一、简介
媒体子系统为开发者提供了媒体相干的很多性能，本文针对其中的视频录制性能做个具体的介绍。首先，我将通过媒体子系统提供的视频录制Test代码作为切入点，给大家梳理一下整个录制的流程。
二、目录
foundation/multimedia/camera_framework
├── frameworks│ ├── js│ │ └── camera_napi #napi实现│ │ └── src│ │ ├── input #Camera输出│ │ ├── output #Camera输入│ │ └── session #会话治理│ └── native #native实现│ └── camera│ ├── BUILD.gn│ ├── src│ │ ├── input #Camera输出│ │ ├── output #Camera输入│ │ └── session #会话治理├── interfaces #接口定义│ ├── inner_api #外部native实现│ │ └── native│ │ ├── camera│ │ │ └── include│ │ │ ├── input│ │ │ ├── output│ │ │ └── session│ └── kits #napi接口│ └── js│ └── camera_napi│ ├── BUILD.gn│ ├── include│ │ ├── input│ │ ├── output│ │ └── session│ └── @ohos.multimedia.camera.d.ts└── services #服务端 └── camera_service ├── binder │ ├── base │ ├── client #IPC的客户端 │ │ └── src │ └── server #IPC的服务端 │ └── src └── src三、录制的总体流程
四、Native接口应用
在OpenAtom OpenHarmony（以下简称“OpenHarmony”）零碎中，多媒体子系统通过N-API接口提供给下层JS调用，N-API相当于是JS和Native之间的桥梁，在OpenHarmony源码中，提供了C++间接调用视频录制性能的例子，foundation/multimedia/camera_framework/interfaces/inner_api/native/test目录中。本文章次要参考了camera_video.cpp文件中的视频录制流程。
首先依据camera_video.cpp的main办法，理解下视频录制的次要流程代码。
int main(int argc, char **argv){ ...... // 创立CameraManager实例 sptr<CameraManager> camManagerObj = CameraManager::GetInstance(); // 设置回调 camManagerObj->SetCallback(std::make_shared<TestCameraMngerCallback>(testName)); // 获取反对的相机设施列表 std::vector<sptr<CameraDevice>> cameraObjList = camManagerObj->GetSupportedCameras(); // 创立采集会话 sptr<CaptureSession> captureSession = camManagerObj->CreateCaptureSession(); // 开始配置采集会话 captureSession->BeginConfig(); // 创立CameraInput sptr<CaptureInput> captureInput = camManagerObj->CreateCameraInput(cameraObjList[0]); sptr<CameraInput> cameraInput = (sptr<CameraInput> &)captureInput; // 开启CameraInput cameraInput->Open(); // 设置CameraInput的Error回调 cameraInput->SetErrorCallback(std::make_shared<TestDeviceCallback>(testName)); // 增加CameraInput实例到采集会话中 ret = captureSession->AddInput(cameraInput); sptr<Surface> videoSurface = nullptr; std::shared_ptr<Recorder> recorder = nullptr; // 创立Video的Surface videoSurface = Surface::CreateSurfaceAsConsumer(); sptr<SurfaceListener> videoListener = new SurfaceListener("Video", SurfaceType::VIDEO, g_videoFd, videoSurface); // 注册Surface的事件监听 videoSurface->RegisterConsumerListener((sptr<IBufferConsumerListener> &)videoListener); // 视频的配置 VideoProfile videoprofile = VideoProfile(static_cast<CameraFormat>(videoFormat), videosize, videoframerates); // 创立VideoOutput实例 sptr<CaptureOutput> videoOutput = camManagerObj->CreateVideoOutput(videoprofile, videoSurface); // 设置VideoOutput的回调 ((sptr<VideoOutput> &)videoOutput)->SetCallback(std::make_shared<TestVideoOutputCallback>(testName)); // 增加videoOutput到采集会话中 ret = captureSession->AddOutput(videoOutput); // 提交会话配置 ret = captureSession->CommitConfig(); // 开始录制 ret = ((sptr<VideoOutput> &)videoOutput)->Start(); sleep(videoPauseDuration); MEDIA_DEBUG_LOG("Resume video recording"); // 暂停录制 ret = ((sptr<VideoOutput> &)videoOutput)->Resume(); MEDIA_DEBUG_LOG("Wait for 5 seconds before stop"); sleep(videoCaptureDuration); MEDIA_DEBUG_LOG("Stop video recording"); // 