MediaStream图像捕捉
Posted
tags:
篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了MediaStream图像捕捉相关的知识,希望对你有一定的参考价值。
参考技术A MediaStream图像捕捉API 是一种用于从照相设备捕捉图像或视频的API。1.通过调用获取对设备的引用 MediaDevices.getUserMedia() ;
2.隔离媒体流的可视部分;
通过调用来做到这一点 MediaStream.getVideoTracks() 。这将返回一个 MediaStreamTrack 对象数组。
3.如果希望在捕获图像之前配置设备功能,可以通过 applyConstraints() 在执行任何其他操作之前调用track 对象来完成此操作;
4.将MediaStreamTrack对象传递给 ImageCapture() 构造函数。
尽管 aMediaStream包含多种类型的轨道并提供多种检索它们的方法,但 ImageCapture 构造函数将抛出 a DOMException of type NotSupportedErrorif MediaStreamTrack.kind is not "video"。
ImageCapture 图像捕获,用于从通过有效引用的摄影设备捕获图像的接口 MediaStreamTrack 。
ImageCapture() 构造函数
创建一个新ImageCapture对象,该对象可用于从MediaStreamTrack表示视频流的给定对象中捕获 静止帧 (照片)。
特性:ImageCapture.track 只读,返回对 MediaStreamTrack (媒体流跟踪)传递给构造函数的引用。
ImageCapture.track 图像捕捉跟踪,接口的 track 只读属性 ImageCapture返回对MediaStreamTrack传递给 ImageCapture()构造函数的引用 。
该 MediaStreamTrack 接口表示流中的单个媒体轨道;通常,这些是音频或视频轨道,但也可能存在其他轨道类型。
接口的 takePhoto() 方法 ImageCapture 使用视频捕获设备进行单次曝光 MediaStreamTrack 并返回 Promise 一个 Blob 包含数据的解析。
licode学习之erizo篇--MediaStream
MediaStream是erizo进行流数据处理的核心模块。当网络数据,经过DtlsTransport进行srtp解密后,得到的rtp裸数据与rtcp裸数据,都要进入MediaStream进行处理;需要发送给对方的rtp数据与rtcp裸数据也要经过MediaStream处理后,才会给DtlsTransport进行加密并发送。
MediaStream也是个人认为erizo的最为复杂的一个部分。先看一看其集成体系:
class MediaStream: public MediaSink, public MediaSource, public FeedbackSink, public FeedbackSource, public LogContext, public HandlerManagerListener, public std::enable_shared_from_this<MediaStream>, public Service
整个继承体系里面,涉及处理的基类有:MediaSink,MediaSource,FeedbackSink,FeedbackSource, HandlerManagerListener,Service
MediaStream同时承载着收数据处理,和发送数据处理两部分内容。其中和丢包重传等结合起来,就变为:接收rtp数据,发送rtcp重传信息;发送rtp数据,接收rtcp重传信息。
在接口上面,需要对外提供发送和接收到的裸数据回调。
这里面erizo的实现,是将MediaSink与MediaSource纠缠到一起的。但是从宏观上,我的理解是:
MediaSink:负责发送数据(write to client)
FeedbackSink:负责发送数据(write to client)
MediaSource:负责read出来rtp数据 (read from client)
FeedbackSource:负责read出来数据(read from client)
MediaStream继承MediaSink和FeedbackSink,所以直接调用MediaStream对象的deliverVideoData,deliverAudioData,deliverFeedback即可直接向对端发送数据。
要想接收对方的数据,需要MediaSource,FeedbackSource进行数据回调,怎么回调,需要进一步看一下MediaSource的定义:
** * A MediaSource is any class that produces audio or video data. */ class MediaSource: public virtual Monitor { protected: // SSRCs coming from the source uint32_t audio_source_ssrc_; std::vector<uint32_t> video_source_ssrc_list_; MediaSink* video_sink_; MediaSink* audio_sink_; MediaSink* event_sink_; // can it accept feedback FeedbackSink* source_fb_sink_; public: void setAudioSink(MediaSink* audio_sink) { boost::mutex::scoped_lock lock(monitor_mutex_); this->audio_sink_ = audio_sink; } void setVideoSink(MediaSink* video_sink) { boost::mutex::scoped_lock lock(monitor_mutex_); this->video_sink_ = video_sink; } void setEventSink(MediaSink* event_sink) { boost::mutex::scoped_lock lock(monitor_mutex_); this->event_sink_ = event_sink; } FeedbackSink* getFeedbackSink() { boost::mutex::scoped_lock lock(monitor_mutex_); return