GeographicLib 2.1.2
TransverseMercatorExact.cpp
Go to the documentation of this file.
1/**
2 * \file TransverseMercatorExact.cpp
3 * \brief Implementation for GeographicLib::TransverseMercatorExact class
4 *
5 * Copyright (c) Charles Karney (2008-2022) <charles@karney.com> and licensed
6 * under the MIT/X11 License. For more information, see
7 * https://geographiclib.sourceforge.io/
8 *
9 * The relevant section of Lee's paper is part V, pp 67--101,
10 * <a href="https://doi.org/10.3138/X687-1574-4325-WM62">Conformal
11 * Projections Based On Jacobian Elliptic Functions</a>;
12 * <a href="https://archive.org/details/conformalproject0000leel/page/92">
13 * borrow from archive.org</a>.
14 *
15 * The method entails using the Thompson Transverse Mercator as an
16 * intermediate projection. The projections from the intermediate
17 * coordinates to [\e phi, \e lam] and [\e x, \e y] are given by elliptic
18 * functions. The inverse of these projections are found by Newton's method
19 * with a suitable starting guess.
20 *
21 * This implementation and notation closely follows Lee, with the following
22 * exceptions:
23 * <center><table>
24 * <tr><th>Lee <th>here <th>Description
25 * <tr><td>x/a <td>xi <td>Northing (unit Earth)
26 * <tr><td>y/a <td>eta <td>Easting (unit Earth)
27 * <tr><td>s/a <td>sigma <td>xi + i * eta
28 * <tr><td>y <td>x <td>Easting
29 * <tr><td>x <td>y <td>Northing
30 * <tr><td>k <td>e <td>eccentricity
31 * <tr><td>k^2 <td>mu <td>elliptic function parameter
32 * <tr><td>k'^2 <td>mv <td>elliptic function complementary parameter
33 * <tr><td>m <td>k <td>scale
34 * <tr><td>zeta <td>zeta <td>complex longitude = Mercator = chi in paper
35 * <tr><td>s <td>sigma <td>complex GK = zeta in paper
36 * </table></center>
37 *
38 * Minor alterations have been made in some of Lee's expressions in an
39 * attempt to control round-off. For example atanh(sin(phi)) is replaced by
40 * asinh(tan(phi)) which maintains accuracy near phi = pi/2. Such changes
41 * are noted in the code.
42 **********************************************************************/
43
45
46#if defined(_MSC_VER)
47// Squelch warnings about constant conditional and enum-float expressions
48# pragma warning (disable: 4127 5055)
49#endif
50
51namespace GeographicLib {
52
53 using namespace std;
54
56 bool extendp)
57 : tol_(numeric_limits<real>::epsilon())
58 , tol2_(real(0.1) * tol_)
59 , taytol_(pow(tol_, real(0.6)))
60 , _a(a)
61 , _f(f)
62 , _k0(k0)
63 , _mu(_f * (2 - _f)) // e^2
64 , _mv(1 - _mu) // 1 - e^2
65 , _e(sqrt(_mu))
66 , _extendp(extendp)
67 , _eEu(_mu)
68 , _eEv(_mv)
69 {
70 if (!(isfinite(_a) && _a > 0))
71 throw GeographicErr("Equatorial radius is not positive");
72 if (!(_f > 0))
73 throw GeographicErr("Flattening is not positive");
74 if (!(_f < 1))
75 throw GeographicErr("Polar semi-axis is not positive");
76 if (!(isfinite(_k0) && _k0 > 0))
77 throw GeographicErr("Scale is not positive");
78 }
79
84 return utm;
85 }
86
87 void TransverseMercatorExact::zeta(real /*u*/, real snu, real cnu, real dnu,
88 real /*v*/, real snv, real cnv, real dnv,
89 real& taup, real& lam) const {
90 // Lee 54.17 but write
91 // atanh(snu * dnv) = asinh(snu * dnv / sqrt(cnu^2 + _mv * snu^2 * snv^2))
92 // atanh(_e * snu / dnv) =
93 // asinh(_e * snu / sqrt(_mu * cnu^2 + _mv * cnv^2))
94 // Overflow value s.t. atan(overflow) = pi/2
95 static const real
96 overflow = 1 / Math::sq(numeric_limits<real>::epsilon());
97 real
98 d1 = sqrt(Math::sq(cnu) + _mv * Math::sq(snu * snv)),
99 d2 = sqrt(_mu * Math::sq(cnu) + _mv * Math::sq(cnv)),
100 t1 = (d1 != 0 ? snu * dnv / d1 : (signbit(snu) ? -overflow : overflow)),
101 t2 = (d2 != 0 ? sinh( _e * asinh(_e * snu / d2) ) :
102 (signbit(snu) ? -overflow : overflow));
103 // psi = asinh(t1) - asinh(t2)
104 // taup = sinh(psi)
105 taup = t1 * hypot(real(1), t2) - t2 * hypot(real(1), t1);
106 lam = (d1 != 0 && d2 != 0) ?
