public static void Normalize(Matrix A, Matrix B) { B.Scale(A, 1.0 / A.Norm()); }
public static Matrix Homography(List<Matrix> worldPoints, List<System.Drawing.PointF> imagePoints) { int n = worldPoints.Count; // normalize image coordinates var mu = new Matrix(2, 1); for (int i = 0; i < n; i++) { mu[0] += imagePoints[i].X; mu[1] += imagePoints[i].Y; } mu.Scale(1.0 / n); var muAbs = new Matrix(2, 1); for (int i = 0; i < n; i++) { muAbs[0] += Math.Abs(imagePoints[i].X - mu[0]); muAbs[1] += Math.Abs(imagePoints[i].Y - mu[1]); } muAbs.Scale(1.0 / n); var Hnorm = Matrix.Identity(3, 3); Hnorm[0, 0] = 1 / muAbs[0]; Hnorm[1, 1] = 1 / muAbs[1]; Hnorm[0, 2] = -mu[0] / muAbs[0]; Hnorm[1, 2] = -mu[1] / muAbs[1]; var invHnorm = Matrix.Identity(3, 3); invHnorm[0, 0] = muAbs[0]; invHnorm[1, 1] = muAbs[1]; invHnorm[0, 2] = mu[0]; invHnorm[1, 2] = mu[1]; var A = Matrix.Zero(2 * n, 9); for (int i = 0; i < n; i++) { var X = worldPoints[i]; var imagePoint = imagePoints[i]; var x = new Matrix(3, 1); x[0] = imagePoint.X; x[1] = imagePoint.Y; x[2] = 1; var xn = new Matrix(3, 1); xn.Mult(Hnorm, x); // Zhang's formulation; Hartley's is similar int ii = 2 * i; A[ii, 0] = X[0]; A[ii, 1] = X[1]; A[ii, 2] = 1; A[ii, 6] = -xn[0] * X[0]; A[ii, 7] = -xn[0] * X[1]; A[ii, 8] = -xn[0]; ii++; // next row A[ii, 3] = X[0]; A[ii, 4] = X[1]; A[ii, 5] = 1; A[ii, 6] = -xn[1] * X[0]; A[ii, 7] = -xn[1] * X[1]; A[ii, 8] = -xn[1]; } // h is the eigenvector of ATA with the smallest eignvalue var h = new Matrix(9, 1); { var ATA = new Matrix(9, 9); ATA.MultATA(A, A); var V = new Matrix(9, 9); var ww = new Matrix(9, 1); ATA.Eig(V, ww); h.CopyCol(V, 0); } var Hn = new Matrix(3, 3); Hn.Reshape(h); var H = new Matrix(3, 3); H.Mult(invHnorm, Hn); return H; }
// Use DLT to obtain estimate of calibration rig pose; in our case this is the pose of the Kinect camera. // This pose estimate will provide a good initial estimate for subsequent projector calibration. // Note for a full PnP solution we should probably refine with Levenberg-Marquardt. // DLT is described in Hartley and Zisserman p. 178 public static void DLT(Matrix cameraMatrix, Matrix distCoeffs, List<Matrix> worldPoints, List<System.Drawing.PointF> imagePoints, out Matrix R, out Matrix t) { int n = worldPoints.Count; var A = Matrix.Zero(2 * n, 12); for (int j = 0; j < n; j++) { var X = worldPoints[j]; var imagePoint = imagePoints[j]; double x, y; Undistort(cameraMatrix, distCoeffs, imagePoint.X, imagePoint.Y, out x, out y); int ii = 2 * j; A[ii, 4] = -X[0]; A[ii, 5] = -X[1]; A[ii, 6] = -X[2]; A[ii, 7] = -1; A[ii, 8] = y * X[0]; A[ii, 9] = y * X[1]; A[ii, 10] = y * X[2]; A[ii, 11] = y; ii++; // next row A[ii, 0] = X[0]; A[ii, 1] = X[1]; A[ii, 2] = X[2]; A[ii, 3] = 1; A[ii, 8] = -x * X[0]; A[ii, 9] = -x * X[1]; A[ii, 10] = -x * X[2]; A[ii, 11] = -x; } // Pcolumn is the eigenvector of ATA with the smallest eignvalue var Pcolumn = new Matrix(12, 1); { var ATA = new Matrix(12, 12); ATA.MultATA(A, A); var V = new Matrix(12, 12); var ww = new Matrix(12, 1); ATA.Eig(V, ww); Pcolumn.CopyCol(V, 0); } // reshape into 3x4 projection matrix var P = new Matrix(3, 4); P.Reshape(Pcolumn); R = new Matrix(3, 3); for (int i = 0; i < 3; i++) for (int j = 0; j < 3; j++) R[i, j] = P[i, j]; if (R.Det3x3() < 0) { R.Scale(-1); P.Scale(-1); } // orthogonalize R { var U = new Matrix(3, 3); var V = new Matrix(3, 3); var ww = new Matrix(3, 1); R.SVD(U, ww, V); R.MultAAT(U, V); } // determine scale factor var RP = new Matrix(3, 3); for (int i = 0; i < 3; i++) for (int j = 0; j < 3; j++) RP[i, j] = P[i, j]; double s = RP.Norm() / R.Norm(); t = new Matrix(3, 1); for (int i = 0; i < 3; i++) t[i] = P[i, 3]; t.Scale(1.0 / s); }
public static void PlanarDLT(Matrix cameraMatrix, Matrix distCoeffs, List<Matrix> worldPoints, List<System.Drawing.PointF> imagePoints, out Matrix R, out Matrix t) { int n = worldPoints.Count; var undistortedImagePoints = new List<System.Drawing.PointF>(); for (int i = 0; i < n; i++) { var imagePoint = imagePoints[i]; double x, y; Undistort(cameraMatrix, distCoeffs, imagePoint.X, imagePoint.Y, out x, out y); var undistorted = new System.Drawing.PointF(); undistorted.X = (float)x; undistorted.Y = (float)y; undistortedImagePoints.Add(undistorted); } var H = Homography(worldPoints, undistortedImagePoints); H.Scale(1.0 / H[2, 2]); //Console.WriteLine(H); var r1 = new Matrix(3, 1); r1.CopyCol(H, 0); var r2 = new Matrix(3, 1); r2.CopyCol(H, 1); t = new Matrix(3, 1); t.CopyCol(H, 2); t.Scale(1 / ((r1.Norm() + r2.Norm()) / 2.0)); r1.Scale(1 / r1.Norm()); r2.Scale(1 / r2.Norm()); var r3 = new Matrix(3, 1); r3.Cross(r1, r2); R = new Matrix(3, 3); for (int i = 0; i < 3; i++) { R[i, 0] = r1[i]; R[i, 1] = r2[i]; R[i, 2] = r3[i]; } }
