2 for time-independent (static) and time-dependent ORM relationships. The general case assumes that the source coordinate corresponds to a location that exists in both the source and target object spaces.
In the general case, ORMS and ORMT may differ, and the coordinate systems, CSS and CST, may differ. The formulation simplifies in the special case23 for which ORMS = ORMT or, more generally, in the case for which the associated normal embeddings match. This case is presented in 10.4.3. In a further specialization of the ORMS = ORMT case, it is assumed that CSS and CST are geodetic and/or map projection CSs. These assumptions produce further simplifications (see 10.4.4).
The case for which CSS = CST and ORMS and ORMT differ24 does not generally produce a computational simplification of the general case. However, when both the source and target SRFs are based on the CS LOCOCENTRIC_EUCLIDEAN_3D, a simplification is produced and is presented in 10.4.5. This case is important for change of direction operations (10.5.4).
An extension of the change SRF operation to the case of unrelated source and target object-spaces is presented in 10.4.6 for linear SRFs. In that case, the ORM transformation is only restricted to an invertible affine transformation.
SRFS and SRFT are two object-fixed SRFs for a spatial object and p is a point in object-space that is in the coordinate system domains for both SRFs. denotes the coordinate of p in SRFS, and denotes the coordinate of p in SRFT. The determination of as a function of is an operation on the SRF pair (SRFS, SRFT). The most general form of the operation is:
where:
See Figure 10.2. CS generating and inverse generation functions are specified in Clause 5.
Figure 10.2 — Change coordinate SRF operation
Equation (10) is known as the Helmert transformation when HST is approximated with the Bursa-Wolfe equation (see Annex B).
In the time-dependent case, Equation (10) may be generalized to:
.
EXAMPLE 1 If SRFS and SRFT are two celestiodetic SRFs for the same spatial object with different ellipsoid RDs, Equation (10) transforms the coordinate with respect to one oblate ellipsoid to with respect to the other oblate ellipsoid.
NOTE A transformation between two celestiodetic SRFs for the spatial object Earth is known as a horizontal datum shift. A number of numerical approximations developed to implement this operation have been published. Under the assumption of zero rotations and no scale differences (), a widely used approximation25 to directly transform to , is the standard Molodensky transformation formula [83502T] as follows:
where:
The quantities are defined in Table 5.6.
Equation (10) is only defined for a value of in the CSS domain if its corresponding position belongs to the CST range. If is the domain of the inverse generating function and is the domain of the inverse generating function, Equation (10) is only defined for in the set:
EXAMPLE 2 SRFS is SRF GEOCENTRIC_WGS_1984 and SRFT is an instance of SRFT MERCATOR, with ORM WGS_1984. Equation (10) is not defined for any that is on the z-axis of SRFS, because the z-axis is not contained in the set in Equation (11).
SRFT may optionally specify a valid-region and may optionally specify an extended-valid region (see 8.3.2.4). If is the domain of the generating function, then . If Equation (10) is defined for, may be valid, or extended valid or neither. The set of coordinates for which is valid is:
where:
.
In applications that functionally conform to an SRM profile, the domain of an SRF operation is restricted to the accuracy domain of the SRF as specified by that profile (see Clause 12).
If both ORMs are the same23 , or, more generally, if the corresponding parameters of the seven-parameter reference transformations of ORMS and ORMT match, is the identity transformation. Consequently, Equation (10) simplifies to:
|
. |
EXAMPLE 1 If SRFS is a celestiodetic SRF (see 8.4) and SRFT is the celestiocentric SRF for the same ORM (ORMS = ORMT), then is the identity and Equation (12) reduces to the geodetic generating function: .
EXAMPLE 2 If SRFS is an induced surface celestiodetic SRF (see 8.4) and SRFT is the 3D celestiodetic SRF for the same ORM (ORMS = ORMT), Equation (12) changes from a coordinate of CS type surface to a coordinate of CS type 3D.
If SRFT is a 3D SRF that has ellipsoidal height designated as the vertical coordinate-component of the SRF (see 8.4), and SRFS is the induced zero height surface SRF, the promotion operation converts a surface coordinate in SRFS to a 3D coordinate in SRFT by setting the 1st and 2nd coordinate-components of to the 1st and 2nd coordinate-components of and setting the 3rd coordinate-component, ellipsoidal height, to 0. Coordinate promotion is a special case of Equation (12). Applicable SRFs include those based on SRFT CELESTIODETIC, PLANETODETIC, and all map projection SRFTs
EXAMPLE 3 Reversing the roles of source and target SRFs in Example 2, if SRFS is a celestiodetic 3D SRF and SRFT is the (induced) surface celestiodetic SRF for the same ORM, Equation (12) is not defined for, unless. Equivalently, only coordinates of the form belong to the set in Equation (11). Coordinates in SRFS that are not on the oblate ellipsoid (or sphere) RD instance surface, can be projected to the surface along a coordinate curve by setting.