107 atan2(dnu * snv, cnu * cnv) - _e * atan2(_e * cnu * snv, dnu * cnv) :
108 0;
109 }
110
111 void TransverseMercatorExact::dwdzeta(real /*u*/,
112 real snu, real cnu, real dnu,
113 real /*v*/,
114 real snv, real cnv, real dnv,
115 real& du, real& dv) const {
116 // Lee 54.21 but write (1 - dnu^2 * snv^2) = (cnv^2 + _mu * snu^2 * snv^2)
117 // (see A+S 16.21.4)
118 real d = _mv * Math::sq(Math::sq(cnv) + _mu * Math::sq(snu * snv));
119 du = cnu * dnu * dnv * (Math::sq(cnv) - _mu * Math::sq(snu * snv)) / d;
120 dv = -snu * snv * cnv * (Math::sq(dnu * dnv) + _mu * Math::sq(cnu)) / d;
121 }
122
123 // Starting point for zetainv
124 bool TransverseMercatorExact::zetainv0(real psi, real lam,
125 real& u, real& v) const {
126 bool retval = false;
127 if (psi < -_e * Math::pi()/4 &&
128 lam > (1 - 2 * _e) * Math::pi()/2 &&
129 psi < lam - (1 - _e) * Math::pi()/2) {
130 // N.B. this branch is normally not taken because psi < 0 is converted
131 // psi > 0 by Forward.
132 //
133 // There's a log singularity at w = w0 = Eu.K() + i * Ev.K(),
134 // corresponding to the south pole, where we have, approximately
135 //
136 // psi = _e + i * pi/2 - _e * atanh(cos(i * (w - w0)/(1 + _mu/2)))
137 //
138 // Inverting this gives:
139 real
140 psix = 1 - psi / _e,
141 lamx = (Math::pi()/2 - lam) / _e;
142 u = asinh(sin(lamx) / hypot(cos(lamx), sinh(psix))) *
143 (1 + _mu/2);
144 v = atan2(cos(lamx), sinh(psix)) * (1 + _mu/2);
145 u = _eEu.K() - u;
146 v = _eEv.K() - v;
147 } else if (psi < _e * Math::pi()/2 &&
148 lam > (1 - 2 * _e) * Math::pi()/2) {
149 // At w = w0 = i * Ev.K(), we have
150 //
151 // zeta = zeta0 = i * (1 - _e) * pi/2
152 // zeta' = zeta'' = 0
153 //
154 // including the next term in the Taylor series gives:
155 //
156 // zeta = zeta0 - (_mv * _e) / 3 * (w - w0)^3
157 //
158 // When inverting this, we map arg(w - w0) = [-90, 0] to
159 // arg(zeta - zeta0) = [-90, 180]
160 real
161 dlam = lam - (1 - _e) * Math::pi()/2,
162 rad = hypot(psi, dlam),
163 // atan2(dlam-psi, psi+dlam) + 45d gives arg(zeta - zeta0) in range
164 // [-135, 225). Subtracting 180 (since multiplier is negative) makes
165 // range [-315, 45). Multiplying by 1/3 (for cube root) gives range
166 // [-105, 15). In particular the range [-90, 180] in zeta space maps
167 // to [-90, 0] in w space as required.
