Why does the computed diagonal FOV of the VI-sensor differ from its nominal value?

[JzHuai0108]

I have a quick question about the diagonal FOV of the VI-sensor. From the euroc dataset, I find the focal length in pixel and horizontal and vertical length of the image, so the diagonal focal length is computed as the below python code.

# visensor, see Nikolic 14 paper A Synchronized Visual-Inertial Sensor System with FPGA

# Pre-Processing for Accurate Real-Time SLAM

from __future__ import print_function

import math

import numpy as np



print('VI-sensor', '*'*60)

w = 752

h = 480

d = math.sqrt(w*w + h*h)

focalpixel = 457.0 # approximated from euroc dataset MH_01_easy

actualdfov = 2*math.atan2(d/2, focalpixel)*180/math.pi

print('Diagonal FOV computed from calibrated focal length in pixel {}: {} deg'.format(focalpixel, actualdfov))

The above code gives the result:

Diagonal FOV computed from calibrated focal length in pixel 457.0: 88.612 deg

However the Nikolic ICRA14 paper mentions that the diagonal FOV of the VI-sensor lenses are either 122 or 132 deg.

Can you please clarify this difference on the diagonal FOV for me? Thanks in advance.

[ZacharyTaylor]

I think they played around a lot with different lenses when the vi-sensor was first developed and so the lenses for the EuRoC dataset might be different than the ones in the first paper.

[JzHuai0108]

Alright. I just want to make sure that my method to compute FOV is correct.