Handling SEGMENT event when switching between two cameras

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Handling SEGMENT event when switching between two cameras

jean-philippe
Hi,

I have two RTSP cameras #1 and #2 which are displayed and recorded to MP4
(using qtmux). I need to be able to switch from one RTSP camera to the other
and keep on appending the h.264 stream to the MP4 and display the live
video. I also need to be able to pause and resume recording.

Both cameras use the same encoder, which is programmed with the exact same
settings.

To pause the recording, I use a pad probe on the src pad of the mux which
drops buffers whilst paused. I also set the DTS and PTS to a monotonic
counter so that the MP4 mux sees continuous time.

To switch, from #1 to #2, I use a blocking pad probe on the sink of the mux
to wait until the mux has received the full GOP, then I disconnect #1,
destroy it, create #2, connect it and set its element to PLAYING.

At that point, I receive a new SEGMENT event, sent from #2, with a new
position (a large number) and the mux starts to clip all subsequent buffers.
The end result is that the total time in the MP4 no longer increases (even
though its size increases, suggesting the data carried by the buffers is
appended to it).

I have worked out that if I drop this new SEGMENT event, then MP4 mux no
longer clips new buffers and I can play the MP4 as expected.

My understanding is that when setting the RTSP source to PLAYING it sends a
SEGMENT event before any BUFFER. This event carries a position value which
is used by downstream elements to decide whether buffers should be clipped
or not. In my case, since I set the DTS and PTS to my own monotonic counter
which started at 0 and the position carried by the SEGMENT event is a much
larger number, the timestamps in the buffers are always smaller than then
position which causes the MP4 mux to clip the buffers as they fall outside
the segment.

Is my understanding correct?

If so, this makes sense to me, but I'm not sure how to use this knowledge to
best to structure my app so I'm asking for guidance as to how to best
implement this application.

Dropping the SEGMENT events seems very hacky, or is it? Should I be dropping
all SEGMENT events and pushing my own once only, when I connect to the first
RTSP camera.

Note that the very first SEGMENT element I currently let through also
carries a position set to a larger number than the timestamp I set in the
buffers. This should cause subsequent buffers to be clipped, but they are
not. Why is that?

Are there elements already available to do what I need?

I have found the concat element, but the following pipeline produces errors
and the MP4 is not fully playable, so I did not investigate further:
.\gst-launch-1.0.exe concat name=c ! x264enc ! queue ! qtmux ! filesink
location=test.mp4 videotestsrc num-buffers=100 ! c.   videotestsrc
num-buffers=100 pattern=ball ! c. -e
ERROR                  qtmux gstqtmux.c:4713:gst_qt_mux_add_buffer:
decreasing DTS value 0:00:06.633333332 < 0:00:06.699999999

Thanks



--
Sent from: http://gstreamer-devel.966125.n4.nabble.com/
_______________________________________________
gstreamer-devel mailing list
[hidden email]
https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel