Poor accuracy of test_multi_noncentral_relative_pose_sac

[francescoli]

Hi,
@laurentkneip When I running test_multi_noncentral_relative_pose_sac, as shown below, the RANSAC result is quite different from the groundtruth. Could you tell me the cause of this problem?

the random position is:
-0.045566
 0.577806
  1.52469

the random rotation is:
 0.912585  -0.27358  0.303879
 0.352483  0.903025 -0.245563
-0.207229  0.331209   0.92052

the noise in the data is:
0.3
the outlier fraction is:
0.1
the ransac threshold is: 7.8125e-07
the ransac results is: 
 -0.639841   0.713275    0.28608    1.82647
 -0.142771    0.25545  -0.956223   0.310284
 -0.755129  -0.652675 -0.0616126  -0.112605

Ransac needed 101 iterations and 0.219001 seconds

the number of inliers is: 5

the found inliers are: 
2 15 9 13 17