Hi Nghia, excellent work. I have a qtesuion. I have implemented your code in combination with colour segmentation based point tracking. I am tracking a 100x100mm square, with the corners indicated by coloured dots, using a HD webcam. I have defined the (planar) object as follows:Object model:// Object model// First row holds x-coordinates, second row holds corresponding y-coordinatesdouble modelPoints_data[8] = { -0.05, 0.05, -0.05, 0.05, 0.05, 0.05, -0.05, -0.05};.. Later .. // Insert data into RPP model input matrix for(int i=0; i < 4; i++) { modelPoints.at(0,i) = modelPoints_data[i]; // x modelPoints.at(1,i) = modelPoints_data[i+4]; // y }E.g. similar to your demo code. You see the square center is the origin of the coordinate system, and the points as follows: ___________| 1 2 || || ||_3______4_|With the program running, I output the modelPoints and imagePoints matrices (the latter holds the normalized image points, using a camera matrix equal to your example of 60 deg FoV, which seems to be accurate for this camera):objectPoints:-0.050000 0.050000 -0.050000 0.0500000.050000 0.050000 -0.050000 -0.0500000.000000 0.000000 0.000000 0.000000 imagePoints:-0.049345 0.114326 -0.040100 0.123716-0.042747 -0.052067 0.119452 0.1123521.000000 1.000000 1.000000 1.000000 The object is in this case straight in front of the camera, and I manually select the points by mouse in the order as they are defined in the object model. Thus I would expect near-zero x and y translation, say 0.3m z-translation, and an near-identity rotation matrix.The resulting translation vector seems to be very accurate according to initial measurements, however I am somewhat puzzled by the rotation matrix:Rotation matrix:0.994831 -0.056219 -0.084559-0.054371 -0.998232 0.024007-0.085760 -0.019286 -0.996129This is not approximately an identity matrix as I expected, as two of the diagonals are -1. The values correspond to a rotation of +- pi radians about the x-axis of the object. If I select the points in wrong order, that corresponds to this rotation, I do obtain the expected identity matrix.Do you have any thoughts on this?

online cialis nexium