进行录制 ret = ((sptr<VideoOutput> &)videoOutput)->Stop(); MEDIA_DEBUG_LOG("Closing the session"); // 进行采集会话 ret = captureSession->Stop(); MEDIA_DEBUG_LOG("Releasing the session"); // 开释会话采集 captureSession->Release(); // Close video file TestUtils::SaveVideoFile(nullptr, 0, VideoSaveMode::CLOSE, g_videoFd); cameraInput->Release(); camManagerObj->SetCallback(nullptr); return 0;}以上是视频录制的整体流程，其过程次要通过Camera模块反对的能力来实现，其中波及几个重要的类：CaptureSession、CameraInput、VideoOutput。CaptureSession是整个过程的控制者，CameraInput和VideoOutput相当于是设施的输出和输入。
五、调用流程
后续次要针对下面的调用流程，梳理具体的调用流程，不便咱们对理解视频录制的整顿架构有一个更加深刻的理解。
创立CameraManager实例
通过CameraManager::GetInstance()获取CameraManager的实例，后续的一些接口都是通过该实例进行调用的。GetInstance应用了单例模式，在OpenHarmony代码中这种形式很常见。sptr<CameraManager> &CameraManager::GetInstance(){ if (CameraManager::cameraManager_ == nullptr) { MEDIA_INFO_LOG("Initializing camera manager for first time!"); CameraManager::cameraManager_ = new(std::nothrow) CameraManager(); if (CameraManager::cameraManager_ == nullptr) { MEDIA_ERR_LOG("CameraManager::GetInstance failed to new CameraManager"); } } return CameraManager::cameraManager_;}获取反对的相机设施列表
通过调用CameraManager的GetSupportedCameras()接口，获取设施反对的CameraDevice列表。跟踪代码能够发现serviceProxy_->GetCameras最终会调用到Camera服务端的对应接口。std::vector<sptr<CameraDevice>> CameraManager::GetSupportedCameras(){ CAMERA_SYNC_TRACE; std::lock_guard<std::mutex> lock(mutex_); std::vector<std::string> cameraIds; std::vector<std::shared_ptr<Camera::CameraMetadata>> cameraAbilityList; int32_t retCode = -1; sptr<CameraDevice> cameraObj = nullptr; int32_t index = 0; if (cameraObjList.size() > 0) { cameraObjList.clear(); } if (serviceProxy_ == nullptr) { MEDIA_ERR_LOG("CameraManager::GetCameras serviceProxy_ is null, returning empty list!"); return cameraObjList; } std::vector<sptr<CameraDevice>> supportedCameras; retCode = serviceProxy_->GetCameras(cameraIds, cameraAbilityList); if (retCode == CAMERA_OK) { for (auto& it : cameraIds) { cameraObj = new(std::nothrow) CameraDevice(it, cameraAbilityList[index++]); if (cameraObj == nullptr) { MEDIA_ERR_LOG("CameraManager::GetCameras new CameraDevice failed for id={public}%s", it.c_str()); continue; } supportedCameras.emplace_back(cameraObj); } } else { MEDIA_ERR_LOG("CameraManager::GetCameras failed!, retCode: %{public}d", retCode); } ChooseDeFaultCameras(supportedCameras); return cameraObjList;}创立采集会话
上面是比拟重要的环节，通过调用CameraManager的CreateCaptureSession接口创立采集会话。CameraManager创立采集会话，是通过serviceProxy_->CreateCaptureSession形式进行调用，这里波及到了OpenHarmony中的IPC的调用，serviceProxy_是远端服务在本地的代理，通过这个代理能够调用到具体的服务端，这里是HCameraService。sptr<CaptureSession> CameraManager::CreateCaptureSession(){ CAMERA_SYNC_TRACE; sptr<ICaptureSession> captureSession = nullptr; sptr<CaptureSession> result = nullptr; int32_t retCode = CAMERA_OK; if (serviceProxy_ == nullptr) { MEDIA_ERR_LOG("CameraManager::CreateCaptureSession serviceProxy_ is null"); return nullptr; } retCode = serviceProxy_->CreateCaptureSession(captureSession); if (retCode == CAMERA_OK && captureSession != nullptr) { result = new(std::nothrow) CaptureSession(captureSession); if (result == nullptr) { MEDIA_ERR_LOG("Failed to new CaptureSession"); } } else { MEDIA_ERR_LOG("Failed to get capture session object from hcamera service!, %{public}d", retCode); } return result;}