static public void PlaneFit(IList<Matrix> X, out Matrix R, out Matrix t, out Matrix d2) { int n = X.Count; var mu = new Matrix(3, 1); for (int i = 0; i < n; i++) mu.Add(X[i]); mu.Scale(1f / (float)n); var A = new Matrix(3, 3); var xc = new Matrix(3, 1); var M = new Matrix(3, 3); for (int i = 0; i < X.Count; i++) { var x = X[i]; xc.Sub(x, mu); M.Outer(xc, xc); A.Add(M); } var V = new Matrix(3, 3); var d = new Matrix(3, 1); A.Eig(V, d); // eigenvalues in ascending order // arrange in descending order so that z = 0 var V2 = new Matrix(3, 3); for (int i = 0; i < 3; i++) { V2[i, 2] = V[i, 0]; V2[i, 1] = V[i, 1]; V2[i, 0] = V[i, 2]; } d2 = new Matrix(3, 1); d2[2] = d[0]; d2[1] = d[1]; d2[0] = d[2]; R = new Matrix(3, 3); R.Transpose(V2); if (R.Det3x3() < 0) R.Scale(-1); t = new Matrix(3, 1); t.Mult(R, mu); t.Scale(-1); // eigenvalues are the sum of squared distances in each direction // i.e., min eigenvalue is the sum of squared distances to the plane = d2[2] // compute the distance to the plane by transforming to the plane and take z-coordinate: // xPlane = R*x + t; distance = xPlane[2] }
public void SaveToOBJ(string directory, string objPath) { var objFilename = Path.GetFileNameWithoutExtension(objPath); var objDirectory = Path.GetDirectoryName(objPath); if (!Directory.Exists(objDirectory)) Directory.CreateDirectory(objDirectory); // Because we need to form triangles, we go back to the depth image var quadOffsets = new System.Drawing.Point[] { new System.Drawing.Point(0, 0), new System.Drawing.Point(1, 0), new System.Drawing.Point(0, 1), new System.Drawing.Point(1, 0), new System.Drawing.Point(1, 1), new System.Drawing.Point(0, 1), }; var streamWriter = new CultureInvariantStreamWriter(objDirectory + "/" + objFilename + ".obj"); var mtlFileWriter = new CultureInvariantStreamWriter(objDirectory + "/" + objFilename + ".mtl"); streamWriter.WriteLine("mtllib " + objFilename + ".mtl"); uint nextVertexIndex = 1; var depthImage = new FloatImage(Kinect2Calibration.depthImageWidth, Kinect2Calibration.depthImageHeight); foreach (var camera in cameras) { mtlFileWriter.WriteLine("newmtl camera" + camera.name); mtlFileWriter.WriteLine("Ka 1.000000 1.000000 1.000000"); mtlFileWriter.WriteLine("Kd 1.000000 1.000000 1.000000"); mtlFileWriter.WriteLine("Ks 0.000000 0.000000 0.000000"); mtlFileWriter.WriteLine("Tr 1.000000"); mtlFileWriter.WriteLine("illum 1"); mtlFileWriter.WriteLine("Ns 0.000000"); mtlFileWriter.WriteLine("map_Kd " + objFilename + "_" + camera.name + ".jpg"); File.Copy(directory + "/camera" + camera.name + "/color.jpg", objDirectory + "/" + objFilename + "_" + camera.name + ".jpg", true); streamWriter.WriteLine("usemtl camera" + camera.name); // load depth image string cameraDirectory = directory + "/camera" + camera.name; depthImage.LoadFromFile(cameraDirectory + "/mean.bin"); var calibration = camera.calibration; var depthFrameToCameraSpaceTable = calibration.ComputeDepthFrameToCameraSpaceTable(); var vertices = new Vertex[Kinect2Calibration.depthImageWidth * Kinect2Calibration.depthImageHeight]; var colorCamera = new Matrix(4, 1); var depthCamera = new Matrix(4, 1); var world = new Matrix(4, 1); for (int y = 0; y < Kinect2Calibration.depthImageHeight; y++) for (int x = 0; x < Kinect2Calibration.depthImageWidth; x++) { // depth camera coords var depth = depthImage[x, y] / 1000f; // m // convert to depth camera space var point = depthFrameToCameraSpaceTable[Kinect2Calibration.depthImageWidth * y + x]; depthCamera[0] = point.X * depth; depthCamera[1] = point.Y * depth; depthCamera[2] = depth; depthCamera[3] = 1; // world coordinates world.Mult(camera.pose, depthCamera); //world.Scale(1.0 / world[3]); not necessary for this transform // convert to color camera space colorCamera.Mult(calibration.depthToColorTransform, depthCamera); colorCamera.Scale(1.0 / colorCamera[3]); // project to color image double colorU, colorV; CameraMath.Project(calibration.colorCameraMatrix, calibration.colorLensDistortion, colorCamera[0], colorCamera[1], colorCamera[2], out colorU, out colorV); colorU /= (double)Kinect2Calibration.colorImageWidth; colorV /= (double)Kinect2Calibration.colorImageHeight; var vertex = new Vertex(); vertex.x = (float)world[0]; vertex.y = (float)world[1]; vertex.z = (float)world[2]; vertex.u = (float)colorU; vertex.v = (float)colorV; vertices[Kinect2Calibration.depthImageWidth * y + x] = vertex; } streamWriter.WriteLine("g camera" + camera.name); streamWriter.WriteLine("usemtl camera" + camera.name); // examine each triangle for (int y = 0; y < Kinect2Calibration.depthImageHeight - 1; y++) for (int x = 0; x < Kinect2Calibration.depthImageWidth - 1; x++) { int offseti = 0; for (int tri = 0; tri < 2; tri++) { // the indexes of the vertices of this triangle var i0 = Kinect2Calibration.depthImageWidth * (y + quadOffsets[offseti].Y) + (x + quadOffsets[offseti].X); var i1 = Kinect2Calibration.depthImageWidth * (y + quadOffsets[offseti + 1].Y) + (x + quadOffsets[offseti + 1].X); var i2 = Kinect2Calibration.depthImageWidth * (y + quadOffsets[offseti + 2].Y) + (x + quadOffsets[offseti + 2].X); // is triangle valid? bool nonZero = (vertices[i0].z != 0) && (vertices[i1].z != 0) && (vertices[i2].z != 0); bool jump01 = Vertex.DistanceSquared(vertices[i0], vertices[i1]) < 0.2 * 0.2; bool jump02 = Vertex.DistanceSquared(vertices[i0], vertices[i2]) < 0.2 * 0.2; bool jump12 = Vertex.DistanceSquared(vertices[i1], vertices[i2]) < 0.2 * 0.2; bool valid = nonZero && jump01 && jump02 && jump12; if (valid) { // only add the vertex if we haven't already if (vertices[i0].index == 0) { streamWriter.WriteLine(vertices[i0]); vertices[i0].index = nextVertexIndex++; } if (vertices[i1].index == 0) { streamWriter.WriteLine(vertices[i1]); vertices[i1].index = nextVertexIndex++; } if (vertices[i2].index == 0) { streamWriter.WriteLine(vertices[i2]); vertices[i2].index = nextVertexIndex++; } streamWriter.WriteLine("f {0}/{0} {1}/{1} {2}/{2}", vertices[i0].index, vertices[i1].index, vertices[i2].index); } offseti += 3; } } } streamWriter.Close(); mtlFileWriter.Close(); }
public void OptimizePose() { UnifyPose(); // joint estimate of projector and camera pose // minimize wrt T_CjW, T_WPk: Sum_ijk v_ijk [ p_k( T_WPk T_CjW x_i ) - y_ik ]^2 // cameras observe points x_i (in camera coords) // point x_i is observed to project to point y_ik in projector k // v_ijk === 1 if point i is observed by camera j and imaged by projector k // p_k(x) projects point x in projector k; x in projector coordinates // T_CjW camera j local coordinates to world coordinates // T_WPk world to projector k coorindates // efficient implementation: list of points x_ijk for which v_ijk != 0; store j, k with each point x_i // solve for C_j, P_k; C_0 is not in the set of parameters // parameters: for each projector and camera: 1 rotation + 1 translation = 6 parameters // We leave T_C0W fixed, so have 6 * (numProjectors + numCameras - 1) parameters int nParameters = 6 * (projectors.Count + cameras.Count - 1); //double[] parameters = new double[nParameters]; var parameters = new Matrix(nParameters, 1); // loop over room.cameras, room.projectors to form up parameters array { int pi = 0; // index into our parameter array for (int i = 1; i < cameras.Count; i++) // skip first one, which is our root { var T = cameras[i].pose; var R = new Matrix(3, 3); var t = new Matrix(3, 1); for (int ii = 0; ii < 3; ii++) { t[ii] = T[ii, 3]; for (int jj = 0; jj < 3; jj++) R[ii, jj] = T[ii, jj]; } var r = CameraMath.RotationVectorFromRotationMatrix(R); for (int ii = 0; ii < 3; ii++) parameters[pi++] = r[ii]; for (int ii = 0; ii < 3; ii++) parameters[pi++] = t[ii]; } for (int i = 0; i < projectors.Count; i++) { var T = projectors[i].pose; var R = new Matrix(3, 3); var t = new Matrix(3, 1); for (int ii = 0; ii < 3; ii++) { t[ii] = T[ii, 3]; for (int jj = 0; jj < 3; jj++) R[ii, jj] = T[ii, jj]; } var r = CameraMath.RotationVectorFromRotationMatrix(R); for (int ii = 0; ii < 3; ii++) parameters[pi++] = r[ii]; for (int ii = 0; ii < 3; ii++) parameters[pi++] = t[ii]; } } // count the number of values // use only inliers from previous step int nValues = 0; foreach (var projector in projectors) foreach (var camera in projector.calibrationPointSets.Keys) nValues += projector.calibrationPointSets[camera].worldPointInliers.Count * 2; // count components LevenbergMarquardt.Function optimize = delegate(Matrix p) { var fvec = new Matrix(nValues, 1); // convert p to transforms etc. // convert back to transforms and put back in our structures int pi = 0; // index into our parameter array for (int i = 1; i < cameras.Count; i++) // skip first one, which is our root { var r = new Matrix(3, 1); r[0] = p[pi++]; r[1] = p[pi++]; r[2] = p[pi++]; var R = CameraMath.RotationMatrixFromRotationVector(r); var t = new Matrix(3, 1); t[0] = p[pi++]; t[1] = p[pi++]; t[2] = p[pi++]; var T = new Matrix(4, 4); T.Identity(); for (int ii = 0; ii < 3; ii++) { for (int jj = 0; jj < 3; jj++) T[ii, jj] = R[ii, jj]; T[ii, 3] = t[ii]; } cameras[i].pose = T; } for (int i = 0; i < projectors.Count; i++) { var r = new Matrix(3, 1); r[0] = p[pi++]; r[1] = p[pi++]; r[2] = p[pi++]; var R = CameraMath.RotationMatrixFromRotationVector(r); var t = new Matrix(3, 1); t[0] = p[pi++]; t[1] = p[pi++]; t[2] = p[pi++]; var T = new Matrix(4, 4); T.Identity(); for (int ii = 0; ii < 3; ii++) { for (int jj = 0; jj < 3; jj++) T[ii, jj] = R[ii, jj]; T[ii, 3] = t[ii]; } projectors[i].pose = T; } int fveci = 0; // index into our fvec array foreach (var projector in projectors) { // T_WPk is inverse of T_PkW, projector pose var T_WPk = new Matrix(4, 4); T_WPk.Inverse(projector.pose); foreach (var camera in projector.calibrationPointSets.Keys) { var cameraPoints = projector.calibrationPointSets[camera].worldPointInliers; var projectorPoints = projector.calibrationPointSets[camera].imagePointInliers; // transforms camera to projector coordinates var T_CjW = camera.pose; var T_CjPk = new Matrix(4, 4); T_CjPk.Mult(T_WPk, T_CjW); var