python - 如何在 Gstreamer 中检索流统计信息？

Question

首先我想说我真的是新手Gstreamer 及其功能请原谅我的无知，如果我的理解或执行错误，我还在学习。

我想用 PYTHON 或 JAVA(因为我不精通 C)和 QOS 构建一个小型流应用程序集成在其中，特别是包丢弃计数统计和 RTP和 RTCP 似乎是绝配。

为此我实现了一个服务器

#! /usr/bin/env python
import gi
import sys
gi.require_version('Gst', '1.0')
from gi.repository import GObject, Gst

#gst-launch -v rtpbin name=rtpbin audiotestsrc ! audioconvert ! alawenc ! rtppcmapay ! rtpbin.send_rtp_sink_0 \
#                rtpbin.send_rtp_src_0 ! udpsink port=10000 host=xxx.xxx.xxx.xxx \
#                rtpbin.send_rtcp_src_0 ! udpsink port=10001 host=xxx.xxx.xxx.xxx sync=false async=false \
#                udpsrc port=10002 ! rtpbin.recv_rtcp_sink_0

DEST_HOST = '127.0.0.1'

AUDIO_SRC = 'audiotestsrc'
AUDIO_ENC = 'alawenc'
AUDIO_PAY = 'rtppcmapay'

RTP_SEND_PORT = 5002
RTCP_SEND_PORT = 5003
RTCP_RECV_PORT = 5007

GObject.threads_init()
Gst.init(sys.argv)

# the pipeline to hold everything
pipeline = Gst.Pipeline.new('rtp_server')

# the pipeline to hold everything
audiosrc = Gst.ElementFactory.make(AUDIO_SRC, 'audiosrc')
audioconv = Gst.ElementFactory.make('audioconvert', 'audioconv')
audiores = Gst.ElementFactory.make('audioresample', 'audiores')

# the pipeline to hold everything
audioenc = Gst.ElementFactory.make(AUDIO_ENC, 'audioenc')
audiopay = Gst.ElementFactory.make(AUDIO_PAY, 'audiopay')

# add capture and payloading to the pipeline and link
pipeline.add(audiosrc)
pipeline.add(audioconv)
pipeline.add(audiores)
pipeline.add(audioenc)
pipeline.add(audiopay)


audiosrc.link(audioconv)
audioconv.link(audiores)
audiores.link(audioenc)
audioenc.link(audiopay)

# the rtpbin element
rtpbin = Gst.ElementFactory.make('rtpbin', 'rtpbin')

pipeline.add(rtpbin)

# the udp sinks and source we will use for RTP and RTCP
rtpsink = Gst.ElementFactory.make('udpsink', 'rtpsink')
rtpsink.set_property('port', RTP_SEND_PORT)
rtpsink.set_property('host', DEST_HOST)

rtcpsink = Gst.ElementFactory.make('udpsink', 'rtcpsink')
rtcpsink.set_property('port', RTCP_SEND_PORT)
rtcpsink.set_property('host', DEST_HOST)
# no need for synchronisation or preroll on the RTCP sink
rtcpsink.set_property('async', False)
rtcpsink.set_property('sync', False)

rtcpsrc = Gst.ElementFactory.make('udpsrc', 'rtcpsrc')
rtcpsrc.set_property('port', RTCP_RECV_PORT)

pipeline.add(rtpsink)
pipeline.add(rtcpsink)
pipeline.add(rtcpsrc)

# now link all to the rtpbin, start by getting an RTP sinkpad for session 0
sinkpad = Gst.Element.get_request_pad(rtpbin, 'send_rtp_sink_0')
srcpad = Gst.Element.get_static_pad(audiopay, 'src')
lres = Gst.Pad.link(srcpad, sinkpad)

# get the RTP srcpad that was created when we requested the sinkpad above and
# link it to the rtpsink sinkpad
srcpad = Gst.Element.get_static_pad(rtpbin, 'send_rtp_src_0')
sinkpad = Gst.Element.get_static_pad(rtpsink, 'sink')
lres = Gst.Pad.link(srcpad, sinkpad)

# get an RTCP srcpad for sending RTCP to the receiver
srcpad = Gst.Element.get_request_pad(rtpbin, 'send_rtcp_src_0')
sinkpad = Gst.Element.get_static_pad(rtcpsink, 'sink')
lres = Gst.Pad.link(srcpad, sinkpad)