代码最终来到HCameraService::CreateCaptureSession中，该办法中new了一个HCaptureSession对象，并且将该对象传递给了参数session，所以后面的captureSession对象就是这里new进去的HCaptureSession，后面的CameraManager的CreateCaptureSession()办法中将captureSession封装成CaptureSession对象返回给应用层应用。
int32_t HCameraService::CreateCaptureSession(sptr<ICaptureSession> &session){ CAMERA_SYNC_TRACE; sptr<HCaptureSession> captureSession; if (streamOperatorCallback_ == nullptr) { streamOperatorCallback_ = new(std::nothrow) StreamOperatorCallback(); if (streamOperatorCallback_ == nullptr) { MEDIA_ERR_LOG("HCameraService::CreateCaptureSession streamOperatorCallback_ allocation failed"); return CAMERA_ALLOC_ERROR; } } std::lock_guard<std::mutex> lock(mutex_); OHOS::Security::AccessToken::AccessTokenID callerToken = IPCSkeleton::GetCallingTokenID(); captureSession = new(std::nothrow) HCaptureSession(cameraHostManager_, streamOperatorCallback_, callerToken); if (captureSession == nullptr) { MEDIA_ERR_LOG("HCameraService::CreateCaptureSession HCaptureSession allocation failed"); return CAMERA_ALLOC_ERROR; } session = captureSession; return CAMERA_OK;}开始配置采集会话
调用CaptureSession的BeginConfig进行采集会话的配置工作。这个工作最终调用到被封装的HCaptureSession中。int32_t HCaptureSession::BeginConfig(){ CAMERA_SYNC_TRACE; if (curState_ == CaptureSessionState::SESSION_CONFIG_INPROGRESS) { MEDIA_ERR_LOG("HCaptureSession::BeginConfig Already in config inprogress state!"); return CAMERA_INVALID_STATE; } std::lock_guard<std::mutex> lock(sessionLock_); prevState_ = curState_; curState_ = CaptureSessionState::SESSION_CONFIG_INPROGRESS; tempCameraDevices_.clear(); tempStreams_.clear(); deletedStreamIds_.clear(); return CAMERA_OK;}创立CameraInput
应用层通过camManagerObj->CreateCameraInput(cameraObjList[0])的形式进行CameraInput的创立，cameraObjList[0]就是后面获取反对设施的第一个。依据CameraDevice创立对应的CameraInput对象。sptr<CameraInput> CameraManager::CreateCameraInput(sptr<CameraDevice> &camera){ CAMERA_SYNC_TRACE; sptr<CameraInput> cameraInput = nullptr; sptr<ICameraDeviceService> deviceObj = nullptr; if (camera != nullptr) { deviceObj = CreateCameraDevice(camera->GetID()); if (deviceObj != nullptr) { cameraInput = new(std::nothrow) CameraInput(deviceObj, camera); if (cameraInput == nullptr) { MEDIA_ERR_LOG("failed to new CameraInput Returning null in CreateCameraInput"); return cameraInput; } } else { MEDIA_ERR_LOG("Returning null in CreateCameraInput"); } } else { MEDIA_ERR_LOG("CameraManager::CreateCameraInput: Camera object is null"); } return cameraInput;}- 开启CameraInput
调用了CameraInput的Open办法，进行输出设施的启动关上。
void CameraInput::Open(){ int32_t retCode = deviceObj_->Open(); if (retCode != CAMERA_OK) { MEDIA_ERR_LOG("Failed to open Camera Input, retCode: %{public}d", retCode); }}增加CameraInput实例到采集会话中
通过调用captureSession的AddInput办法，将创立的CameraInput对象增加到采集会话的输出中，这样采集会话就晓得采集输出的设施。int32_t CaptureSession::AddInput(sptr<CaptureInput> &input){ CAMERA_SYNC_TRACE; if (input == nullptr) { MEDIA_ERR_LOG("CaptureSession::AddInput input is null"); return CAMERA_INVALID_ARG; } input->SetSession(this); inputDevice_ = input; return captureSession_->AddInput(((sptr<CameraInput> &)input)->GetCameraDevice());}