cameraInProjector4 = new Matrix(4, 1); cameraInProjector4[3] = 1; var cameraPoint4 = new Matrix(4, 1); cameraPoint4[3] = 1; for (int i = 0; i < cameraPoints.Count; i++) { var cameraPoint = cameraPoints[i]; cameraPoint4[0] = cameraPoint[0]; cameraPoint4[1] = cameraPoint[1]; cameraPoint4[2] = cameraPoint[2]; cameraInProjector4.Mult(T_CjPk, cameraPoint4); cameraInProjector4.Scale(1.0 / cameraInProjector4[3]); // fvec_i = y_i - p_k( T_CjPk x_i ); double u, v; CameraMath.Project(projector.cameraMatrix, projector.lensDistortion, cameraInProjector4[0], cameraInProjector4[1], cameraInProjector4[2], out u, out v); var projectorPoint = projectorPoints[i]; fvec[fveci++] = projectorPoint.X - u; fvec[fveci++] = projectorPoint.Y - v; } } } //double sum = 0; //for (int i = 0; i < nValues; i++) // sum += fvec[i] * fvec[i]; //double rms = Math.Sqrt(sum / (double)nValues); //Console.WriteLine("in functor, rms == " + rms); return fvec; }; // TODO: maybe compute error before final optimization var calibrate = new LevenbergMarquardt(optimize); calibrate.minimumReduction = 1.0e-4; while (calibrate.State == LevenbergMarquardt.States.Running) { double rmsError = calibrate.MinimizeOneStep(parameters); Console.WriteLine("rms error = " + rmsError); } //for (int i = 0; i < nParameters; i++) // Console.WriteLine(parameters[i] + "\t"); //Console.WriteLine(); // convert back to transforms and put back in our structures { int pi = 0; // index into our parameter array for (int i = 1; i < cameras.Count; i++) // skip first one, which is our root { var r = new Matrix(3, 1); r[0] = parameters[pi++]; r[1] = parameters[pi++]; r[2] = parameters[pi++]; var R = CameraMath.RotationMatrixFromRotationVector(r); var t = new Matrix(3, 1); t[0] = parameters[pi++]; t[1] = parameters[pi++]; t[2] = parameters[pi++]; var T = new Matrix(4, 4); T.Identity(); for (int ii = 0; ii < 3; ii++) { for (int jj = 0; jj < 3; jj++) T[ii, jj] = R[ii, jj]; T[ii, 3] = t[ii]; } cameras[i].pose = T; } for (int i = 0; i < projectors.Count; i++) { var r = new Matrix(3, 1); r[0] = parameters[pi++]; r[1] = parameters[pi++]; r[2] = parameters[pi++]; var R = CameraMath.RotationMatrixFromRotationVector(r); var t = new Matrix(3, 1); t[0] = parameters[pi++]; t[1] = parameters[pi++]; t[2] = parameters[pi++]; var T = new Matrix(4, 4); T.Identity(); for (int ii = 0; ii < 3; ii++) { for (int jj = 0; jj < 3; jj++) T[ii, jj] = R[ii, jj]; T[ii, 3] = t[ii]; } projectors[i].pose = T; } } Console.WriteLine("elapsed time " + stopWatch.ElapsedMilliseconds); }
// Use DLT to obtain estimate of calibration rig pose; in our case this is the pose of the Kinect camera. // This pose estimate will provide a good initial estimate for subsequent projector calibration. // Note for a full PnP solution we should probably refine with Levenberg-Marquardt. // DLT is described in Hartley and Zisserman p. 178 public static void DLT(Matrix cameraMatrix, Matrix distCoeffs, List <Matrix> worldPoints, List <System.Drawing.PointF> imagePoints, out Matrix R, out Matrix t) { int n = worldPoints.Count; var A = Matrix.Zero(2 * n, 12); for (int j = 0; j < n; j++) { var X = worldPoints[j]; var imagePoint = imagePoints[j]; double x, y; Undistort(cameraMatrix, distCoeffs, imagePoint.X, imagePoint.Y, out x, out y); int ii = 2 * j; A[ii, 4] = -X[0]; A[ii, 5] = -X[1]; A[ii, 6] = -X[2]; A[ii, 7] = -1; A[ii, 8] = y * X[0]; A[ii, 9] = y * X[1]; A[ii, 10] = y * X[2]; A[ii, 11] = y; ii++; // next row A[ii, 0] = X[0]; A[ii, 1] = X[1]; A[ii, 2] = X[2]; A[ii, 3] = 1; A[ii, 8] = -x * X[0]; A[ii, 9] = -x * X[1]; A[ii, 10] = -x * X[2]; A[ii, 11] = -x; } // Pcolumn is the eigenvector of ATA with the smallest eignvalue var Pcolumn = new Matrix(12, 1); { var ATA = new Matrix(12, 12); ATA.MultATA(A, A); var V = new Matrix(12, 12); var ww = new Matrix(12, 1); ATA.Eig(V, ww); Pcolumn.CopyCol(V, 0); } // reshape into 3x4 projection matrix var P = new Matrix(3, 4); P.Reshape(Pcolumn); R = new Matrix(3, 3); for (int i = 0; i < 3; i++) { for (int j = 0; j < 3; j++) { R[i, j] = P[i, j]; } } if (R.Det3x3() < 0) { R.Scale(-1); P.Scale(-1); } // orthogonalize R { var U = new Matrix(3, 3); var V = new Matrix(3, 3); var ww = new Matrix(3, 1); R.SVD(U, ww, V); R.MultAAT(U, V); } // determine scale factor var RP = new Matrix(3, 3); for (int i = 0; i < 3; i++) { for (int j = 0; j < 3; j++) { RP[i, j] = P[i, j]; } } double s = RP.Norm() / R.Norm(); t = new Matrix(3, 1); for (int i = 0; i < 3; i++) { t[i] = P[i, 3]; } t.Scale(1.0 / s); }
