1.opencv中3D點根據相機引數投影成2D點，直接上程式碼：

輸入：3D座標+旋轉，平移矩陣（所謂的相機姿態）+相機內參（包括畸變矩陣）

輸出：2D座標

（1.投影函式：根據相機引數（包括畸變矩陣），把3D點投影成2D點

2.搞清楚R和t的具體含義。

R的第i行 表示攝像機座標系中的第i個座標軸方向的單位向量在世界座標系裡的座標；
R的第i列 表示世界座標系中的第i個座標軸方向的單位向量在攝像機座標系裡的座標；
t 表示世界座標系的原點在攝像機座標系的座標；
-R的轉置 * t 表示攝像機座標系的原點在世界座標系的座標。

#include "opencv2/core/core.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include "opencv2/calib3d/calib3d.hpp"
#include "opencv2/highgui/highgui.hpp"
#include <iostream>
#include <string>
std::vector<cv::Point3d> Generate3DPoints();

int main(int argc, char* argv[])
{
// Read 3D points
std::vector<cv::Point3d> objectPoints = Generate3DPoints();
std::vector<cv::Point2d> imagePoints;
cv::Mat intrisicMat(3, 3, cv::DataType<double>::type); // Intrisic matrix
intrisicMat.at<double>(0, 0) = 1.6415318549788924e+003;
intrisicMat.at<double>(1, 0) = 0;
intrisicMat.at<double>(2, 0) = 0;
intrisicMat.at<double>(0, 1) = 0;
intrisicMat.at<double>(1, 1) = 1.7067753507885654e+003;
intrisicMat.at<double>(2, 1) = 0;
intrisicMat.at<double>(0, 2) = 5.3262822453148601e+002;
intrisicMat.at<double>(1, 2) = 3.8095355839052968e+002;
intrisicMat.at<double>(2, 2) = 1;
cv::Mat rVec(3, 1, cv::DataType<double>::type); // Rotation vector
rVec.at<double>(0) = -3.9277902400761393e-002;
rVec.at<double>(1) = 3.7803824407602084e-002;
rVec.at<double>(2) = 2.6445674487856268e-002;
cv::Mat tVec(3, 1, cv::DataType<double>::type); // Translation vector
tVec.at<double>(0) = 2.1158489381208221e+000;
tVec.at<double>(1) = -7.6847683212704716e+000;
tVec.at<double>(2) = 2.6169795190294256e+001;
cv::Mat distCoeffs(5, 1, cv::DataType<double>::type);   // Distortion vector
distCoeffs.at<double>(0) = -7.9134632415085826e-001;
distCoeffs.at<double>(1) = 1.5623584435644169e+000;
distCoeffs.at<double>(2) = -3.3916502741726508e-002;
distCoeffs.at<double>(3) = -1.3921577146136694e-002;
distCoeffs.at<double>(4) = 1.1430734623697941e+002;
std::cout << "Intrisic matrix: " << intrisicMat << std::endl << std::endl;
std::cout << "Rotation vector: " << rVec << std::endl << std::endl;
std::cout << "Translation vector: " << tVec << std::endl << std::endl;
std::cout << "Distortion coef: " << distCoeffs << std::endl << std::endl;
cv::projectPoints(objectPoints, rVec, tVec, intrisicMat, distCoeffs, imagePoints);
std::cout << "Press any key to exit.";
std::cin.ignore();
std::cin.get();
return 0;
}
std::vector<cv::Point3d> Generate3DPoints()
{
std::vector<cv::Point3d> points;
double x, y, z;
x = .5; y = .5; z = -.5;
points.push_back(cv::Point3d(x, y, z));
x = .5; y = .5; z = .5;
points.push_back(cv::Point3d(x, y, z));
x = -.5; y = .5; z = .5;
points.push_back(cv::Point3d(x, y, z));
x = -.5; y = .5; z = -.5;
points.push_back(cv::Point3d(x, y, z));
x = .5; y = -.5; z = -.5;
points.push_back(cv::Point3d(x, y, z));
x = -.5; y = -.5; z = -.5;
points.push_back(cv::Point3d(x, y, z));
x = -.5; y = -.5; z = .5;
points.push_back(cv::Point3d(x, y, z));
for (unsigned int i = 0; i < points.size(); ++i)
{
std::cout << points[i] << std::endl << std::endl;
}
return points;
}

2.opencv中solvePnP函式計算相機姿態：http://blog.csdn.net/aptx704610875/article/details/48915149

或者http://www.cnblogs.com/gaoxiang12/p/4659805.html

輸入：3D座標，2D座標+相機內參（包括畸變矩陣）

輸出：相機姿態（旋轉平移矩陣）

3.超級好的一個部落格！：http://www.cnblogs.com/gaoxiang12/p/4652478.html（這個部落格還包括根據RGB-D影象，點雲拼接（三維重建），很詳細！）

（涉及到了三維點雲，除了opencv,還用到PCL）

原理+程式碼實現從（2D->3D）

輸入：2D座標，相機內參（包括畸變矩陣），相機姿態

輸出：3D座標

（1,2,3其實就是3D，2D座標，相機姿態（在已經知道相機引數情況下），知道其中兩個可以求得另外一個！）

4.用kinect1來拍攝三維模型，很重要的一步就是先要標定相機（因為深度相機與彩色相機不是一個位置，存在偏差）：

（補充：