最终调用到HCaptureSession的AddInput办法，该办法中外围的代码是tempCameraDevices_.emplace_back(localCameraDevice)，将须要增加的CameraDevice插入到tempCameraDevices_容器中。
int32_t HCaptureSession::AddInput(sptr<ICameraDeviceService> cameraDevice){ CAMERA_SYNC_TRACE; sptr<HCameraDevice> localCameraDevice = nullptr; if (cameraDevice == nullptr) { MEDIA_ERR_LOG("HCaptureSession::AddInput cameraDevice is null"); return CAMERA_INVALID_ARG; } if (curState_ != CaptureSessionState::SESSION_CONFIG_INPROGRESS) { MEDIA_ERR_LOG("HCaptureSession::AddInput Need to call BeginConfig before adding input"); return CAMERA_INVALID_STATE; } if (!tempCameraDevices_.empty() || (cameraDevice_ != nullptr && !cameraDevice_->IsReleaseCameraDevice())) { MEDIA_ERR_LOG("HCaptureSession::AddInput Only one input is supported"); return CAMERA_INVALID_SESSION_CFG; } localCameraDevice = static_cast<HCameraDevice*>(cameraDevice.GetRefPtr()); if (cameraDevice_ == localCameraDevice) { cameraDevice_->SetReleaseCameraDevice(false); } else { tempCameraDevices_.emplace_back(localCameraDevice); CAMERA_SYSEVENT_STATISTIC(CreateMsg("CaptureSession::AddInput")); } sptr<IStreamOperator> streamOperator; int32_t rc = localCameraDevice->GetStreamOperator(streamOperatorCallback_, streamOperator); if (rc != CAMERA_OK) { MEDIA_ERR_LOG("HCaptureSession::GetCameraDevice GetStreamOperator returned %{public}d", rc); localCameraDevice->Close(); return rc; } return CAMERA_OK;}创立Video的Surface
通过Surface::CreateSurfaceAsConsumer创立Surface。sptr<Surface> Surface::CreateSurfaceAsConsumer(std::string name, bool isShared){ sptr<ConsumerSurface> surf = new ConsumerSurface(name, isShared); GSError ret = surf->Init(); if (ret != GSERROR_OK) { BLOGE("Failure, Reason: consumer surf init failed"); return nullptr; } return surf;}创立VideoOutput实例
通过调用CameraManager的CreateVideoOutput来创立VideoOutput实例。sptr<VideoOutput> CameraManager::CreateVideoOutput(VideoProfile &profile, sptr<Surface> &surface){ CAMERA_SYNC_TRACE; sptr<IStreamRepeat> streamRepeat = nullptr; sptr<VideoOutput> result = nullptr; int32_t retCode = CAMERA_OK; camera_format_t metaFormat; metaFormat = GetCameraMetadataFormat(profile.GetCameraFormat()); retCode = serviceProxy_->CreateVideoOutput(surface->GetProducer(), metaFormat, profile.GetSize().width, profile.GetSize().height, streamRepeat); if (retCode == CAMERA_OK) { result = new(std::nothrow) VideoOutput(streamRepeat); if (result == nullptr) { MEDIA_ERR_LOG("Failed to new VideoOutput"); } else { std::vector<int32_t> videoFrameRates = profile.GetFrameRates(); if (videoFrameRates.size() >= 2) { // vaild frame rate range length is 2 result->SetFrameRateRange(videoFrameRates[0], videoFrameRates[1]); } POWERMGR_SYSEVENT_CAMERA_CONFIG(VIDEO, profile.GetSize().width, profile.GetSize().height); } } else { MEDIA_ERR_LOG("VideoOutpout: Failed to get stream repeat object from hcamera service! %{public}d", retCode); } return result;}该办法中通过IPC的调用最终调用到了HCameraService的CreateVideoOutput(surface->GetProducer(), format, streamRepeat)。