// Use DLT to obtain estimate of calibration rig pose; in our case this is the pose of the Kinect camera. // This pose estimate will provide a good initial estimate for subsequent projector calibration. // Note for a full PnP solution we should probably refine with Levenberg-Marquardt. // DLT is described in Hartley and Zisserman p. 178 public static void DLT(Matrix cameraMatrix, Matrix distCoeffs, List <Matrix> worldPoints, List <System.Drawing.PointF> imagePoints, out Matrix R, out Matrix t) { int n = worldPoints.Count; var A = Matrix.Zero(2 * n, 12); for (int j = 0; j < n; j++) { var X = worldPoints[j]; var imagePoint = imagePoints[j]; double x, y; Undistort(cameraMatrix, distCoeffs, imagePoint.X, imagePoint.Y, out x, out y); double w = 1; int ii = 2 * j; A[ii, 4] = -w * X[0]; A[ii, 5] = -w * X[1]; A[ii, 6] = -w * X[2]; A[ii, 7] = -w; A[ii, 8] = y * X[0]; A[ii, 9] = y * X[1]; A[ii, 10] = y * X[2]; A[ii, 11] = y; ii++; // next row A[ii, 0] = w * X[0]; A[ii, 1] = w * X[1]; A[ii, 2] = w * X[2]; A[ii, 3] = w; A[ii, 8] = -x * X[0]; A[ii, 9] = -x * X[1]; A[ii, 10] = -x * X[2]; A[ii, 11] = -x; } var Pcolumn = new Matrix(12, 1); { var U = new Matrix(2 * n, 2 * n); // full SVD, alas, supports small number of points var V = new Matrix(12, 12); var ww = new Matrix(12, 1); A.SVD(U, ww, V); // find smallest singular value int min = 0; ww.Minimum(ref min); // Pcolumn is last column of V Pcolumn.CopyCol(V, min); } // reshape into 3x4 projection matrix var P = new Matrix(3, 4); P.Reshape(Pcolumn); // x = P * X // P = K [ R | t ] // inv(K) P = [ R | t ] //var Kinv = new Matrix(3, 3); //Kinv.Inverse(cameraMatrix); //var Rt = new Matrix(3, 4); //Rt.Mult(Kinv, P); var Rt = new Matrix(3, 4); Rt.Copy(P); // P does not contain camera matrix (by earlier undistort) R = new Matrix(3, 3); t = new Matrix(3, 1); for (int ii = 0; ii < 3; ii++) { t[ii] = Rt[ii, 3]; for (int jj = 0; jj < 3; jj++) { R[ii, jj] = Rt[ii, jj]; } } //R.Copy(0, 0, Rt); //t.CopyCol(Rt, 3); if (R.Det3x3() < 0) { R.Scale(-1); t.Scale(-1); } // orthogonalize R { var U = new Matrix(3, 3); var Vt = new Matrix(3, 3); var V = new Matrix(3, 3); var ww = new Matrix(3, 1); R.SVD(U, ww, V); Vt.Transpose(V); R.Mult(U, Vt); double s = ww.Sum() / 3.0; t.Scale(1.0 / s); } // compute error? }
static public void PlaneFit(IList <Matrix> X, out Matrix R, out Matrix t, out Matrix d2) { int n = X.Count; var mu = new Matrix(3, 1); for (int i = 0; i < n; i++) { mu.Add(X[i]); } mu.Scale(1f / (float)n); var A = new Matrix(3, 3); var xc = new Matrix(3, 1); var M = new Matrix(3, 3); for (int i = 0; i < X.Count; i++) { var x = X[i]; xc.Sub(x, mu); M.Outer(xc, xc); A.Add(M); } var V = new Matrix(3, 3); var d = new Matrix(3, 1); A.Eig(V, d); // eigenvalues in ascending order // arrange in descending order so that z = 0 var V2 = new Matrix(3, 3); for (int i = 0; i < 3; i++) { V2[i, 2] = V[i, 0]; V2[i, 1] = V[i, 1]; V2[i, 0] = V[i, 2]; } d2 = new Matrix(3, 1); d2[2] = d[0]; d2[1] = d[1]; d2[0] = d[2]; R = new Matrix(3, 3); R.Transpose(V2); if (R.Det3x3() < 0) { R.Scale(-1); } t = new Matrix(3, 1); t.Mult(R, mu); t.Scale(-1); // eigenvalues are the sum of squared distances in each direction // i.e., min eigenvalue is the sum of squared distances to the plane = d2[2] // compute the distance to the plane by transforming to the plane and take z-coordinate: // xPlane = R*x + t; distance = xPlane[2] }
public static Matrix Homography(List <Matrix> worldPoints, List <System.Drawing.PointF> imagePoints) { int n = worldPoints.Count; // normalize image coordinates var mu = new Matrix(2, 1); for (int i = 0; i < n; i++) { mu[0] += imagePoints[i].X; mu[1] += imagePoints[i].Y; } mu.Scale(1.0 / n); var muAbs = new Matrix(2, 1); for (int i = 0; i < n; i++) { muAbs[0] += Math.Abs(imagePoints[i].X - mu[0]); muAbs[1] += Math.Abs(imagePoints[i].Y - mu[1]); } muAbs.Scale(1.0 / n); var Hnorm = Matrix.Identity(3, 3); Hnorm[0, 0] = 1 / muAbs[0]; Hnorm[1, 1] = 1 / muAbs[1]; Hnorm[0, 2] = -mu[0] / muAbs[0]; Hnorm[1, 2] = -mu[1] / muAbs[1]; var invHnorm = Matrix.Identity(3, 3); invHnorm[0, 0] = muAbs[0]; invHnorm[1, 1] = muAbs[1]; invHnorm[0, 2] = mu[0]; invHnorm[1, 2] = mu[1]; var A = Matrix.Zero(2 * n, 9); for (int i = 0; i < n; i++) { var X = worldPoints[i]; var imagePoint = imagePoints[i]; var x = new Matrix(3, 1); x[0] = imagePoint.X; x[1] = imagePoint.Y; x[2] = 1; var xn = new Matrix(3, 1); xn.Mult(Hnorm, x); // Zhang's formulation; Hartley's is similar int ii = 2 * i; A[ii, 0] = X[0]; A[ii, 1] = X[1]; A[ii, 2] = 1; A[ii, 6] = -xn[0] * X[0]; A[ii, 7] = -xn[0] * X[1]; A[ii, 8] = -xn[0]; ii++; // next row A[ii, 3] = X[0]; A[ii, 4] = X[1]; A[ii, 5] = 1; A[ii, 6] = -xn[1] * X[0]; A[ii, 7] = -xn[1] * X[1]; A[ii, 8] = -xn[1]; } // h is the eigenvector of ATA with the smallest eignvalue var h = new Matrix(9, 1); { var ATA = new Matrix(9, 9); ATA.MultATA(A, A); var V = new Matrix(9, 9); var ww = new Matrix(9, 1); ATA.Eig(V, ww); h.CopyCol(V, 0); } var Hn = new Matrix(3, 3); Hn.Reshape(h); var H = new Matrix(3, 3); H.Mult(invHnorm, Hn); return(H); }
public static void Normalize(Matrix A, Matrix B) { B.Scale(A, 1.0/A.Norm()); }
