Timestamps, clock and duration of recorded bag files

asked 2015-03-29 22:34:07 -0500

Valerio gravatar image

I am working on a dataset recorded in a bag file, that contains a sequence of images acquired at 10 fps with stereo cameras. I don't understand the time information contained in this dataset. The image timestamps start at 0 seconds and end at 201.577 seconds, while the output of rosbag play displays a duration of 196.911 seconds. I have the following questions:

1) Why is the duration of the recording smaller than the difference between the first and the last time stamp of the images? I would expect them to be the same, or the duration to be slightly larger than the difference between the time stamps.

2) Which is the source of the recorded clock in the bag file and the timestamps of the images? Do they come from the wall clock of the computed used for the recording? Do the cameras have any direct role in computing the timestamps?

3) The timestamps of the images coming from the left and right cameras agree at the nanosecond level. This sounds a little ridiculous to me. Is it possible that the system used for recording computed only one timestamp for each image pair? If this is what happened, how am I supposed to know if the left and right cameras are actually synchronized? I guess there's no way...

Thanks in advance!

edit retag flag offensive close merge delete