I am working with stereo camera calibration using checkerboard.
I am using about 50 images in left and right, respectively.
In order to get the translation parameter, I use cv2.calibrateCamera function.
as follows : ret_1, k1, d1, r_1, t_1 = cv2.calibrateCamera(
threedpoints, twodpoints_c1,(width, height), None, None)
I get translation vector. in some code, the problem is that the parameter is restored in npz file format and scale shape of t_1 is N times 3 times 1, not 3 times 1 vector. I don’t know why. I need correct translation vector, and then I can estimate baseline for depth estimation. what can I do? translation tvecs of left camera is like this:
left translation matrix [[[ -1.45023029]
[-14.32631881]
[ 70.75464883]]
[[ 5.31966134]
[ -1.76426191]
[ 74.89211154]]
[[ -4.96704786]
[ -0.27187788]
[ 74.32857874]]
[[ -0.17176461]
[ 4.64189051]
[ 81.01646601]]
[[ 5.19483659]
[ 5.07308422]
[ 81.51979487]]
[[ 7.73597941]
[ -7.90230406]
[ 82.09726143]]
[[ 6.9505047 ]
[-16.38893344]
[ 82.70428372]]
[[ -4.40334536]
[-15.96919736]
[ 80.3917108 ]]
[[-18.62022825]
[-17.45662902]
[107.07112184]]
[[ -3.41950405]
[-19.73944268]
[109.79771039]]
[[ 9.09591381]
[-20.36718436]
[110.84339971]]
[[ 9.62038964]
[ -9.9015084 ]
[109.97886006]]
[[ -4.03804921]
[-10.06447627]
[108.10038068]]
[[-16.78069469]
[ -8.92345387]
[107.31875784]]
[[-20.89555805]
[ 2.93358763]
[107.21048845]]
[[ -6.87036094]
[ 4.41176443]
[109.31964222]]
[[ 9.52674567]
[ 3.95541506]
[111.97490365]]
[[ 10.70938144]
[ 1.68018021]
[113.04744195]]
[[ -3.89920638]
[ 1.63860167]
[109.41954086]]
[[-20.85657341]
[ 2.10307594]
[106.13929225]]
[[-23.04552747]
[-12.37606267]
[109.66673468]]
[[ -6.98387619]
[-14.35512823]
[110.55349138]]
[[ 12.90681553]
[-15.87215479]
[113.86482201]]
[[ 12.17722363]
[-29.70365156]
[113.028406 ]]
[[ -3.94408241]
[-31.70859421]
[108.93008251]]]