168 ang = atan2(dlam-psi, psi+dlam) - real(0.75) * Math::pi();
169 // Error using this guess is about 0.21 * (rad/e)^(5/3)
170 retval = rad < _e * taytol_;
171 rad = cbrt(3 / (_mv * _e) * rad);
172 ang /= 3;
173 u = rad * cos(ang);
174 v = rad * sin(ang) + _eEv.K();
175 } else {
176 // Use spherical TM, Lee 12.6 -- writing atanh(sin(lam) / cosh(psi)) =
177 // asinh(sin(lam) / hypot(cos(lam), sinh(psi))). This takes care of the
178 // log singularity at zeta = Eu.K() (corresponding to the north pole)
179 v = asinh(sin(lam) / hypot(cos(lam), sinh(psi)));
180 u = atan2(sinh(psi), cos(lam));
181 // But scale to put 90,0 on the right place
182 u *= _eEu.K() / (Math::pi()/2);
183 v *= _eEu.K() / (Math::pi()/2);
184 }
185 return retval;
186 }
187
188 // Invert zeta using Newton's method
189 void TransverseMercatorExact::zetainv(real taup, real lam,
190 real& u, real& v) const {
191 real
192 psi = asinh(taup),
193 scal = 1/hypot(real(1), taup);
194 if (zetainv0(psi, lam, u, v))
195 return;
196 real stol2 = tol2_ / Math::sq(fmax(psi, real(1)));
197 // min iterations = 2, max iterations = 6; mean = 4.0
198 for (int i = 0, trip = 0; i < numit_ || GEOGRAPHICLIB_PANIC; ++i) {
199 real snu, cnu, dnu, snv, cnv, dnv;
200 _eEu.sncndn(u, snu, cnu, dnu);
201 _eEv.sncndn(v, snv, cnv, dnv);
202 real tau1, lam1, du1, dv1;
203 zeta(u, snu, cnu, dnu, v, snv, cnv, dnv, tau1, lam1);
204 dwdzeta(u, snu, cnu, dnu, v, snv, cnv, dnv, du1, dv1);
205 tau1 -= taup;
206 lam1 -= lam;
207 tau1 *= scal;
208 real
209 delu = tau1 * du1 - lam1 * dv1,
210 delv = tau1 * dv1 + lam1 * du1;
211 u -= delu;
212 v -= delv;
213 if (trip)
214 break;
215 real delw2 = Math::sq(delu) + Math::sq(delv);
216 if (!(delw2 >= stol2))
217 ++trip;
218 }
219 }
220
221 void TransverseMercatorExact::sigma(real /*u*/, real snu, real cnu, real dnu,
222 real v, real snv, real cnv, real dnv,
223 real& xi, real& eta) const {
224 // Lee 55.4 writing
225 // dnu^2 + dnv^2 - 1 = _mu * cnu^2 + _mv * cnv^2
226 real d = _mu * Math::sq(cnu) + _mv * Math::sq(cnv);
227 xi = _eEu.E(snu, cnu, dnu) - _mu * snu * cnu * dnu / d;
228 eta = v - _eEv.E(snv, cnv, dnv) + _mv * snv * cnv * dnv / d;
229 }
230
231 void TransverseMercatorExact::dwdsigma(real /*u*/,
232 real snu, real cnu, real dnu,
233 real /*v*/,
234 real snv, real cnv, real dnv,
235 real& du, real& dv) const {
236 // Reciprocal of 55.9: dw/ds = dn(w)^2/_mv, expanding complex dn(w) using
237 // A+S 16.21.4
238 real d = _mv * Math::sq(Math::sq(cnv) + _mu * Math::sq(snu * snv));
239 real
240 dnr = dnu * cnv * dnv,
241 dni = - _mu * snu * cnu * snv;
242 du = (Math::sq(dnr) - Math::sq(dni)) / d;
243 dv = 2 * dnr * dni / d;
244 }
245
246 // Starting point for sigmainv
