zoukankan      html  css  js  c++  java
  • 基于LIVE555的RTSP QoS实现

    如何从OnDemandServerMediaSubsession类以及继承类对象中获取RTCP信息(句柄)

    OnDemandServerMediaSubsession.cpp void StreamState::startPlaying函数中添加:

    fRTCPInstance->setRRHandler(fMaster.fRRHandlerTask, fMaster.fRRHandlerClientData);

    OnDemandServerMediaSubsession.hh 中OnDemandServerMediaSubsession添加两个成员:

     TaskFunc* fRRHandlerTask;
     void* fRRHandlerClientData;

    以及成员函数

    setRTCPRRPacketHandler(TaskFunc* handler, void* clientData) {
      fRRHandlerTask = handler;
      fRRHandlerClientData = clientData;
    }

    创建CamServerMediaSubsession 对象时,设置回调。(CamServerMediaSubsession 是继承于 OnDemandServerMediaSubsession,重写createNewStreamSource和createNewRTPSink即可)

    CamServerMediaSubsession   *sub = CamServerMediaSubsession::createNew(*env, inputDevice, &device);
    ...
    sub->setRTCPRRPacketHandler(RTCPRRHandler, (void *)sub);
    ...

    函调函数中获取RTCP RR信息

    void RTCPRRHandler(void* clientData)
    {
        using namespace CamStream;
        CamServerMediaSubsession *sub = (CamServerMediaSubsession *)clientData;
        RTPSink *sink = sub->get_rtp_sink();
        if (!sink) {
            std::cout<<"unable to get sink obj, not ready"<<std::endl;
            return;
        }
        bool ignore_firstRR = true;
    
        RTPTransmissionStatsDB& transmissionStats = sink->transmissionStatsDB();
        RTPTransmissionStatsDB::Iterator iter(transmissionStats);
        RTPTransmissionStats* substat;
    
        while ((substat = iter.next()) != NULL) {
            auto cam = sub->get_cam_instance();
            auto jitter = substat->jitter();
            auto loss_ratio = ((float)substat->packetLossRatio()/256)*100;   // %
            auto rtt = (int)(((float)substat->roundTripDelay()/65536)*1000);  //ms
            auto last_bitrate = cam->get_bitrate();
    
            std::cout<<"SSRC "<<substat->SSRC()
                <<" RTT "<<rtt<<" ms"
                <<" jitter "<<jitter
                <<" loss "<<(int)loss_ratio<<"%"<<std::endl;
      } }

    最后,根据丢帧率以及RTTD等信息,我们可以调整视频源的码率,实现QoS。

    关于CamServerMediaSubsession的实现(实现下面两个函数,就可以蒋H264视频流转为RTP传输流,从而实现RTSP服务器) 

      /*source */
      FramedSource* CamServerMediaSubsession::createNewStreamSource(unsigned clientSessionId, unsigned& estBitrate) { estBitrate = static_cast<unsigned int>(this->bit_rate_); FramedSource *source = replicator_->createStreamReplica(); //H264VideoStreamDiscreteFramer的输入是离散的NALU //H264VideoStreamFramer的输入是stream bit流 FramedSource *h264_source = H264VideoStreamDiscreteFramer::createNew(envir(), source); return h264_source; } /*sink */ RTPSink* CamServerMediaSubsession::createNewRTPSink(Groupsock* rtpGroupsock, unsigned char rtpPayloadTypeIfDynamic, FramedSource* inputSource) { auto sink = H264VideoRTPSink::createNew(envir(), rtpGroupsock, rtpPayloadTypeIfDynamic, sps_nal_, sps_nal_size_, pps_nal_, pps_nal_size_); return sink; }

    总结,上面的实现修改了live555源码,官方推荐的方式是通过继承现有类重写方法来实现,不过代码看了半天没头绪,有知道怎么弄的告诉我声(vslinux@qq.com)

  • 相关阅读:
    吴恩达深度学习与神经网络
    吴恩达机器学习的ppt以及作业编程练习题答案(别人总结的)
    关于机器学习的小科普
    质因数分解
    FFT
    Luogu P1262 间谍网络
    关于次短路
    Luogu P1955 [NOI2015]程序自动分析
    Luogu P1041传染病控制
    Bzoj 1731 POJ 3169 Luogu P4878 Layout
  • 原文地址:https://www.cnblogs.com/rayfloyd/p/11720642.html
Copyright © 2011-2022 走看看