sptr<VideoOutput> CameraManager::CreateVideoOutput(VideoProfile &profile, sptr<Surface> &surface){ CAMERA_SYNC_TRACE; sptr<IStreamRepeat> streamRepeat = nullptr; sptr<VideoOutput> result = nullptr; int32_t retCode = CAMERA_OK; camera_format_t metaFormat; metaFormat = GetCameraMetadataFormat(profile.GetCameraFormat()); retCode = serviceProxy_->CreateVideoOutput(surface->GetProducer(), metaFormat, profile.GetSize().width, profile.GetSize().height, streamRepeat); if (retCode == CAMERA_OK) { result = new(std::nothrow) VideoOutput(streamRepeat); if (result == nullptr) { MEDIA_ERR_LOG("Failed to new VideoOutput"); } else { std::vector<int32_t> videoFrameRates = profile.GetFrameRates(); if (videoFrameRates.size() >= 2) { // vaild frame rate range length is 2 result->SetFrameRateRange(videoFrameRates[0], videoFrameRates[1]); } POWERMGR_SYSEVENT_CAMERA_CONFIG(VIDEO, profile.GetSize().width, profile.GetSize().height); } } else { MEDIA_ERR_LOG("VideoOutpout: Failed to get stream repeat object from hcamera service! %{public}d", retCode); } return result;}HCameraService的CreateVideoOutput办法中次要创立了HStreamRepeat，并且通过参数传递给后面的CameraManager应用，CameraManager通过传递的HStreamRepeat对象，进行封装，创立出VideoOutput对象。
- 增加videoOutput到采集会话中，并且提交采集会话
该步骤相似增加CameraInput到采集会话的过程，能够参考后面的流程。 开始录制
通过调用VideoOutput的Start进行录制的操作。int32_t VideoOutput::Start(){ return static_cast<IStreamRepeat *>(GetStream().GetRefPtr())->Start();}
该办法中会调用到HStreamRepeat的Start办法。
int32_t HStreamRepeat::Start(){ CAMERA_SYNC_TRACE; if (streamOperator_ == nullptr) { return CAMERA_INVALID_STATE; } if (curCaptureID_ != 0) { MEDIA_ERR_LOG("HStreamRepeat::Start, Already started with captureID: %{public}d", curCaptureID_); return CAMERA_INVALID_STATE; } int32_t ret = AllocateCaptureId(curCaptureID_); if (ret != CAMERA_OK) { MEDIA_ERR_LOG("HStreamRepeat::Start Failed to allocate a captureId"); return ret; } std::vector<uint8_t> ability; OHOS::Camera::MetadataUtils::ConvertMetadataToVec(cameraAbility_, ability); CaptureInfo captureInfo; captureInfo.streamIds_ = {streamId_}; captureInfo.captureSetting_ = ability; captureInfo.enableShutterCallback_ = false; MEDIA_INFO_LOG("HStreamRepeat::Start Starting with capture ID: %{public}d", curCaptureID_); CamRetCode rc = (CamRetCode)(streamOperator_->Capture(curCaptureID_, captureInfo, true)); if (rc != HDI::Camera::V1_0::NO_ERROR) { ReleaseCaptureId(curCaptureID_); curCaptureID_ = 0; MEDIA_ERR_LOG("HStreamRepeat::Start Failed with error Code:%{public}d", rc); ret = HdiToServiceError(rc); } return ret;}外围的代码是streamOperator_->Capture，其中最初一个参数true，示意采集间断数据。
- 录制完结，保留录制文件
六、总结
本文次要对OpenHarmony 3.2 Beta多媒体子系统的视频录制进行介绍，首先梳理了整体的录制流程，而后对录制过程中的次要步骤进行了具体地剖析。视频录制次要分为以下几个步骤：
（1） 获取CameraManager实例。
（2） 创立采集会话CaptureSession。
（3） 创立CameraInput实例，并且将输出设施增加到CaptureSession中。
（4） 创立Video录制须要的Surface。
（5） 创立VideoOutput实例，并且将输入增加到CaptureSession中。
（6） 提交采集会话的配置。
（7） 调用VideoOutput的Start办法，进行视频的录制。
（8） 录制完结，保留录制的文件。
对于OpenHarmony 3.2 Beta多媒体系列开发，我之前还分享过
《OpenHarmony 3.2 Beta源码剖析之MediaLibrary》
《OpenHarmony 3.2 Beta多媒体系列——音视频播放框架》
《OpenHarmony 3.2 Beta多媒体系列——音视频播放gstreamer》
这几篇文章，欢送感兴趣的开发者进行浏览。