android6.0源码分析之Camera API2.0下的Capture流程分析

Posted 2022-11-24 古冥

前面对Camera2的初始化以及预览的相关流程进行了详细分析，本文将会对Camera2的capture(拍照)流程进行分析。

Camera2相关文章的目录如下：
android6.0源码分析之Camera API2.0简介
android6.0源码分析之Camera2 HAL分析
android6.0源码分析之Camera API2.0下的初始化流程分析
android6.0源码分析之Camera API2.0下的Preview(预览)流程分析
android6.0源码分析之Camera API2.0下的Capture流程分析
android6.0源码分析之Camera API2.0下的video流程分析
Camera API2.0的应用

---

前面分析preview的时候，当预览成功后，会使能ShutterButton，即可以进行拍照，定位到ShutterButton的监听事件为onShutterButtonClick方法：

//CaptureModule.java
@Override
public void onShutterButtonClick() 
    //Camera未打开
    if (mCamera == null) 
        return;
    

    int countDownDuration = mSettingsManager.getInteger(SettingsManager
        .SCOPE_GLOBAL,Keys.KEY_COUNTDOWN_DURATION);
    if (countDownDuration > 0) 
        // 开始倒计时
        mAppController.getCameraAppUI().transitionToCancel();
        mAppController.getCameraAppUI().hideModeOptions();
        mUI.setCountdownFinishedListener(this);
        mUI.startCountdown(countDownDuration);
        // Will take picture later via listener callback.
     else 
        //即刻拍照
        takePictureNow();
    

首先，读取Camera的配置，判断配置是否需要延时拍照，此处分析不需延时的情况，即调用takePictureNow方法：

//CaptureModule.java
private void takePictureNow() 
    if (mCamera == null) 
        Log.i(TAG, "Not taking picture since Camera is closed.");
        return;
    
    //创建Capture会话并开启会话
    CaptureSession session = createAndStartCaptureSession();
    //获取Camera的方向
    int orientation = mAppController.getOrientationManager()
        .getDeviceOrientation().getDegrees();
    //初始化图片参数
    PhotoCaptureParameters params = new PhotoCaptureParameters(
            session.getTitle(), orientation, session.getLocation(),
            mContext.getExternalCacheDir(), this, mPictureSaverCallback,
            mHeadingSensor.getCurrentHeading(), mZoomValue, 0);
    //装配Session
    decorateSessionAtCaptureTime(session);
    //拍照
    mCamera.takePicture(params, session);

它首先调用createAndStartCaptureSession来创建一个CaptureSession并且启动会话,这里并且会进行初始参数的设置，譬如设置CaptureModule(此处实参为this)为图片处理的回调(后面再分析)：

//CaptureModule.java
private CaptureSession createAndStartCaptureSession() 
    //获取会话时间
    long sessionTime = getSessionTime();
    //当前位置
    Location location = mLocationManager.getCurrentLocation();
    //设置picture name
    String title = CameraUtil.instance().createJpegName(sessionTime);
    //创建会话
    CaptureSession session = getServices().getCaptureSessionManager()
           .createNewSession(title, sessionTime, location);
    //开启会话
    session.startEmpty(new CaptureStats(mHdrPlusEnabled),new Size(
        (int) mPreviewArea.width(), (int) mPreviewArea.height()));
    return session;

首先，获取会话的相关参数，包括会话时间，拍照的照片名字以及位置信息等，然后调用Session管理来创建CaptureSession，最后将此CaptureSession启动。到这里，会话就创建并启动了，所以接着分析上面的拍照流程，它会调用OneCameraImpl的takePicture方法来进行拍照：

//OneCameraImpl.java
@Override
public void takePicture(final PhotoCaptureParameters params, final CaptureSession session) 
    ...
    // 除非拍照已经返回，否则就广播一个未准备好状态的广播，即等待本次拍照结束
    broadcastReadyState(false);
    //创建一个线程
    mTakePictureRunnable = new Runnable() 
        @Override
        public void run() 
            //拍照
            takePictureNow(params, session);
        
    ;
    //设置回调，此回调后面将分析，它其实就是CaptureModule,它实现了PictureCallback
    mLastPictureCallback = params.callback;
    mTakePictureStartMillis = SystemClock.uptimeMillis();