public void RecoverCalibrationFromSensor(KinectSensor kinectSensor) { var stopWatch = new System.Diagnostics.Stopwatch(); stopWatch.Start(); var objectPoints1 = new List <RoomAliveToolkit.Matrix>(); var colorPoints1 = new List <System.Drawing.PointF>(); var depthPoints1 = new List <System.Drawing.PointF>(); int n = 0; for (float x = -2f; x < 2f; x += 0.2f) { for (float y = -2f; y < 2f; y += 0.2f) { for (float z = 0.4f; z < 4.5f; z += 0.4f) { var kinectCameraPoint = new CameraSpacePoint(); kinectCameraPoint.X = x; kinectCameraPoint.Y = y; kinectCameraPoint.Z = z; // use SDK's projection // adjust Y to make RH cooridnate system that is a projection of Kinect 3D points var kinectColorPoint = kinectSensor.CoordinateMapper.MapCameraPointToColorSpace(kinectCameraPoint); kinectColorPoint.Y = colorImageHeight - kinectColorPoint.Y; var kinectDepthPoint = kinectSensor.CoordinateMapper.MapCameraPointToDepthSpace(kinectCameraPoint); kinectDepthPoint.Y = depthImageHeight - kinectDepthPoint.Y; if ((kinectColorPoint.X >= 0) && (kinectColorPoint.X < colorImageWidth) && (kinectColorPoint.Y >= 0) && (kinectColorPoint.Y < colorImageHeight) && (kinectDepthPoint.X >= 0) && (kinectDepthPoint.X < depthImageWidth) && (kinectDepthPoint.Y >= 0) && (kinectDepthPoint.Y < depthImageHeight)) { n++; var objectPoint = new RoomAliveToolkit.Matrix(3, 1); objectPoint[0] = kinectCameraPoint.X; objectPoint[1] = kinectCameraPoint.Y; objectPoint[2] = kinectCameraPoint.Z; objectPoints1.Add(objectPoint); var colorPoint = new System.Drawing.PointF(); colorPoint.X = kinectColorPoint.X; colorPoint.Y = kinectColorPoint.Y; colorPoints1.Add(colorPoint); //Console.WriteLine(objectPoint[0] + "\t" + objectPoint[1] + "\t" + colorPoint.X + "\t" + colorPoint.Y); var depthPoint = new System.Drawing.PointF(); depthPoint.X = kinectDepthPoint.X; depthPoint.Y = kinectDepthPoint.Y; depthPoints1.Add(depthPoint); } } } } colorCameraMatrix[0, 0] = 1000; //fx colorCameraMatrix[1, 1] = 1000; //fy colorCameraMatrix[0, 2] = colorImageWidth / 2; //cx colorCameraMatrix[1, 2] = colorImageHeight / 2; //cy colorCameraMatrix[2, 2] = 1; var rotation = new Matrix(3, 1); var translation = new Matrix(3, 1); var colorError = CalibrateColorCamera(objectPoints1, colorPoints1, colorCameraMatrix, colorLensDistortion, rotation, translation); //var rotationMatrix = Orientation.Rodrigues(rotation); var rotationMatrix = RoomAliveToolkit.ProjectorCameraEnsemble.RotationMatrixFromRotationVector(rotation); depthToColorTransform = Matrix.Identity(4, 4); for (int i = 0; i < 3; i++) { depthToColorTransform[i, 3] = translation[i]; for (int j = 0; j < 3; j++) { depthToColorTransform[i, j] = rotationMatrix[i, j]; } } depthCameraMatrix[0, 0] = 360; //fx depthCameraMatrix[1, 1] = 360; //fy depthCameraMatrix[0, 2] = depthImageWidth / 2; //cx depthCameraMatrix[1, 2] = depthImageHeight / 2; //cy depthCameraMatrix[2, 2] = 1; var depthError = CalibrateDepthCamera(objectPoints1, depthPoints1, depthCameraMatrix, depthLensDistortion); //// latest SDK gives access to depth intrinsics directly -- this gives slightly higher projection error; not sure why //var depthIntrinsics = kinectSensor.CoordinateMapper.GetDepthCameraIntrinsics(); //depthCameraMatrix[0, 0] = depthIntrinsics.FocalLengthX; //depthCameraMatrix[1, 1] = depthIntrinsics.FocalLengthY; //depthCameraMatrix[0, 2] = depthIntrinsics.PrincipalPointX; //depthCameraMatrix[1, 2] = depthImageHeight - depthIntrinsics.PrincipalPointY; // note flip in Y! //depthDistCoeffs[0] = depthIntrinsics.RadialDistortionSecondOrder; //depthDistCoeffs[1] = depthIntrinsics.RadialDistortionFourthOrder; // check projections double depthProjectionError = 0; double colorProjectionError = 0; var color = new RoomAliveToolkit.Matrix(4, 1); var testObjectPoint4 = new RoomAliveToolkit.Matrix(4, 1); for (int i = 0; i < n; i++) { var testObjectPoint = objectPoints1[i]; var testDepthPoint = depthPoints1[i]; var testColorPoint = colorPoints1[i]; // "camera space" == depth camera space // depth camera projection double depthU, depthV; CameraMath.Project(depthCameraMatrix, depthLensDistortion, testObjectPoint[0], testObjectPoint[1], testObjectPoint[2], out depthU, out depthV); double dx = testDepthPoint.X - depthU; double dy = testDepthPoint.Y - depthV; depthProjectionError += (dx * dx) + (dy * dy); // color camera projection testObjectPoint4[0] = testObjectPoint[0]; testObjectPoint4[1] = testObjectPoint[1]; testObjectPoint4[2] = testObjectPoint[2]; testObjectPoint4[3] = 1; color.Mult(depthToColorTransform, testObjectPoint4); color.Scale(1.0 / color[3]); // not necessary for this transform double colorU, colorV; CameraMath.Project(colorCameraMatrix, colorLensDistortion, color[0], color[1], color[2], out colorU, out colorV); dx = testColorPoint.X - colorU; dy = testColorPoint.Y - colorV; colorProjectionError += (dx * dx) + (dy * dy); } depthProjectionError /= n; colorProjectionError /= n; stopWatch.Stop(); Console.WriteLine("FakeCalibration :"); Console.WriteLine("n = " + n); Console.WriteLine("color error = " + colorError); Console.WriteLine("depth error = " + depthError); Console.WriteLine("depth