247 bool TransverseMercatorExact::sigmainv0(real xi, real eta,
248 real& u, real& v) const {
249 bool retval = false;
250 if (eta > real(1.25) * _eEv.KE() ||
251 (xi < -real(0.25) * _eEu.E() && xi < eta - _eEv.KE())) {
252 // sigma as a simple pole at w = w0 = Eu.K() + i * Ev.K() and sigma is
253 // approximated by
254 //
255 // sigma = (Eu.E() + i * Ev.KE()) + 1/(w - w0)
256 real
257 x = xi - _eEu.E(),
258 y = eta - _eEv.KE(),
259 r2 = Math::sq(x) + Math::sq(y);
260 u = _eEu.K() + x/r2;
261 v = _eEv.K() - y/r2;
262 } else if ((eta > real(0.75) * _eEv.KE() && xi < real(0.25) * _eEu.E())
263 || eta > _eEv.KE()) {
264 // At w = w0 = i * Ev.K(), we have
265 //
266 // sigma = sigma0 = i * Ev.KE()
267 // sigma' = sigma'' = 0
268 //
269 // including the next term in the Taylor series gives:
270 //
271 // sigma = sigma0 - _mv / 3 * (w - w0)^3
272 //
273 // When inverting this, we map arg(w - w0) = [-pi/2, -pi/6] to
274 // arg(sigma - sigma0) = [-pi/2, pi/2]
275 // mapping arg = [-pi/2, -pi/6] to [-pi/2, pi/2]
276 real
277 deta = eta - _eEv.KE(),
278 rad = hypot(xi, deta),
279 // Map the range [-90, 180] in sigma space to [-90, 0] in w space. See
280 // discussion in zetainv0 on the cut for ang.
281 ang = atan2(deta-xi, xi+deta) - real(0.75) * Math::pi();
282 // Error using this guess is about 0.068 * rad^(5/3)
283 retval = rad < 2 * taytol_;
284 rad = cbrt(3 / _mv * rad);
285 ang /= 3;
286 u = rad * cos(ang);
287 v = rad * sin(ang) + _eEv.K();
288 } else {
289 // Else use w = sigma * Eu.K/Eu.E (which is correct in the limit _e -> 0)
290 u = xi * _eEu.K()/_eEu.E();
291 v = eta * _eEu.K()/_eEu.E();
292 }
293 return retval;
294 }
295
296 // Invert sigma using Newton's method
297 void TransverseMercatorExact::sigmainv(real xi, real eta,
298 real& u, real& v) const {
299 if (sigmainv0(xi, eta, u, v))
300 return;
301 // min iterations = 2, max iterations = 7; mean = 3.9
302 for (int i = 0, trip = 0; i < numit_ || GEOGRAPHICLIB_PANIC; ++i) {
303 real snu, cnu, dnu, snv, cnv, dnv;
304 _eEu.sncndn(u, snu, cnu, dnu);
305 _eEv.sncndn(v, snv, cnv, dnv);
306 real xi1, eta1, du1, dv1;
307 sigma(u, snu, cnu, dnu, v, snv, cnv, dnv, xi1, eta1);
308 dwdsigma(u, snu, cnu, dnu, v, snv, cnv, dnv, du1, dv1);
309 xi1 -= xi;
310 eta1 -= eta;
311 real
312 delu = xi1 * du1 - eta1 * dv1,
313 delv = xi1 * dv1 + eta1 * du1;
314 u -= delu;
315 v -= delv;
316 if (trip)
317 break;
318 real delw2 = Math::sq(delu) + Math::sq(delv);
319 if (!(delw2 >= tol2_))
320 ++trip;
321 }
322 }
323
324 void TransverseMercatorExact::Scale(real tau, real /*lam*/,
325 real snu, real cnu, real dnu,
326 real snv, real cnv, real dnv,
327 real& gamma, real& k) const {