    //如果需要自动聚焦
    if (mLastResultAFState == AutoFocusState.ACTIVE_SCAN) 
        mTakePictureWhenLensIsStopped = true;
     else 
        //拍照
        takePictureNow(params, session);
    

在拍照里，首先广播一个未准备好的状态广播，然后进行拍照的回调设置，并且判断是否有自动聚焦，如果是则将mTakePictureWhenLensIsStopped 设为ture，即即刻拍照被停止了，否则则调用OneCameraImpl的takePictureNow方法来发起拍照请求：

//OneCameraImpl.java
public void takePictureNow(PhotoCaptureParameters params, CaptureSession 
        session) 
    long dt = SystemClock.uptimeMillis() - mTakePictureStartMillis;
    try 
        // 构造JPEG图片拍照的请求
        CaptureRequest.Builder builder = mDevice.createCaptureRequest(
            CameraDevice.TEMPLATE_STILL_CAPTURE);
        builder.setTag(RequestTag.CAPTURE);
        addBaselineCaptureKeysToRequest(builder);

        // Enable lens-shading correction for even better DNGs.
        if (sCaptureImageFormat == ImageFormat.RAW_SENSOR) 
            builder.set(CaptureRequest.STATISTICS_LENS_SHADING_MAP_MODE,
                CaptureRequest.STATISTICS_LENS_SHADING_MAP_MODE_ON);
         else if (sCaptureImageFormat == ImageFormat.JPEG) 
            builder.set(CaptureRequest.JPEG_QUALITY, JPEG_QUALITY);
                .getJpegRotation(params.orientation, mCharacteristics));
        
        //用于preview的控件
        builder.addTarget(mPreviewSurface);
        //用于图片显示的控件
        builder.addTarget(mCaptureImageReader.getSurface());
        CaptureRequest request = builder.build();

        if (DEBUG_WRITE_CAPTURE_DATA) 
            final String debugDataDir = makeDebugDir(params.debugDataFolder,
                        "normal_capture_debug");
            Log.i(TAG, "Writing capture data to: " + debugDataDir);
            CaptureDataSerializer.toFile("Normal Capture", request, 
                new File(debugDataDir,"capture.txt"));
        
        //拍照，mCaptureCallback为回调
        mCaptureSession.capture(request, mCaptureCallback, mCameraHandler);
     catch (CameraAccessException e) 
        Log.e(TAG, "Could not access camera for still image capture.");
        broadcastReadyState(true);
        params.callback.onPictureTakingFailed();
        return;
    
    synchronized (mCaptureQueue) 
        mCaptureQueue.add(new InFlightCapture(params, session));
    

与preview类似，都是通过CaptureRequest来与Camera进行通信的，通过session的capture来进行拍照，并设置拍照的回调函数为mCaptureCallback：

//CameraCaptureSessionImpl.java
@Override
public synchronized int capture(CaptureRequest request,CaptureCallback callback,Handler handler)throws CameraAccessException
    ...
    handler = checkHandler(handler,callback);
    return addPendingSequence(mDeviceImpl.capture(request,createCaptureCallbackProxy(
        handler,callback),mDeviceHandler));

代码与preview中的类似，都是将请求加入到待处理的请求集，现在看CaptureCallback回调：

//OneCameraImpl.java
private final CameraCaptureSession.CaptureCallback mCaptureCallback = new CameraCaptureSession.CaptureCallback()
    @Override
    public void onCaptureStarted(CameraCaptureSession session,CaptureRequest request,long 
            timestamp,long frameNumber)
　　　　　//与preview类似
        if(request.getTag() == RequestTag.CAPTURE&&mLastPictureCallback!=null)
            mLastPictureCallback.onQuickExpose();
        
    
    ...
    @Override
    public void onCaptureCompleted(CameraCaptureSession session,CaptureRequest request
            ,TotalCaptureResult result)
        autofocusStateChangeDispatcher(result);
        if(result.get(CaptureResult.CONTROL_AF_STATE) == null)
　　　　　　　//检查自动聚焦的状态
            AutoFocusHelper.checkControlAfState(result);
        
        ...
        if(request.getTag() == RequestTag.CAPTURE)
            synchronized(mCaptureQueue)
                if(mCaptureQueue.getFirst().setCaptureResult(result).isCaptureComplete())
                    capture = mCaptureQueue.removeFirst();
                
            
            if(capture != null)
　　　　　　　　　//拍照结束
                OneCameraImpl.this.onCaptureCompleted(capture);
            
        
        super.onCaptureCompleted(session,request,result);
    
    ...

