Rectify maps

[jprodriguezg]

Hello, thanks for the cool dataset!

I am wondering how the maps to rectify distorted pixels were computed. I am trying to obtain the same results using cv2.fisheye.undistortPoints as suggested here. However, the results are different from those provided in $SEQUENCE_(left/right)_(x/y)_map.txt.

For instance, these are some results:

Pixels to rectify

[[[0. 0.]
[0. 1.]
[0. 2.]]]

Output from indoor_flying_left_(left/right)_(x/y)_map.txt

[[[-106.6652 -83.8143]
[-102.6674 -82.0455]
[ -98.7668 -80.3292]]]

Output from cv2.fisheye.undistortPoints

[[[-143.5242 -110.53386]
[-141.45755 -107.12762]
[-139.44424 -103.78565]]]

Could someone please explain how these maps were computed?

Thanks!

[vbrebion]

Hi!

In case it might help anyone, here is how I was able to get the same values as the provided maps:

import cv2
import numpy as np
import yaml

with open("/path/to/camchain-imucam-indoor_flying.yaml", "r") as stream:
  yaml_data = yaml.safe_load(stream)
  davis_intrinsics = np.array(yaml_data["cam0"]["intrinsics"])
  davis_dist = np.array(yaml_data["cam0"]["distortion_coeffs"])
  davis_rect = np.array(yaml_data["cam0"]["rectification_matrix"])
  davis_proj = np.array(yaml_data["cam0"]["projection_matrix"])

K = np.array([[davis_intrinsics[0], 0, davis_intrinsics[2]],
              [0, davis_intrinsics[1], davis_intrinsics[3]],
              [0, 0, 1]])

pts_to_rectify = np.array([[[0., 0.],
                            [0., 1.],
                            [0., 2.]]])

rectified_pts = cv2.fisheye.undistortPoints(pts_to_rectify, K, davis_dist, None, davis_rect, davis_proj)
print(rectified_pts)