SampleOptimization.cpp

This is an example of how to use the Optimization class by fitting curves of increasing degree to random data.

#include "Optimization.h"
#include "cv.h"
#include "highgui.h"
#include <time.h>
#include <vector>
#include <iostream>
#include <string>
using namespace std;
using namespace alvar;
const int res=640;
const double poly_res=8.0;
double random(int dist_type, double param1, double param2) {
static CvRNG rng=0;
if (rng == 0) rng = cvRNG(time(0));
double m_data;
CvMat m = cvMat(1, 1, CV_64F, &m_data);
cvRandArr(&rng, &m, dist_type, cvScalar(param1), cvScalar(param2));
return m_data;
}
double get_y(double x, double a, double b, double c, double d, double e) {
return (a*x*x*x*x + b*x*x*x + c*x*x + d*x + e);
}
// Just generate some random data that can be used as sensor input
bool get_measurement(double *x, double *y, double a, double b, double c, double d, double e) {
double xx = random(CV_RAND_UNI, -(poly_res/2), +(poly_res/2)); //(rand()*poly_res/RAND_MAX)-(poly_res/2);
double yy = get_y(xx, a, b, c, d, e);
double ry = random(CV_RAND_NORMAL, 0, poly_res/8); //(rand()*(poly_res/4)/RAND_MAX)-(poly_res/8);
yy += ry;
*x = xx;
*y = yy;
if (*y < -(poly_res/2)) return false;
if (*y >= (poly_res/2)) return false;
return true;
}
void Estimate(CvMat* state, CvMat *projection, void *param) {
double *measx=(double *)param;
int data_degree = state->rows-1;
double a = (data_degree >= 4? cvmGet(state, 4, 0) : 0);
double b = (data_degree >= 3? cvmGet(state, 3, 0) : 0);
double c = (data_degree >= 2? cvmGet(state, 2, 0) : 0);
double d = (data_degree >= 1? cvmGet(state, 1, 0) : 0);
double e = (data_degree >= 0? cvmGet(state, 0, 0) : 0);
for (int i=0; i<projection->rows; i++) {
cvmSet(projection, i, 0, get_y(measx[i], a, b, c, d, e));
}
}
int main(int argc, char *argv[])
{
try {
// Output usage message
std::string filename(argv[0]);
filename = filename.substr(filename.find_last_of('\\') + 1);
std::cout << "SampleOptimization" << std::endl;
std::cout << "==================" << std::endl;
std::cout << std::endl;
std::cout << "Description:" << std::endl;
std::cout << " This is an example of how to use the 'Optimization' class. Random data" << std::endl;
std::cout << " is generated and approximated using curves of increasing degrees." << std::endl;
std::cout << std::endl;
std::cout << "Usage:" << std::endl;
std::cout << " " << filename << std::endl;
std::cout << std::endl;
std::cout << "Keyboard Shortcuts:" << std::endl;
std::cout << " any key: cycle through datasets" << std::endl;
std::cout << " q: quit" << std::endl;
std::cout << std::endl;
// Processing loop
IplImage *img = cvCreateImage(cvSize(res,res), IPL_DEPTH_8U, 3);
cvNamedWindow("SampleOptimization");
for (int data_degree=0; data_degree<5; data_degree++) {
double a = (data_degree >= 4? random(CV_RAND_UNI, -0.5, 0.5) : 0);
double b = (data_degree >= 3? random(CV_RAND_UNI, -0.5, 0.5) : 0);
double c = (data_degree >= 2? random(CV_RAND_UNI, -0.5, 0.5) : 0);
double d = (data_degree >= 1? random(CV_RAND_UNI, -0.5, 0.5) : 0);
double e = (data_degree >= 0? random(CV_RAND_UNI, -0.5, 0.5) : 0);
cvZero(img);
vector<CvPoint2D32f> measvec;
for (int i=0; i<1000; i++) {
double x, y;
if (get_measurement(&x, &y, a, b, c, d, e)) {
measvec.push_back(cvPoint2D32f(x, y));
x = (x*res/poly_res)+(res/2);
y = (y*res/poly_res)+(res/2);
cvCircle(img, cvPoint(int(x), int(y)), 1, CV_RGB(0,255,0));
}
}
cvShowImage("SampleOptimization", img);
cvWaitKey(10);
double measx[1000];
CvMat *meas = cvCreateMat(measvec.size(), 1, CV_64F);
for (size_t i=0; i<measvec.size(); i++) {
measx[i] = measvec[i].x;
cvmSet(meas, i, 0, measvec[i].y);
}
for (int degree=0; degree<5; degree++)
{
double param_data[5]={0};
CvMat param = cvMat(degree+1, 1, CV_64F, param_data);
Optimization opt(param.rows, meas->rows);
opt.Optimize(&param, meas, 0.1, 100, Estimate, measx);
double a = (degree >= 4? cvmGet(&param, 4, 0) : 0);
double b = (degree >= 3? cvmGet(&param, 3, 0) : 0);
double c = (degree >= 2? cvmGet(&param, 2, 0) : 0);
double d = (degree >= 1? cvmGet(&param, 1, 0) : 0);
double e = (degree >= 0? cvmGet(&param, 0, 0) : 0);
const int step=5;
for (int x2=step; x2<res; x2+=step) {
int x1 = x2-step;
double xx1 = (x1*poly_res/res)-(poly_res/2);
double xx2 = (x2*poly_res/res)-(poly_res/2);
double yy1 = get_y(xx1, a, b, c, d, e);
double yy2 = get_y(xx2, a, b, c, d, e);
int y1 = int((yy1*res/poly_res)+(res/2));
int y2 = int((yy2*res/poly_res)+(res/2));
cvLine(img, cvPoint(x1,y1), cvPoint(x2,y2), CV_RGB(degree*50,255-(degree*50),255));
}
cvShowImage("SampleOptimization", img);
cvWaitKey(10);
}
cvReleaseMat(&meas);
cvShowImage("SampleOptimization", img);
int key = cvWaitKey(0);
if (key == 'q') {
break;
}
}
cvReleaseImage(&img);
return 0;
}
catch (const std::exception &e) {
std::cout << "Exception: " << e.what() << endl;
}
catch (...) {
std::cout << "Exception: unknown" << std::endl;
}
}

ar_track_alvar
Author(s): Scott Niekum
autogenerated on Mon Jun 10 2019 12:47:03