source_fb_sink_; } virtual int sendPLI() = 0; uint32_t getVideoSourceSSRC() { boost::mutex::scoped_lock lock(monitor_mutex_); if (video_source_ssrc_list_.empty()) { return 0; } return video_source_ssrc_list_[0]; } void setVideoSourceSSRC(uint32_t ssrc) { boost::mutex::scoped_lock lock(monitor_mutex_); if (video_source_ssrc_list_.empty()) { video_source_ssrc_list_.push_back(ssrc); return; } video_source_ssrc_list_[0] = ssrc; } std::vector<uint32_t> getVideoSourceSSRCList() { boost::mutex::scoped_lock lock(monitor_mutex_); return video_source_ssrc_list_; // return by copy to avoid concurrent access } void setVideoSourceSSRCList(const std::vector<uint32_t>& new_ssrc_list) { boost::mutex::scoped_lock lock(monitor_mutex_); video_source_ssrc_list_ = new_ssrc_list; } uint32_t getAudioSourceSSRC() { boost::mutex::scoped_lock lock(monitor_mutex_); return audio_source_ssrc_; } void setAudioSourceSSRC(uint32_t ssrc) { boost::mutex::scoped_lock lock(monitor_mutex_); audio_source_ssrc_ = ssrc; } bool isVideoSourceSSRC(uint32_t ssrc) { auto found_ssrc = std::find_if(video_source_ssrc_list_.begin(), video_source_ssrc_list_.end(), [ssrc](uint32_t known_ssrc) { return known_ssrc == ssrc; }); return (found_ssrc != video_source_ssrc_list_.end()); } bool isAudioSourceSSRC(uint32_t ssrc) { return audio_source_ssrc_ == ssrc; } MediaSource() : audio_source_ssrc_{0}, video_source_ssrc_list_{std::vector<uint32_t>(1, 0)}, video_sink_{nullptr}, audio_sink_{nullptr}, event_sink_{nullptr}, source_fb_sink_{nullptr} {} virtual ~MediaSource() {} virtual void close() = 0; };
MediaSource定义了4个MediaSink对象,分别对应video,audio,event,feedback。
MediaStream继承了MediaSource同样的4个set接口,让调用者可以设置MediaSource的这4个对象。当发生读取事件时,MediaStream会调用这4个设置的MediaSink对象的相应方法,来向外传递数据。
FeedbackSource也是同样的道理。
void MediaStream::read(std::shared_ptr<DataPacket> packet) { char* buf = packet->data; int len = packet->length; // PROCESS RTCP RtpHeader *head = reinterpret_cast<RtpHeader*> (buf); RtcpHeader *chead = reinterpret_cast<RtcpHeader*> (buf); uint32_t recvSSRC = 0; if (!chead->isRtcp()) { recvSSRC = head->getSSRC(); } else if (chead->packettype == RTCP_Sender_PT || chead->packettype == RTCP_SDES_PT) { // Sender Report recvSSRC = chead->getSSRC(); } // DELIVER FEEDBACK (RR, FEEDBACK PACKETS) if (chead->isFeedback()) { if (fb_sink_ != nullptr && should_send_feedback_) { fb_sink_->deliverFeedback(std::move(packet)); } } else { // RTP or RTCP Sender Report if (bundle_) { // Check incoming SSRC // Deliver data if (isVideoSourceSSRC(recvSSRC) && video_sink_) { parseIncomingPayloadType(buf, len, VIDEO_PACKET); video_sink_->deliverVideoData(std::move(packet)); } else if (isAudioSourceSSRC(recvSSRC) && audio_sink_) { parseIncomingPayloadType(buf, len, AUDIO_PACKET); audio_sink_->deliverAudioData(std::move(packet)); } else { ELOG_DEBUG("%s read video unknownSSRC: %u, localVideoSSRC: %u, localAudioSSRC: %u", toLog(), recvSSRC, this->getVideoSourceSSRC(), this->getAudioSourceSSRC()); } } else { if (packet->type == AUDIO_PACKET && audio_sink_) { parseIncomingPayloadType(buf, len, AUDIO_PACKET); // Firefox does not send SSRC in SDP if (getAudioSourceSSRC() == 0) { ELOG_DEBUG("%s discoveredAudioSourceSSRC:%u", toLog(), recvSSRC); this->setAudioSourceSSRC(recvSSRC); } audio_sink_->deliverAudioData(std::move(packet)); } else if (packet->type == VIDEO_PACKET && video_sink_) { parseIncomingPayloadType(buf, len, VIDEO_PACKET); // Firefox does not send SSRC in SDP if (getVideoSourceSSRC() == 0) { ELOG_DEBUG("%s discoveredVideoSourceSSRC:%u", toLog(), recvSSRC); this->setVideoSourceSSRC(recvSSRC); } // change ssrc for RTP packets, don‘t touch here if RTCP video_sink_->deliverVideoData(std::move(packet)); } } // if not bundle } // if not Feedback }
这里,MediaStream的读写就清楚了,如果我们需要使用MediaStream,则需要做:
1、定义一个MediaSink的子类,将之设置给MediaStream,用于接收MediaStream的数据
2、直接调用MediaStream的deliver方法,让其向外发送数据。
class Receiver : public MediaSink { public: virtual void close() {}; private: virtual int deliverAudioData_(std::shared_ptr<DataPacket> data_packet) { printf("now receiver audio packet "); }; virtual int deliverVideoData_(std::shared_ptr<DataPacket> data_packet) { printf("now receive video packet "); }; virtual int deliverEvent_(MediaEventPtr event) { printf("now receive event packet "); }; }; class StreamControl { public: void init() { m_workerPool = std::make_shared<ThreadPool>(2); m_ioWorkerPool = std::make_shared<IOThreadPool>(2); std::string connid = "1"; IceConfig cfg;//you may need init the cfg value std::vector<RtpMap> rtp_mappings;//you may need to init the mappings std::vector<erizo::ExtMap> ext_mappings; //you may need to init the ext mappings WebRtcConnectionEventListener* listener = new SendOfferEvtListener; m_conn = std::make_shared<WebRtcConnection>(workerPool->getLessUsedWorker(), m_ioWorkerPool->getLessUsedIOWorker(), connid, cfg, rtp_mappings, ext_mappings, listener); std::string stream_id = "1"; std::string stream_label = "1"; m_pStream = new MediaStream(m_workerPool->getLessUsedWorker(), m_conn, stream_id, stream_label, true); m_conn->addMediaStream(m_pStream); m_pStream->setVideoSink(&m_receiver); m_pStream->setAudioSink(&m_receiver); m_pStream->setEventSink(&m_receiver); } void sendRtpData(const char* buf, int len, bool isVideo) { std::shared_ptr<DataPacket> dp = std::make_shared<DataPacket>(); dp->length = len; memcpy(dp->data, buf); dp->comp = 1; dp->type = isVideo ? VIDEO_PACKET : AUDIO_PACKET; if (isVideo) { m_pStream->deliverVideoData(dp); } else { m_pStream->deliverAudioData(dp); } } private: Receiver m_receiver; MediaStream* m_pStream; std::shared_ptr<ThreadPool> m_workerPool; std::shared_ptr<IOThreadPool> m_ioWorkerPool; WebRtcConnection* m_conn; };
MediaStream还继承了HandlerManagerListener,Service这两部分是媒体处理的核心模块,pipeline使用的内容,之后学习时再写吧。pipeline更是复杂,代码好难读。
以上是关于MediaStream图像捕捉的主要内容,如果未能解决你的问题,请参考以下文章