# we also want to receive RTCP, request an RTCP sinkpad for session 0 and
# link it to the srcpad of the udpsrc for RTCP
srcpad = Gst.Element.get_static_pad(rtcpsrc, 'src')
sinkpad = Gst.Element.get_request_pad(rtpbin, 'recv_rtcp_sink_0')
lres = Gst.Pad.link(srcpad, sinkpad)

# set the pipeline to playing
Gst.Element.set_state(pipeline, Gst.State.PLAYING)

# we need to run a GLib main loop to get the messages
mainloop = GObject.MainLoop()
mainloop.run()
Gst.Element.set_state(pipeline, Gst.State.NULL)

和一个客户

#! /usr/bin/env python

import gi
import sys
gi.require_version('Gst', '1.0')
from gi.repository import GObject, Gst

#
# A simple RTP receiver
#
#  receives alaw encoded RTP audio on port 5002, RTCP is received on  port 5003.
#  the receiver RTCP reports are sent to port 5007
#
#             .-------.      .----------.     .---------.   .-------.   .--------.
#  RTP        |udpsrc |      | rtpbin   |     |pcmadepay|   |alawdec|   |alsasink|
#  port=5002  |      src->recv_rtp recv_rtp->sink     src->sink   src->sink      |
#             '-------'      |          |     '---------'   '-------'   '--------'
#                            |          |
#                            |          |     .-------.
#                            |          |     |udpsink|  RTCP
#                            |    send_rtcp->sink     | port=5007
#             .-------.      |          |     '-------' sync=false
#  RTCP       |udpsrc |      |          |               async=false
#  port=5003  |     src->recv_rtcp      |
#             '-------'      '----------'

AUDIO_CAPS = 'application/x-rtp,media=(string)audio,clock-rate=(int)8000,encoding-name=(string)PCMA'
AUDIO_DEPAY = 'rtppcmadepay'
AUDIO_DEC = 'alawdec'
AUDIO_SINK = 'autoaudiosink'

DEST = '127.0.0.1'

RTP_RECV_PORT = 5002
RTCP_RECV_PORT = 5003
RTCP_SEND_PORT = 5007

GObject.threads_init()
Gst.init(sys.argv)

#gst-launch -v rtpbin name=rtpbin                                                \
#       udpsrc caps=$AUDIO_CAPS port=$RTP_RECV_PORT ! rtpbin.recv_rtp_sink_0              \
#             rtpbin. ! rtppcmadepay ! alawdec ! audioconvert ! audioresample ! autoaudiosink \
#           udpsrc port=$RTCP_RECV_PORT ! rtpbin.recv_rtcp_sink_0                              \
#         rtpbin.send_rtcp_src_0 ! udpsink port=$RTCP_SEND_PORT host=$DEST sync=false async=false

def pad_added_cb(rtpbin, new_pad, depay):
    sinkpad = Gst.Element.get_static_pad(depay, 'sink')
    lres = Gst.Pad.link(new_pad, sinkpad)

# the pipeline to hold eveything
pipeline = Gst.Pipeline.new('rtp_client')

# the udp src and source we will use for RTP and RTCP
rtpsrc = Gst.ElementFactory.make('udpsrc', 'rtpsrc')
rtpsrc.set_property('port', RTP_RECV_PORT)

# we need to set caps on the udpsrc for the RTP data
caps = Gst.caps_from_string(AUDIO_CAPS)
rtpsrc.set_property('caps', caps)

rtcpsrc = Gst.ElementFactory.make('udpsrc', 'rtcpsrc')
rtcpsrc.set_property('port', RTCP_RECV_PORT)

rtcpsink = Gst.ElementFactory.make('udpsink', 'rtcpsink')
rtcpsink.set_property('port', RTCP_SEND_PORT)
rtcpsink.set_property('host', DEST)

# no need for synchronisation or preroll on the RTCP sink
rtcpsink.set_property('async', False)
rtcpsink.set_property('sync', False)

pipeline.add(rtpsrc)
pipeline.add(rtcpsrc)
pipeline.add(rtcpsink)

# the depayloading and decoding
audiodepay = Gst.ElementFactory.make(AUDIO_DEPAY, 'audiodepay')
audiodec = Gst.ElementFactory.make(AUDIO_DEC, 'audiodec')

# the audio playback and format conversion
audioconv = Gst.ElementFactory.make('audioconvert', 'audioconv')
audiores = Gst.ElementFactory.make('audioresample', 'audiores')
audiosink = Gst.ElementFactory.make(AUDIO_SINK, 'audiosink')

