These are chat archives for l0g1x/DUO-Camera-ROS

17th
Aug 2015
Krystian Gebis
@l0g1x
Aug 17 2015 21:51
so i pretty much have it done, im just trying to figure out what the duo intrinsic, extrinsic, rotational, and translation matricies are equivilant to in the camera_info documentation
gitunit
@gitunit
Aug 17 2015 22:15
ok. how about the information provided by sai?
Krystian Gebis
@l0g1x
Aug 17 2015 22:16
forgot about that, looking at that now
gitunit
@gitunit
Aug 17 2015 22:18
also important would be how matrices are read in ROS msgs, column-wise or row-wise? i'll look that up
Krystian Gebis
@l0g1x
Aug 17 2015 22:18
row major
gitunit
@gitunit
Aug 17 2015 22:19
ok, the same with DUO?
Krystian Gebis
@l0g1x
Aug 17 2015 22:20
not sure actually, didnt seem to run across it actually.
gitunit
@gitunit
Aug 17 2015 22:21
ok, lets asume its the same
Krystian Gebis
@l0g1x
Aug 17 2015 22:21
for now yeah
ill double check with duo
gitunit
@gitunit
Aug 17 2015 22:22
do you have info on the distortion model?
Krystian Gebis
@l0g1x
Aug 17 2015 22:23
i messaged them about it, assuming they have something proprietary
didnt get back to me yet
gitunit
@gitunit
Aug 17 2015 22:24
i think they used the same as in openCV, i'll search for it
Krystian Gebis
@l0g1x
Aug 17 2015 22:24
even with the new circle calibration?
gitunit
@gitunit
Aug 17 2015 22:25
yes
Krystian Gebis
@l0g1x
Aug 17 2015 22:25
ok let me know if you find something
gitunit
@gitunit
Aug 17 2015 22:25
i've seen it for quite a time
Krystian Gebis
@l0g1x
Aug 17 2015 22:31
ill push the code that i have in a second. itll have incorrect setting of camera_info parameters though, but just want to give it to you incase you can also figure it out
gitunit
@gitunit
Aug 17 2015 22:32
ok. i've looked into the openCV docu but couldn't find a different distortion model, so let's assume the standard for now.
gitunit
@gitunit
Aug 17 2015 22:39
do you have skype installed? maybe you can reach (since its in your timezone) someone from DUO who can give us information on the distortion model and if the information about the matrix coefficients provided by Sai above are correct.
Krystian Gebis
@l0g1x
Aug 17 2015 22:40
i have direct number, they respond whenever they get a chance
gitunit
@gitunit
Aug 17 2015 22:40
ok
Krystian Gebis
@l0g1x
Aug 17 2015 22:40
okay pushed
let me know if it builds because i didnt want to commit some dense3d changes i made
also make sure to read the github README
and if the build fails, read the outputted logs
gitunit
@gitunit
Aug 17 2015 22:42
i've build the latest version, did the installation change somehow with your latest push?
Krystian Gebis
@l0g1x
Aug 17 2015 22:42
oh shit hold on i think i pushed to wrong branch fml
installation meaning?
okay try now
gitunit
@gitunit
Aug 17 2015 22:46
ok
not building because of name change for the dense3d.cpp file
Krystian Gebis
@l0g1x
Aug 17 2015 22:49
okay hold up
gitunit
@gitunit
Aug 17 2015 22:50
line 74, right?
Krystian Gebis
@l0g1x
Aug 17 2015 22:51
no it builds on mine because i have all the files, doesnt build on your because i didnt push them
gitunit
@gitunit
Aug 17 2015 22:51
the files are there, its just cmakelists i guess
now its building
or not ..
Krystian Gebis
@l0g1x
Aug 17 2015 22:52
git pull now
gitunit
@gitunit
Aug 17 2015 22:52
at least one error is gone
Krystian Gebis
@l0g1x
Aug 17 2015 22:53
after pull?
gitunit
@gitunit
Aug 17 2015 22:53
no, new error:
No such file or directory

include "driverDUOstereo.h"

