mrcal-show-projection-uncertainty - Visualize the expected projection error due to noise in calibration-time input
$ mrcal-show-projection-uncertainty left.cameramodel
... a plot pops up showing the projection uncertainty of the intrinsics in
... this modelThe operation of this tool is documented at http://mrcal.secretsauce.net/uncertainty.html
A calibration process produces the best-fitting camera parameters. To be able to use these parameters we must know how trustworthy they are. This tool examines the uncertainty of projection using a given camera model. The projection operation uses the intrinsics only, but the uncertainty must take into account the calibration-time extrinsics and the calibration-time observed object poses as well. This tool visualizes the expected value of projection error across the imager. Areas with a high expected projection error are unreliable, and observations in those regions cannot be used for further work (localization, mapping, etc).
There are several modes of operation:
- By default we look at projection of points some distance away from the camera (given by --distance). We evaluate the uncertainty of these projections everywhere across the imager, and display the results as a heatmap with overlaid contours
- With --vs-distance-at we evaluate the uncertainty along an observation ray mapping to a single pixel. We show the uncertainty vs distances from the camera along this ray
See http://mrcal.secretsauce.net/uncertainty.html for a full description of the computation performed here
model Input camera model. If "-' is given, we read standard
input-h, --help show this help message and exit
--vs-distance-at VS_DISTANCE_AT
If given, we don't compute the uncertainty everywhere
in the image at a constant distance from the camera,
but instead we look at different distances at one
pixel. This option takes a single argument: the "X,Y"
pixel coordinate we care about, or "center" to look at
the center of the imager or "centroid" to look at the
center of the calibration-time chessboards. This is
exclusive with --gridn and --distance and
--observations and --cbmax
--gridn GRIDN GRIDN How densely we should sample the imager. By default we
use a 60x40 grid
--distance DISTANCE By default we display the projection uncertainty
infinitely far away from the camera. If we want to
look closer in, the desired observation distance can
be given in this argument
--isotropic By default I display the expected value of the
projection error in the worst possible direction of
this error. If we want to plot the RMS of the worst
and best directions, pass --isotropic. If we assume
the errors will apply evenly in all directions, then
we can use this metric, which is potentially easier to
compute
--method {mean-frames,cross-reprojection--rrp-Jfp}
Multiple uncertainty quantification methods are
available. We default to 'mean-frames'
--observations If given, I display the pixel observations at
calibration time. This should correspond to the low-
uncertainty regions.
--valid-intrinsics-region
If given, I overlay the valid-intrinsics region onto
the plot
--cbmax CBMAX Maximum range of the colorbar
--title TITLE Title string for the plot. Overrides the default
title. Exclusive with --extratitle
--extratitle EXTRATITLE
Additional string for the plot to append to the
default title. Exclusive with --title
--hardcopy HARDCOPY Write the output to disk, instead of an interactive
plot
--terminal TERMINAL gnuplotlib terminal. The default is good almost
always, so most people don't need this option
--set SET Extra 'set' directives to gnuplotlib. Can be given
multiple times
--unset UNSET Extra 'unset' directives to gnuplotlib. Can be given
multiple timeshttps://www.github.com/dkogan/mrcal
Dima Kogan, <dima@secretsauce.net>
Copyright (c) 2017-2023 California Institute of Technology ("Caltech"). U.S. Government sponsorship acknowledged. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0