Equation(5.7) page 82:
Jacobian of p evaluated at the origin (alpha = 0,belt = 0, r = 0), which means J_p is always a constant matrix at every point. In this case, i think we cant just set gaussian noise(mean = 0, scale = 4) for image points.
Because if we use gaussian noise to set image points, the value of covariance matrices V_i is always similar(because of the same variance). This leads to equation (5.8) covariance matrices of alpha,belt, r also get a similar value.This is not correct.
So im thinking the noise of the 4 image points should be different at different cam pose, which means the covariance matrices V_i is different at different cam pose.
So my question is: In the real case, the noise of the image points should be different at different cam pose, right? And can we make a meeting next week before 08.03.2018?
Paper : Tracking Errors in Augmented Reality
Equation(5.7) page 82: Jacobian of p evaluated at the origin (alpha = 0,belt = 0, r = 0), which means J_p is always a constant matrix at every point. In this case, i think we cant just set gaussian noise(mean = 0, scale = 4) for image points. Because if we use gaussian noise to set image points, the value of covariance matrices V_i is always similar(because of the same variance). This leads to equation (5.8) covariance matrices of alpha,belt, r also get a similar value.This is not correct.
So im thinking the noise of the 4 image points should be different at different cam pose, which means the covariance matrices V_i is different at different cam pose.
So my question is: In the real case, the noise of the image points should be different at different cam pose, right? And can we make a meeting next week before 08.03.2018?