这是Native层在处理请求时，会调用相应的回调，如capture开始时，会回调onCaptureStarted,具体的在preview中有过分析，当拍照结束时，会回调onCaptureCompleted方法，其中会根据CaptureResult来检查自动聚焦的状态，并通过TAG判断其是Capture动作时，再来看它是否是队列中的第一个请求，如果是，则将请求移除，因为请求已经处理成功，最后再调用OneCameraImpl的onCaptureCompleted方法来进行处理：

//OneCameraImpl.java
private void onCaptureCompleted(InFlightCapture capture)
    if(isCaptureImageFormat == ImageFormat.RAW_SENSOR)
        ...
        File dngFile = new File(RAW_DIRECTORY,capture.session.getTitle()+".dng");
        writeDngBytesAndClose(capture.image,capture.totalCaptureResult,mCharacteristics,dngFile);
    else
        //解析result中的图片数据
        byte[] imageBytes = acquireJpegBytesAndClose(capture.image);
        //保存Jpeg图片
        saveJpegPicture(imageBytes,capture.parameters,capture.session,capture.totalCaptureResult);
    
    broadcastReadyState(true);
    //调用回调
    capture.parameters.callback.onPictureTaken(capture.session);

如代码所示，首先，对result中的图片数据进行了解析，然后调用saveJpegPicture方法将解析得到的图片数据进行保存，最后再调用里面的回调(即CaptureModule，前面在初始化Parameters时说明了，它实现了PictureCallbak接口)的onPictureTaken方法，所以，接下来先分析saveJpegPicture方法：

//OneCameraImpl.java
private void saveJpegPicture(byte[] jpegData,final PhotoCaptureParameters captureParams,CaptureSession session,CaptureResult result)
    ...
    ListenableFuture<Optional<Uri>> futureUri = session.saveAndFinish(jpegData,width,
            height,rotation,exif);
    Futures.addCallback(futureUri,new FutureCallback<Optional<Uri>>()
        @Override
        public void onSuccess(Optional<Uri> uriOptional)
            captureParams.callback.onPictureSaved(mOptional.orNull());
        

        @Override
        public void onFailure(Throwable throwable)
            captureParams.callback.onPictureSaved(null);
        
    );

它最后会回调onPictureSaved方法来对图片进行保存，所以需要分析CaptureModule的onPictureSaved方法：

//CaptureModule.java
@Override
public void onPictureSaved(Uri uri)
    mAppController.notifyNewMedia(uri);

mAppController的实现为CameraActivity，所以分析notifyNewMedia方法：

//CameraActivity.java
@Override
public void notifyNewMedia(Uri uri)
    ...
    if(FilmstripItemUtils.isMimeTypeVideo(mimeType))
　　　　//如果拍摄的是video
        sendBroadcast(new Intent(CameraUtil.ACTION_NEW_VIDEO,uri));
        newData = mVideoItemFactory.queryContentUri(uri);
        ...
    else if(FilmstripItemUtils.isMimeTypeImage(mimeType))
　　　　//如果是拍摄图片
        CameraUtil.broadcastNewPicture(mAppContext,uri);
        newData = mPhotoItemFactory.queryCotentUri(uri);
        ...
    else
        return;
    
    new AsyncTask<FilmstripItem,Void,FilmstripItem>()
        @Override
        protected FilmstripItem doInBackground(FilmstripItem... Params)
            FilmstripItem data = params[0];
            MetadataLoader.loadMetadata(getandroidContet(),data);
            return data;
        
        ...
    

由代码可知，这里有两种数据的处理，一种是video，另一种是image。而我们这里分析的是capture图片数据，所以首先会根据在回调函数传入的参数Uri和PhotoItemFactory来查询到相应的拍照数据，然后再开启一个异步的Task来对此数据进行处理，即通过MetadataLoader的loadMetadata来加载数据，并返回。至此，capture的流程就基本分析结束了，下面将给出capture流程的整个过程中的时序图：

---