                         ^
Krystian Gebis
@l0g1x
Aug 17 2015 22:53
okay hold on im just going to fresh clone into a new folder
kk works now
gitunit
@gitunit
Aug 17 2015 23:00
yes, it does
i will check if it runs .. is the calibration already publishing?
Krystian Gebis
@l0g1x
Aug 17 2015 23:01
i havent tried running it yet, cuz just wanted to get you the code
but it should
incorrectly though (most definetely)
gitunit
@gitunit
Aug 17 2015 23:03
ok, then i will better research about the coefficients ..
Krystian Gebis
@l0g1x
Aug 17 2015 23:04
cool
likewise
gitunit
@gitunit
Aug 17 2015 23:11
ok, i think the DUO arrays are columns
i think so because the extrinsics are in a struct which consists of rotation and translation
and the translation should be a column, i guess
Krystian Gebis
@l0g1x
Aug 17 2015 23:12
i think your right
would you set the translation matrix in the P3 P[7] indecies?
gitunit
@gitunit
Aug 17 2015 23:16
but if we don't know for sure how the coefficients are saved in DUO we can only guess
idk yet
Krystian Gebis
@l0g1x
Aug 17 2015 23:18
yeah well whenever duo gets back to me ill let you know
gitunit
@gitunit
Aug 17 2015 23:18
at least in openCV the coefficients are saved not as Sai described it above. see: http://docs.opencv.org/modules/imgproc/doc/geometric_transformations.html#undistort
Krystian Gebis
@l0g1x
Aug 17 2015 23:19
okay, so then maybe he specified it for duo
gitunit
@gitunit
Aug 17 2015 23:20
maybe .. we dont know for sure
i haven't seen it documented on the DUO site
Krystian Gebis
@l0g1x
Aug 17 2015 23:20
yeah me neither i tried looking everywhere
gitunit
@gitunit
Aug 17 2015 23:21
we could try using the openCV way since i think they highly build on openCV
Krystian Gebis
@l0g1x
Aug 17 2015 23:22
well like you said, we’d be only guessing.. so ill just wait for duo and if they dont get back by today ill call them tomorrow
they ussually respond to me within a hour or two
unless theyre really busy
gitunit
@gitunit
Aug 17 2015 23:23
ok
did you try the duo on arm?
Krystian Gebis
@l0g1x
Aug 17 2015 23:26
havent had a chance, but shouldnt make a difference whether its x86 or arm, at least on my end
gitunit
@gitunit
Aug 17 2015 23:27
yeah, i just thought you have maybe the ARM driver since im not able to download it from the site
Krystian Gebis
@l0g1x
Aug 17 2015 23:27
really?
gitunit
@gitunit
Aug 17 2015 23:28
yes. its quite strange since they advertised with it from day 1
and i waited for this for more than a year
Krystian Gebis
@l0g1x
Aug 17 2015 23:28
thats wierd.. i should have access to a arm board within the week so ill test it on there if i find time
gitunit
@gitunit
Aug 17 2015 23:29
if you have the driver .. it should i guess
but without access, im hopeless running it on ARM :(
Krystian Gebis
@l0g1x
Aug 17 2015 23:30
yeah they sent it to me to test like 3-4 weeks ago
gitunit
@gitunit
Aug 17 2015 23:30
i asked them yesterday, let's see
so back on the coefficients:
if Sai is correct, and we have k1 .. k6 than it should be (in openCV at least) the CV_CALIB_RATIONAL_MODEL
Krystian Gebis
@l0g1x
Aug 17 2015 23:35
nice
i guess it should be possible since this (http://docs.carnegierobotics.com/SL/api.html) driver's docu says that the rational model is supported in ROS:
Krystian Gebis
@l0g1x
Aug 17 2015 23:57
okay looking into it
gitunit
@gitunit
Aug 17 2015 23:57
ok, that's it for today. let me know if you have something new from duo. read ya!