reprojection error = " + depthProjectionError); Console.WriteLine("color reprojection error = " + colorProjectionError); Console.WriteLine("depth camera matrix = \n" + depthCameraMatrix); Console.WriteLine("depth lens distortion = \n" + depthLensDistortion); Console.WriteLine("color camera matrix = \n" + colorCameraMatrix); Console.WriteLine("color lens distortion = \n" + colorLensDistortion); Console.WriteLine(stopWatch.ElapsedMilliseconds + " ms"); //// get camera space table //// this does not change frame to frame (or so I believe) //var tableEntries = kinectSensor.CoordinateMapper.GetDepthFrameToCameraSpaceTable(); //// compute our own version of the camera space table and compare it to the SDK's //stopWatch.Restart(); //var tableEntries2 = ComputeDepthFrameToCameraSpaceTable(); //Console.WriteLine("ComputeDepthFrameToCameraSpaceTable took " + stopWatch.ElapsedMilliseconds + " ms"); //{ // float error = 0; // for (int framey = 0; framey < depthImageHeight; framey++) // for (int framex = 0; framex < depthImageWidth; framex++) // { // var point1 = tableEntries[depthImageWidth * framey + framex]; // var point2 = tableEntries2[depthImageWidth * framey + framex]; // error += (float)Math.Sqrt((point1.X - point2.X) * (point1.X - point2.X) + (point1.Y - point2.Y) * (point1.Y - point2.Y)); // } // error /= (float)(depthImageHeight * depthImageWidth); // Console.WriteLine("error = " + error); //} }
public void RecoverCalibrationFromSensor(KinectSensor kinectSensor) { var stopWatch = new System.Diagnostics.Stopwatch(); stopWatch.Start(); var objectPoints1 = new List<RoomAliveToolkit.Matrix>(); var colorPoints1 = new List<System.Drawing.PointF>(); var depthPoints1 = new List<System.Drawing.PointF>(); int n = 0; for (float x = -2f; x < 2f; x += 0.2f) for (float y = -2f; y < 2f; y += 0.2f) for (float z = 0.4f; z < 4.5f; z += 0.4f) { var kinectCameraPoint = new CameraSpacePoint(); kinectCameraPoint.X = x; kinectCameraPoint.Y = y; kinectCameraPoint.Z = z; // use SDK's projection // adjust Y to make RH cooridnate system that is a projection of Kinect 3D points var kinectColorPoint = kinectSensor.CoordinateMapper.MapCameraPointToColorSpace(kinectCameraPoint); kinectColorPoint.Y = colorImageHeight - kinectColorPoint.Y; var kinectDepthPoint = kinectSensor.CoordinateMapper.MapCameraPointToDepthSpace(kinectCameraPoint); kinectDepthPoint.Y = depthImageHeight - kinectDepthPoint.Y; if ((kinectColorPoint.X >= 0) && (kinectColorPoint.X < colorImageWidth) && (kinectColorPoint.Y >= 0) && (kinectColorPoint.Y < colorImageHeight) && (kinectDepthPoint.X >= 0) && (kinectDepthPoint.X < depthImageWidth) && (kinectDepthPoint.Y >= 0) && (kinectDepthPoint.Y < depthImageHeight)) { n++; var objectPoint = new RoomAliveToolkit.Matrix(3, 1); objectPoint[0] = kinectCameraPoint.X; objectPoint[1] = kinectCameraPoint.Y; objectPoint[2] = kinectCameraPoint.Z; objectPoints1.Add(objectPoint); var colorPoint = new System.Drawing.PointF(); colorPoint.X = kinectColorPoint.X; colorPoint.Y = kinectColorPoint.Y; colorPoints1.Add(colorPoint); //Console.WriteLine(objectPoint[0] + "\t" + objectPoint[1] + "\t" + colorPoint.X + "\t" + colorPoint.Y); var depthPoint = new System.Drawing.PointF(); depthPoint.X = kinectDepthPoint.X; depthPoint.Y = kinectDepthPoint.Y; depthPoints1.Add(depthPoint); } } colorCameraMatrix[0, 0] = 1000; //fx colorCameraMatrix[1, 1] = 1000; //fy colorCameraMatrix[0, 2] = colorImageWidth / 2; //cx colorCameraMatrix[1, 2] = colorImageHeight / 2; //cy colorCameraMatrix[2, 2] = 1; var rotation = new Matrix(3, 1); var translation = new Matrix(3, 1); var colorError = CalibrateColorCamera(objectPoints1, colorPoints1, colorCameraMatrix, colorLensDistortion, rotation, translation); //var rotationMatrix = Orientation.Rodrigues(rotation); var rotationMatrix = RoomAliveToolkit.ProjectorCameraEnsemble.RotationMatrixFromRotationVector(rotation); depthToColorTransform = Matrix.Identity(4, 4); for (int i = 0; i < 3; i++) { depthToColorTransform[i, 3] = translation[i]; for (int j = 0; j < 3; j++) depthToColorTransform[i, j] = rotationMatrix[i, j]; } depthCameraMatrix[0, 0] = 360; //fx depthCameraMatrix[1, 1] = 360; //fy depthCameraMatrix[0, 2] = depthImageWidth / 2; //cx depthCameraMatrix[1, 2] = depthImageHeight / 2; //cy depthCameraMatrix[2, 2] = 1; var depthError = CalibrateDepthCamera(objectPoints1, depthPoints1, depthCameraMatrix, depthLensDistortion); //// latest SDK gives access to depth intrinsics directly -- this gives slightly higher projection error; not sure why //var depthIntrinsics = kinectSensor.CoordinateMapper.GetDepthCameraIntrinsics(); //depthCameraMatrix[0, 0] = depthIntrinsics.FocalLengthX; //depthCameraMatrix[1, 1] = depthIntrinsics.FocalLengthY; //depthCameraMatrix[0, 2] = depthIntrinsics.PrincipalPointX; //depthCameraMatrix[1, 2] = depthImageHeight - depthIntrinsics.PrincipalPointY; // note flip in