If SRFS is a 3D SRF that has ellipsoidal height designated as the vertical coordinate-component of the SRF (see 8.4), and SRFT is the induced zero height surface SRF, the truncation operation converts a 3D coordinate in SRFS to a surface coordinate , by setting the 1st and 2nd coordinate-components of to the 1st and 2nd coordinate-components of. The point in object-space corresponding to and the point in object-space corresponding to are not the same point unless. Truncation, therefore, does not generally preserve location.
The CS generating function for a map projection SRF (or, respectively, an augmented map projection SRF) is implicitly defined (see 5.8.2 or, respectively, 5.8.6) by the composition of the generating function for the surface geodetic CS (respectively, the geodetic 3D CS) with the inverse mapping equations (respectively,) as:
.
If SRFS and SRFT are map projection SRFs for the same object, and the corresponding seven parameters of their reference transformations match, then Equation (12) becomes:
for:
Furthermore, if ORMS = ORMT, then and Equation (13) simplifies to:
|
. |
NOTE If SRFS is a map projection SRF, and SRFT is the corresponding augmented map projection SRF based on the same ORM, then Equation (14) is equivalent to the promotion operation (see 10.4.3).
If SRFT is a celestiodetic SRF and ORMT = ORMS, Equation (13) simplifies to:
.
Similarly, if SRFS is a celestiodetic SRF and ORMT = ORMS, Equation (13) simplifies to:
.
The special case of source and target SRFs based on the CS LOCOCENTRIC_EUCLIDEAN_3D is important for the treatment of directions (see 10.5). Every linear orthonormal CS may be viewed as an instance of a CS LOCOCENTRIC_EUCLIDEAN_3D. If SRFS and SRFT are two SRFT LOCOCENTRIC_EUCLIDEAN_3D based SRFs (see Table 8.11), then the SRF pair operation on is determined by substituting the CS LOCOCENTRIC_EUCLIDEAN_3D (see Table 5.9) generating function and its inverse in Equation (10). If vectors are the CS binding parameters for the SRFT LOCOCENTRIC_EUCLIDEAN_3D based SRF, may be expressed in the form of the affine transformation:
where:
The inverse generating function is expressed as:
where:.
If vectors are the CS binding parameters for SRFS and SRFT respectively (see Table 8.11), then substituting the expression in Equation (9) for HST, Equation (10) specializes to:
In the case that the corresponding seven parameters of the reference transformations of ORMS and ORMT match, Equation (12) specializes to Equation (16):
|
. |
Engineering designs and other abstract models are often intended for realization in the physical world.
EXAMPLE A building plan is designed in the source SRFS, an abstract space LOCAL_SPACE_RECTANGULAR_3D SRF. A terrestrial site survey establishes the origin of the target SRTT, a LOCAL_TANGENT_SPACE_EUCLIDEAN SRF. Source coordinates are identified to target coordinates by: where is a scale factor.
More generally, abstract models are scaled, rotated, or otherwise transformed by an invertible matrix W before a source coordinate is identified to a target coordinate. This identification may be viewed as a change coordinate SRF operation from in SRFS, an abstract space LOCAL_SPACE_RECTANGULAR_3D SRF, to a coordinate in SRFT, a physical world LOCOCENTRIC_EUCLIDEAN_3D SRF. In the notation of 10.4.5:
Define an invertible affine transformation as (see 10.3.3). Substitute this in Equation (10) and simplify:
This illustrates that the identification may be viewed as a change coordinate SRF operation.
NOTE Equation (17) illustrates that digital graphic composite pattern modelling techniques such as SceneGraph trees that use scale and rotation matrices W together with translation operations at each tree node are special cases of Equation (10). See also 10.5.4 Example 2.
In 3D position-space, a direction is unambiguously specified by a normalized vector. The direction specified is translation independent. This is illustrated by lines through points in a given direction n (see A.7.1.1 Example 1). All such lines are parallel. This translation invariance carries over to the coordinate-space of a linear CS, but not to other CSs with vector space structure. In particular, an augmented map projection inherits the vector space structure of 3D Euclidean coordinate-space, but the “up pointing” vector n = (0, 0, 1) points in different spatial directions (in position-space) depending on the map coordinate location from which n is viewed.
Figure 10.3 — Coordinate-space and position-space directions compared
In Figure 10.3, distinct position points p and q on the ellipsoid surface are projected to augmented map coordinates (s, t, 0) and (u, v, 0). Starting at these map coordinates, the coordinates one unit away in direction n are (s, t, 1) and (u, v, 1) respectively. In an augmented map projection, these coordinates correspond to the position-space points p’ and q’. The direction from p’ to p is not the same as the direction from q’ to q. It is noted in 5.8.6.2 that augmented map projections are not vertically conformal, therefore angular relationships of spatial directions are generally not preserved by augmented map projections.