328 real sec2 = 1 + Math::sq(tau); // sec(phi)^2
329 // Lee 55.12 -- negated for our sign convention. gamma gives the bearing
330 // (clockwise from true north) of grid north
331 gamma = atan2(_mv * snu * snv * cnv, cnu * dnu * dnv);
332 // Lee 55.13 with nu given by Lee 9.1 -- in sqrt change the numerator
333 // from
334 //
335 // (1 - snu^2 * dnv^2) to (_mv * snv^2 + cnu^2 * dnv^2)
336 //
337 // to maintain accuracy near phi = 90 and change the denomintor from
338 //
339 // (dnu^2 + dnv^2 - 1) to (_mu * cnu^2 + _mv * cnv^2)
340 //
341 // to maintain accuracy near phi = 0, lam = 90 * (1 - e). Similarly
342 // rewrite sqrt term in 9.1 as
343 //
344 // _mv + _mu * c^2 instead of 1 - _mu * sin(phi)^2
345 k = sqrt(_mv + _mu / sec2) * sqrt(sec2) *
346 sqrt( (_mv * Math::sq(snv) + Math::sq(cnu * dnv)) /
347 (_mu * Math::sq(cnu) + _mv * Math::sq(cnv)) );
348 }
349
350 void TransverseMercatorExact::Forward(real lon0, real lat, real lon,
351 real& x, real& y,
352 real& gamma, real& k) const {
353 lat = Math::LatFix(lat);
354 lon = Math::AngDiff(lon0, lon);
355 // Explicitly enforce the parity
356 int
357 latsign = (!_extendp && signbit(lat)) ? -1 : 1,
358 lonsign = (!_extendp && signbit(lon)) ? -1 : 1;
359 lon *= lonsign;
360 lat *= latsign;
361 bool backside = !_extendp && lon > Math::qd;
362 if (backside) {
363 if (lat == 0)
364 latsign = -1;
365 lon = Math::hd - lon;
366 }
367 real
368 lam = lon * Math::degree(),
369 tau = Math::tand(lat);
370
371 // u,v = coordinates for the Thompson TM, Lee 54
372 real u, v;
373 if (lat == Math::qd) {
374 u = _eEu.K();
375 v = 0;
376 } else if (lat == 0 && lon == Math::qd * (1 - _e)) {
377 u = 0;
378 v = _eEv.K();
379 } else
380 // tau = tan(phi), taup = sinh(psi)
381 zetainv(Math::taupf(tau, _e), lam, u, v);
382
383 real snu, cnu, dnu, snv, cnv, dnv;
384 _eEu.sncndn(u, snu, cnu, dnu);
385 _eEv.sncndn(v, snv, cnv, dnv);
386
387 real xi, eta;
388 sigma(u, snu, cnu, dnu, v, snv, cnv, dnv, xi, eta);
389 if (backside)
390 xi = 2 * _eEu.E() - xi;
391 y = xi * _a * _k0 * latsign;
392 x = eta * _a * _k0 * lonsign;
393
394 if (lat == Math::qd) {
395 gamma = lon;
396 k = 1;
397 } else {
398 // Recompute (tau, lam) from (u, v) to improve accuracy of Scale
399 zeta(u, snu, cnu, dnu, v, snv, cnv, dnv, tau, lam);
400 tau = Math::tauf(tau, _e);
401 Scale(tau, lam, snu, cnu, dnu, snv, cnv, dnv, gamma, k);
402 gamma /= Math::degree();
403 }
404 if (backside)
405 gamma = Math::hd - gamma;
406 gamma *= latsign * lonsign;
407 k *= _k0;
408 }
409
410 void TransverseMercatorExact::Reverse(real lon0, real x, real y,
411 real& lat, real& lon,
412 real& gamma, real& k) const {
413 // This undoes the steps in Forward.