# add depayloading and playback to the pipeline and link
pipeline.add(audiodepay)
pipeline.add(audiodec)
pipeline.add(audioconv)
pipeline.add(audiores)
pipeline.add(audiosink)


audiodepay.link(audiodec)
audiodec.link(audioconv)
audioconv.link(audiores)
audiores.link(audiosink)

# the rtpbin element
rtpbin = Gst.ElementFactory.make('rtpbin', 'rtpbin')

pipeline.add(rtpbin)

# now link all to the rtpbin, start by getting an RTP sinkpad for session 0
srcpad = Gst.Element.get_static_pad(rtpsrc, 'src')
sinkpad = Gst.Element.get_request_pad(rtpbin, 'recv_rtp_sink_0')
lres = Gst.Pad.link(srcpad, sinkpad)

# get an RTCP sinkpad in session 0
srcpad = Gst.Element.get_static_pad(rtcpsrc, 'src')
sinkpad = Gst.Element.get_request_pad(rtpbin, 'recv_rtcp_sink_0')
lres = Gst.Pad.link(srcpad, sinkpad)

# get an RTCP srcpad for sending RTCP back to the sender
srcpad = Gst.Element.get_request_pad(rtpbin, 'send_rtcp_src_0')
sinkpad = Gst.Element.get_static_pad(rtcpsink, 'sink')
lres = Gst.Pad.link(srcpad, sinkpad)

rtpbin.connect('pad-added', pad_added_cb, audiodepay)



# def newManager():
    # rtpbin.connect('on-ssrc-active', onSSRCActive)

# def onSSRCActive(self):
    # print (self)

# rtpsrc.connect('new-manager', newManager)

Gst.Element.set_state(pipeline, Gst.State.PLAYING)

mainloop = GObject.MainLoop()
mainloop.run()

Gst.Element.set_state(pipeline, Gst.State.NULL)

虽然服务器和客户端工作没有问题，但我找不到方法或有关如何从 RTCP 检索此 RTP 流的丢弃包统计信息的信息，即使它是在命令行中也是如此。

如有任何帮助，我们将不胜感激!

Answer 1

这是一个在 Linux 上的模型，只是为了说明一种监听总线消息的方法。
我没有时间整理 QOS Gst.Message，所以我在里面黑了一些东西来强制 QOS，这样你至少可以看到它。
希望这足以让您朝着正确的方向前进。

#!/usr/bin/python3

from os import path
import time
import gi
gi.require_version('Gst', '1.0')
gi.require_version('Gtk', '3.0')
gi.require_version('GstVideo', '1.0')
from gi.repository import GObject, Gst, Gtk
from gi.repository import GdkX11, GstVideo


GObject.threads_init()
Gst.init(None)
filename = path.join(path.dirname(path.abspath(__file__)), '../H.mkv')
uri = 'file://' + filename


class Player(object):
    def __init__(self):
        self.window = Gtk.Window()
        self.window.connect('destroy', self.quit)
        self.window.set_default_size(800, 450)

        self.drawingarea = Gtk.DrawingArea()
        self.window.add(self.drawingarea)

        # Create GStreamer pipeline
        self.pipeline = Gst.Pipeline()

        # Create bus to get events from GStreamer pipeline
        self.bus = self.pipeline.get_bus()
        self.bus.add_signal_watch()
        self.bus.connect('message::eos', self.on_eos)
        self.bus.connect('message::error', self.on_error)
        self.bus.connect('message::qos', self.on_quality_of_service)

        self.bus.enable_sync_message_emission()
        self.bus.connect('sync-message::element', self.on_sync_message)

        # Create GStreamer elements
        self.playbin = Gst.ElementFactory.make('playbin', None)

        # Add playbin to the pipeline
        self.pipeline.add(self.playbin)

        # Set properties
        self.playbin.set_property('uri', uri)

    def run(self):
        self.window.show_all()
        self.xid = self.drawingarea.get_property('window').get_xid()
        self.pipeline.set_state(Gst.State.PLAYING)
        time.sleep(2)
        self.bus.emit('message::qos',Gst.Message('xxxxx'))
        Gtk.main()

    def quit(self, window):
        self.pipeline.set_state(Gst.State.NULL)
        Gtk.main_quit()

    def on_sync_message(self, bus, msg):
        if msg.get_structure().get_name() == 'prepare-window-handle':
            print('prepare-window-handle')
            msg.src.set_window_handle(self.xid)

    def on_eos(self, bus, msg):
        print('End of Service')

    def on_error(self, bus, msg):
        print('Error', msg.parse_error())

    def on_quality_of_service(self, bus, msg):
        print('Qos Message:', msg.parse_qos())

p = Player()
p.run()