Y! //depthDistCoeffs[0] = depthIntrinsics.RadialDistortionSecondOrder; //depthDistCoeffs[1] = depthIntrinsics.RadialDistortionFourthOrder; // check projections double depthProjectionError = 0; double colorProjectionError = 0; var color = new RoomAliveToolkit.Matrix(4, 1); var testObjectPoint4 = new RoomAliveToolkit.Matrix(4, 1); for (int i = 0; i < n; i++) { var testObjectPoint = objectPoints1[i]; var testDepthPoint = depthPoints1[i]; var testColorPoint = colorPoints1[i]; // "camera space" == depth camera space // depth camera projection double depthU, depthV; CameraMath.Project(depthCameraMatrix, depthLensDistortion, testObjectPoint[0], testObjectPoint[1], testObjectPoint[2], out depthU, out depthV); double dx = testDepthPoint.X - depthU; double dy = testDepthPoint.Y - depthV; depthProjectionError += (dx * dx) + (dy * dy); // color camera projection testObjectPoint4[0] = testObjectPoint[0]; testObjectPoint4[1] = testObjectPoint[1]; testObjectPoint4[2] = testObjectPoint[2]; testObjectPoint4[3] = 1; color.Mult(depthToColorTransform, testObjectPoint4); color.Scale(1.0 / color[3]); // not necessary for this transform double colorU, colorV; CameraMath.Project(colorCameraMatrix, colorLensDistortion, color[0], color[1], color[2], out colorU, out colorV); dx = testColorPoint.X - colorU; dy = testColorPoint.Y - colorV; colorProjectionError += (dx * dx) + (dy * dy); } depthProjectionError /= n; colorProjectionError /= n; stopWatch.Stop(); Console.WriteLine("FakeCalibration :"); Console.WriteLine("n = " + n); Console.WriteLine("color error = " + colorError); Console.WriteLine("depth error = " + depthError); Console.WriteLine("depth reprojection error = " + depthProjectionError); Console.WriteLine("color reprojection error = " + colorProjectionError); Console.WriteLine("depth camera matrix = \n" + depthCameraMatrix); Console.WriteLine("depth lens distortion = \n" + depthLensDistortion); Console.WriteLine("color camera matrix = \n" + colorCameraMatrix); Console.WriteLine("color lens distortion = \n" + colorLensDistortion); Console.WriteLine(stopWatch.ElapsedMilliseconds + " ms"); //// get camera space table //// this does not change frame to frame (or so I believe) //var tableEntries = kinectSensor.CoordinateMapper.GetDepthFrameToCameraSpaceTable(); //// compute our own version of the camera space table and compare it to the SDK's //stopWatch.Restart(); //var tableEntries2 = ComputeDepthFrameToCameraSpaceTable(); //Console.WriteLine("ComputeDepthFrameToCameraSpaceTable took " + stopWatch.ElapsedMilliseconds + " ms"); //{ // float error = 0; // for (int framey = 0; framey < depthImageHeight; framey++) // for (int framex = 0; framex < depthImageWidth; framex++) // { // var point1 = tableEntries[depthImageWidth * framey + framex]; // var point2 = tableEntries2[depthImageWidth * framey + framex]; // error += (float)Math.Sqrt((point1.X - point2.X) * (point1.X - point2.X) + (point1.Y - point2.Y) * (point1.Y - point2.Y)); // } // error /= (float)(depthImageHeight * depthImageWidth); // Console.WriteLine("error = " + error); //} }
public static List<RoomAliveToolkit.Matrix> TransformPoints(RoomAliveToolkit.Matrix A, List<RoomAliveToolkit.Matrix> points) { var transformedPoints = new List<RoomAliveToolkit.Matrix>(); var point4 = new RoomAliveToolkit.Matrix(4, 1); point4[3] = 1; var transformedPoint4 = new RoomAliveToolkit.Matrix(4, 1); foreach (var point in points) { point4[0] = point[0]; point4[1] = point[1]; point4[2] = point[2]; transformedPoint4.Mult(A, point4); transformedPoint4.Scale(1.0f / transformedPoint4[3]); var transformedPoint = new RoomAliveToolkit.Matrix(3, 1); transformedPoint[0] = transformedPoint4[0]; transformedPoint[1] = transformedPoint4[1]; transformedPoint[2] = transformedPoint4[2]; transformedPoints.Add(transformedPoint); } return transformedPoints; }
public static double PlaneFit(IList<Matrix> points, out Matrix X, out double D) { X = new Matrix(3, 1); var mu = new RoomAliveToolkit.Matrix(3, 1); for (int i = 0; i < points.Count; i++) mu.Add(points[i]); mu.Scale(1f / (float)points.Count); var A = new RoomAliveToolkit.Matrix(3, 3); var pc = new RoomAliveToolkit.Matrix(3, 1); var M = new RoomAliveToolkit.Matrix(3, 3); for (int i = 0; i < points.Count; i++) { var p = points[i]; pc.Sub(p, mu); M.Outer(pc, pc); A.Add(M); } var V = new RoomAliveToolkit.Matrix(3, 3); var d = new RoomAliveToolkit.Matrix(3, 1); A.Eig(V, d); // TODO: replace with 3x3 version? //Console.WriteLine("------"); //Console.WriteLine(A); //Console.WriteLine(V); //Console.WriteLine(d); double minEigenvalue = Double.MaxValue; int minEigenvaluei = 0; for (int i = 0; i < 3; i++) if (d[i] < minEigenvalue) { minEigenvalue = d[i]; minEigenvaluei = i; } X.CopyCol(V, minEigenvaluei); D = -X.Dot(mu); // min eigenvalue is the sum of squared distances to the plane // signed distance is: double distance = X.Dot(point) + D; return minEigenvalue; }