/**
* \file transform.c
*
* \brief This file contains routines which perform (affine?)
* transformations from one coordinate system into another.
*
* The second system may be translated, stretched, and rotated relative
* to the first. The input system is system a and the output
* system is b.
*
* This program is free software under the GNU General Public License
* (>=v2). Read the file COPYING that comes with GRASS for details.
*
* \author GRASS GIS Development Team
*
* \date 1987-2007
*/
/****************************************************************
note: uses sqrt() from math library
*****************************************************************
Points from one system may be converted into the second by
use of one of the two equation routines.
transform_a_into_b (ax,ay,bx,by)
double ax,ay; input point from system a
double *bx,*by; resultant point in system b
transform_b_into_a (bx,by,ax,ay)
double bx,by; input point from system b
double *ax,*ay; resultant point in system a
*****************************************************************
Residual analysis on the equation can be run to test how well
the equations work. Either test how well b is predicted by a
or vice versa.
residuals_a_predicts_b (ax,ay,bx,by,use,n,residuals,rms)
residuals_b_predicts_a (ax,ay,bx,by,use,n,residuals,rms)
double ax[], ay[]; coordinate from system a
double bx[], by[]; coordinate from system b
char use[]; use point flags
int n; number of points in ax,ay,bx,by
double residual[] residual error for each point
double *rms; overall root mean square error
****************************************************************/
#include
#include
#include
/* the coefficients */
static double A0, A1, A2, A3, A4, A5;
static double B0, B1, B2, B3, B4, B5;
/* function prototypes */
static int resid(double *, double *, double *, double *, int *, int, double *,
double *, int);
/**
* \fn int compute_transformation_coef (double ax[], double ay[], double bx[], double by[], char *use, int n)
*
* \brief The first step is to compute coefficients for a set of equations
* which are then used to convert from the one system to the other.
*
* A set of x,y points from both systems is input into the equation
* generator which determines the equation coefficients which most
* nearly represent the original points. These coefficients are kept
* in a static variables internal to this file.
*
* NOTE: use[i] must be true for ax[i],ay[i],bx[i],by[i] to be used
* in the equation. Also, the total number of used points must be
* 4 or larger.
*
* \param[in] ax coordinate from system a
* \param[in] ay coordinate from system a
* \param[in] bx coordinate from system b
* \param[in] by coordinate from system b
* \param[in] use use point flags
* \param[in] n number of points in ax, ay, bx, by
* \return int 1 if successful
* \return int -1 if could not solve equation. Points probably colinear.
* \return int -2 if less than 4 points used
*/
int compute_transformation_coef(double ax[], double ay[], double bx[],
double by[], int *use, int n)
{
int i;
int j;
int count;
double aa[3];
double aar[3];
double bb[3];
double bbr[3];
double cc[3][3];
double x;
count = 0;
for (i = 0; i < n; i++)
if (use[i])
count++;
if (count < 4)
return -2; /* must have at least 4 points */
for (i = 0; i < 3; i++) {
aa[i] = bb[i] = 0.0;
for (j = 0; j < 3; j++)
cc[i][j] = 0.0;
}
for (i = 0; i < n; i++) {
if (!use[i])
continue; /* skip this point */
cc[0][0] += 1;
cc[0][1] += bx[i];
cc[0][2] += by[i];
cc[1][1] += bx[i] * bx[i];
cc[1][2] += bx[i] * by[i];
cc[2][2] += by[i] * by[i];
aa[0] += ay[i];
aa[1] += ay[i] * bx[i];
aa[2] += ay[i] * by[i];
bb[0] += ax[i];
bb[1] += ax[i] * bx[i];
bb[2] += ax[i] * by[i];
}
cc[1][0] = cc[0][1];
cc[2][0] = cc[0][2];
cc[2][1] = cc[1][2];
/* aa and bb are solved */
if (inverse(cc) < 0)
return (-1);
if (m_mult(cc, aa, aar) < 0 || m_mult(cc, bb, bbr) < 0)
return (-1);
/* the equation coefficients */
B0 = aar[0];
B1 = aar[1];
B2 = aar[2];
B3 = bbr[0];
B4 = bbr[1];
B5 = bbr[2];
/* the inverse equation */
x = B2 * B4 - B1 * B5;
if (!x)
return (-1);
A0 = (B1 * B3 - B0 * B4) / x;
A1 = -B1 / x;
A2 = B4 / x;
A3 = (B0 * B5 - B2 * B3) / x;
A4 = B2 / x;
A5 = -B5 / x;
return 1;
}
int transform_a_into_b(double ax, double ay, double *bx, double *by)
{
*by = A0 + A1 * ax + A2 * ay;
*bx = A3 + A4 * ax + A5 * ay;
return 0;
}
int transform_b_into_a(double bx, double by, double *ax, double *ay)
{
*ay = B0 + B1 * bx + B2 * by;
*ax = B3 + B4 * bx + B5 * by;
return 0;
}
/**************************************************************
These routines are internal to this source code
solve (a, b)
double a[3][3]
double b[3]
equation solver used by compute_transformation_coef()
**************************************************************/
/* #define abs(xx) (xx >= 0 ? xx : -xx) */
/* #define N 3 */
int residuals_a_predicts_b(double ax[], double ay[], double bx[], double by[],
int use[], int n, double residuals[], double *rms)
{
resid(ax, ay, bx, by, use, n, residuals, rms, 1);
return 0;
}
int residuals_b_predicts_a(double ax[], double ay[], double bx[], double by[],
int use[], int n, double residuals[], double *rms)
{
resid(ax, ay, bx, by, use, n, residuals, rms, 0);
return 0;
}
/**
* \fn int print_transform_matrix (void)
*
* \brief Prints matrix to stdout in human readable format.
*
* \return int 1
*/
int print_transform_matrix(void)
{
fprintf(stdout, "\nTransformation Matrix\n");
fprintf(stdout, "| xoff a b |\n");
fprintf(stdout, "| yoff d e |\n");
fprintf(stdout, "-------------------------------------------\n");
fprintf(stdout, "%f %f %f \n", -B3, B2, -B5);
fprintf(stdout, "%f %f %f \n", -B0, -B1, B4);
fprintf(stdout, "-------------------------------------------\n");
return 1;
}
static int resid(double ax[], double ay[], double bx[], double by[],
int use[], int n, double residuals[], double *rms, int atob)
{
double x, y;
int i;
int count;
double sum;
double delta;
double dx, dy;
count = 0;
sum = 0.0;
for (i = 0; i < n; i++) {
if (!use[i])
continue;
count++;
if (atob) {
transform_a_into_b(ax[i], ay[i], &x, &y);
dx = x - bx[i];
dy = y - by[i];
}
else {
transform_b_into_a(bx[i], by[i], &x, &y);
dx = x - ax[i];
dy = y - ay[i];
}
delta = dx * dx + dy * dy;
residuals[i] = sqrt(delta);
sum += delta;
}
*rms = sqrt(sum / count);
return 0;
}