A linear CS will not preserve angular relationships between directions unless the CS is also orthonormal. In an SRF based on a linear orthonormal CS, the translation invariant vector space structure of the abstract CS carries over to the spatial CS because the underlying normal embedding preserves angles and distances.
The coordinate-space of a curvilinear CS does not have a linear vector space structure so there is no natural way to specify a translation invariant direction with curvilinear coordinates. An SRF based on a curvilinear CS requires a uniform method associating a unique linear orthonormal CS based SRF to each coordinate in the interior of the CS domain. This association is defined in 10.5.2.
In this International Standard, a direction in a 3D orthogonal26 SRFS is expressed as a combination of a normalized vector and a reference coordinate. The normalized vector is in a 3D linear orthonormal SRF, denoted by SRFL. If SRFS is curvilinear, SRFL is uniquely defined for each reference coordinate using the normalized tangent vectors to the coordinate-component curves at the reference coordinate. These vectors are used as SRF parameters for SRFT LOCOCENTRIC_EUCLIDEAN_3D with ORMS to specify SRFL. SRFL is termed the local tangent frame at the reference coordinate.
The same definition is applicable if SRFS is linear. In the linear case, SRFL at reference coordinate (0, 0, 0) coincides with SRFS as a spatial CS. Also in the linear case, the normalized vector representing the direction is independent of the reference coordinate used. The linear case includes SRFs that are based on SRFTs CELESTIOCENTRIC, LOCAL_TANGENT_SPACE_EUCLIDEAN, LOCOCENTRIC_EUCLIDEAN_3D, and LOCAL_SPACE_RECTANGULAR_3D.
Given a coordinate c = (u0, v0, w0) in the interior of the domain of a 3D orthogonal SRFS, the local tangent frame at coordinate c, SRFL, is the SRF specified by the SRFT LOCOCENTRIC_EUCLIDEAN_3D with ORMS and the following SRF parameters:
where:
The vectors r and s are termed the local tangent vectors at c. Coordinate-component curves are defined in 5.5.3.
NOTE 1 The tangent vector to the 3rd coordinate-curve at (u0, v0, w0) points in the same direction as the vector because of the coordinate-component ordering restriction specified in 5.6.4.
A direction in an orthogonal CS based SRFS shall be comprised of:
a) a coordinate c in the interior of the CS domain of SRFS, and
b) a normalized vector n in the local tangent frame at c.
The coordinate c is termed the reference coordinate of the direction and its corresponding position is termed the reference position for the direction. The vector n is termed the direction vector at c.
NOTE 2 The local tangent frame at a coordinate is an instance of the SRFT LOCOCENTRIC_EUCLIDEAN_3D that provides a vector space setting for vector operations on direction vectors at c.
EXAMPLE 1 If SRFS is a LOCOCENTRIC_EUCLIDEAN_3D SRF with SRF parameters q, r and s, and c is an SRFS reference coordinate, then local tangent vectors at c are equal to the SRF parameters r and s. If c = (0,0,0), then SRFL = SRFS.
EXAMPLE 2 SRFS is an EQUATORIAL_INERTIAL SRF. This SRF is based on the SPHERICAL CS. If is a reference coordinate, then the local tangent vectors at c are:
EXAMPLE 3 SRFS is a CELESTIODETIC SRF. This SRF is based on the GEODETIC CS. If is a reference coordinate, then the local tangent vectors at c are:
The vector
In this example, SRFL is an LOCAL_TANGENT_SPACE_EUCLIDEAN SRF with SRF parameters .
EXAMPLE 4 SRFS is based on a conformal map projection CS. If is a reference coordinate, and is the corresponding celestiodetic coordinate, then the local tangent vectors at c are:
In this example, SRFL is an LOCAL_TANGENT_SPACE_EUCLIDEAN SRF with SRF parameters .
Given a direction represented with direction vector n1 at c1, the same direction may be represented at another reference coordinate c2 in the same SRF, with direction vector n2. The direction vector n2 is computed as:
The local tangent vectors are computed as in Equation (18). Equation (19) is derived from Equation (16) by dropping the translation term since directions are translation invariant.
If the SRF is based on a linear CS, then the matrix in Equation (19) is the identity matrix and n1 = n2. This implies that in an SRF based on a linear orthonormal CS, a direction vector is independent of the reference coordinate. Thus, Equation (19) is only of interest in the case of a curvilinear SRF.