414 real
415 xi = y / (_a * _k0),
416 eta = x / (_a * _k0);
417 // Explicitly enforce the parity
418 int
419 xisign = (!_extendp && signbit(xi)) ? -1 : 1,
420 etasign = (!_extendp && signbit(eta)) ? -1 : 1;
421 xi *= xisign;
422 eta *= etasign;
423 bool backside = !_extendp && xi > _eEu.E();
424 if (backside)
425 xi = 2 * _eEu.E()- xi;
426
427 // u,v = coordinates for the Thompson TM, Lee 54
428 real u, v;
429 if (xi == 0 && eta == _eEv.KE()) {
430 u = 0;
431 v = _eEv.K();
432 } else
433 sigmainv(xi, eta, u, v);
434
435 real snu, cnu, dnu, snv, cnv, dnv;
436 _eEu.sncndn(u, snu, cnu, dnu);
437 _eEv.sncndn(v, snv, cnv, dnv);
438 real phi, lam, tau;
439 if (v != 0 || u != _eEu.K()) {
440 zeta(u, snu, cnu, dnu, v, snv, cnv, dnv, tau, lam);
441 tau = Math::tauf(tau, _e);
442 phi = atan(tau);
443 lat = phi / Math::degree();
444 lon = lam / Math::degree();
445 Scale(tau, lam, snu, cnu, dnu, snv, cnv, dnv, gamma, k);
446 gamma /= Math::degree();
447 } else {
448 lat = Math::qd;
449 lon = lam = gamma = 0;
450 k = 1;
451 }
452
453 if (backside)
454 lon = Math::hd - lon;
455 lon *= etasign;
456 lon = Math::AngNormalize(lon + Math::AngNormalize(lon0));
457 lat *= xisign;
458 if (backside)
459 gamma = Math::hd - gamma;
460 gamma *= xisign * etasign;
461 k *= _k0;
462 }
463
464} // namespace GeographicLib
GeographicLib::Math::real real
Definition: GeodSolve.cpp:31
#define GEOGRAPHICLIB_PANIC
Definition: Math.hpp:61
Header for GeographicLib::TransverseMercatorExact class.
void sncndn(real x, real &sn, real &cn, real &dn) const
Exception handling for GeographicLib.
Definition: Constants.hpp:316
static T degree()
Definition: Math.hpp:200
static T tand(T x)
Definition: Math.cpp:172
static T LatFix(T x)
Definition: Math.hpp:300
static T sq(T x)
Definition: Math.hpp:212
static T tauf(T taup, T es)
Definition: Math.cpp:219
static T AngNormalize(T x)
Definition: Math.cpp:71
static T taupf(T tau, T es)
Definition: Math.cpp:209
static T pi()
Definition: Math.hpp:190
static T AngDiff(T x, T y, T &e)
Definition: Math.cpp:82
@ hd
degrees per half turn
Definition: Math.hpp:144
@ qd
degrees per quarter turn
Definition: Math.hpp:141
An exact implementation of the transverse Mercator projection.
void Forward(real lon0, real lat, real lon, real &x, real &y, real &gamma, real &k) const
TransverseMercatorExact(real a, real f, real k0, bool extendp=false)
static const TransverseMercatorExact & UTM()
void Reverse(real lon0, real x, real y, real &lat, real &lon, real &gamma, real &k) const
Namespace for GeographicLib.
Definition: Accumulator.cpp:12