Given a direction represented with direction vector nS at cS in SRFS, the same direction may be represented at reference coordinate cT, with direction vector nT in SRFT. If HST is the similarity transformation from ORMS to ORMT and TST is the matrix in the last term in Equation (15), then the direction vector nT is computed as:
Equation (20), is derived from Equation (15) by dropping the translation term since directions are translation invariant and dropping the scale factor since nT is a normalized vector.
EXAMPLE 1 SRFS is SRF GEODETIC_WGS_1984 and SRFT is SRF GEOCENTRIC_WGS_1984. With SRFS reference coordinate , the Washington monument, an obelisk, points approximately in the direction at cS. In this example, ORMS = ORMT so that TST is the identity matrix, and because SRFT is based on SRFT CELESTIOCENTRIC, is also the identity matrix. Consequently Equation (20) reduces to:
Then using the expression in 10.5.2 Example 3 for t:
So that the direction vector in SRF GEOCENTRIC_WGS_1984 is.
The case of changing an abstract space linear SRF direction vector nS to a direction vector nT in a linear SRF in the space of another object is based on Equation (17). In the notation of 10.4.6:
|
. |
Since a direction vector is normalized, division by the determinant cancels any scaling by matrix W. RS is a rotation matrix and therefore its determinant is 1.
EXAMPLE 2 In ISO/IEC FDIS 18023-1:2005, if an instance of the class <DRM Geometry Model Instance> has a component of class <DRM World Transformation>, that component specifies an invertible matrix W and a coordinate c in the <DRM Environment Root> SRF. If cS and nS are a reference coordinate and a direction vector in an associated LOCAL_SPACE_RECTANGULAR_3D <DRM Geometry Model>, and SRFT is the local tangent frame at c, then Equation (17) and Equation (21) may be used to compute cT and nT, respectively. The methods of 10.4.3 may be used to further change cT from SRFT to the <DRM Environment Root> SRF. This procedure to change <DRM Geometry Model> coordinates and directions to the environment root SRF is termed "model instancing".
This International Standard supports an operation to return the Euclidean distance between two object-space locations using the coordinates of those locations in an SRF.
If c1 and c2 are two coordinates in an SRF, and if G is the generating function of the CS of the SRF, the Euclidean distance dE between the corresponding points in object-space is given by:
where d is the Euclidean metric.
This International Standard supports the geodesic distance and azimuth operations for SRFs that have ellipsoidal height designated as the vertical coordinate-component (see 8.4). These SRFs include those based on SRFT CELESTIODETIC, PLANETODETIC, and all map projection SRFTs.
The zero vertical coordinate-component surface for such an SRF is an oblate ellipsoid. Two distinct points on the surface of the oblate ellipsoid are connected by a surface curve called a geodesic as defined in A.7.3. The distance along the curve between the two points is called the geodesic distance. At each point, the angle between the geodesic and the meridian at the point as defined in 5.8.3.4 is the azimuth at the point with respect to the other point. The operations to return the geodesic distance and azimuths given the surface coordinates of the points are supported in the API.
For an oblate ellipsoid, a geodesic does not, in general, lie completely in any single plane [RAPP1] [RAPP2]. If ( λ1, φ1) and ( λ2, φ2) are the surface celestiodetic coordinates of two points lying on the oblate ellipsoid surface with parameters a and e, and, the geodesic distance between the points [PEAR] is given by:
In the general case, two surface coordinates c1 and c2 are converted to celestiodetic coordinates using the operations defined in 10.4.4. In particular, in the case of a map projection SRF, if Q is inverse mapping equations for the SRF, then:
NOTE This is an elliptic integral and the development of approximation equations for has been the subject of much research. There are approximation formulas for the short distance case where ≤ 200 km, for the medium distance case where ≤ 1000 km and for the long lines case where the points are antipodal or near antipodal. Two points on the oblate ellipsoid are exactly antipodal when |(λ2 - λ1)| = π and φ1 = -φ2. There are also special cases when. A thorough exposition of geodesic distance approximations is given in [RAPP1] [RAPP2].
Geodetic azimuth is defined in 5.8.3.4. On a sphere, a geodesic between two points is an arc of a great circle and the problem of computing the angles of a spherical triangle can be solved in closed form. In the general case of an oblate ellipsoid, the problem of computing the angles of an elliptical triangle does not have a closed solution. Several different approximations are commonly used.
NOTE Some algorithms are designed to compute both the geodesic distance and the azimuths associated with two points.
25 Historically it was thought that these approximations would require less computation than direct conversion. The perceived computational advantage may have been overcome by technology advances. New efficient algorithms for converting celestiocentric coordinates to celestiodetic coordinates have been developed that result in appreciably faster transformations without the attendant loss of accuracy.
26 All of the 3D SRFTs in this International Standard are based on orthogonal CSs.