Camera2 同时打开两个摄像头Demo讲解
背景
首先要指出，虽然强大的谷歌已经兼容了同时打开两个物理相机的场景，但并不是所有机器都具备这样的功能，这个还跟平台有这紧密的关系，这里面可能是ISP的处理能力的关系，本人目前参与做了qcom、MTK、sprd平台的相机开发，但是只有高通支持这样的功能，原因就在于使用的高通平台拥有两个强大的ISP处理器，服服服。。。
同时打开多个摄像头谷歌在官网中已经描述了原理，这里就不在赘述了，附上一张总结性的原理图片，去那边看看吧。
谷歌官网：多摄像头支持 | Android 开源项目 | Android Open Source Project
强大的谷歌让开发成了傻子，同时打开多摄像头也是这样，基本可以理解所有的操作全部double一下就行，主要还是处理线程问题，一个线程尝试过很吃力，然后整两个线程呗！废话不多说，介绍demo 开干！
一、初始化变量
记住，double就行，当然大神不必跟从，你肯定有更好的办法；创建了两个textureView用于接收预览回调数据，两个imageview用于拍照回调显示，分别给两个相机设置一个线程，这样可以分别打开camera1和camera2，避免阻塞挂掉。
private ImageView image_view1;
private ImageView image_view2;
private TextureView textureview1;
private TextureView textureview2;
private ImageReader imagereader1;
private ImageReader imagereader2;
private Surface textureview1_surface;
private Surface textureview2_surface;
private Surface imagereader1_surface;
private Surface imagereader2_surface;
private SurfaceTexture texture_surfacetextture1;
private SurfaceTexture texture_surfacetextture2;
private CameraManager mcameramanager;
private CameraDevice mCameraDevice;
private CameraDevice mCameraDevice1;
private CameraCaptureSession mcamerasession0;
private CameraCaptureSession mcamerasession1;
Button camera1_button;
Button camera2_button;
private HandlerThread mBackgroundforcamera0Thread;
private Handler mBackgroundforcamera0Handler;
private HandlerThread mBackgroundforcamera1Thread;
private Handler mBackgroundforcamera1Handler;
这里面的逻辑就是配置UI，然后获取cameramanager，直接开看textureView的监视器，这俩面分别实例化可一个surface空间，并把预览分辨率设置为1920*1080，以及创建了imagereader，同时设置jpeg的分辨率，且实例化imagereader的surface；最后open_camera()。textureListener2和textureListener1完全是一样的，不表述。
mBackgroundforcamera0Thread = new HandlerThread("CameraBackgroundForCamera0");
mBackgroundforcamera0Thread.start();
mBackgroundforcamera0Handler = new Handler(mBackgroundforcamera0Thread.getLooper());
//创立线程2
mBackgroundforcamera1Thread = new HandlerThread("CameraBackgroundForCamera1");
mBackgroundforcamera1Thread.start();
mBackgroundforcamera1Handler = new Handler(mBackgroundforcamera1Thread.getLooper());
image_view1 = findViewById(R.id.imageView);
image_view2 = findViewById(R.id.imageView2);
textureview1 = findViewById(R.id.textureView);
textureview2 = findViewById(R.id.textureView2);
//按键拍照
camera1_button = findViewById(R.id.button);
camera2_button = findViewById(R.id.button_2);
//按键拍照
camera1_button = findViewById(R.id.button);
camera2_button = findViewById(R.id.button_2);
mcameramanager = (CameraManager) getSystemService(Context.CAMERA_SERVICE);
textureview1.setSurfaceTextureListener(textureListener1);
textureview2.setSurfaceTextureListener(textureListener2);
private final TextureView.SurfaceTextureListener textureListener1 = new TextureView.SurfaceTextureListener() {
@Override
public void onSurfaceTextureAvailable(@NonNull SurfaceTexture surface, int width, int height) {
texture_surfacetextture1 = textureview1.getSurfaceTexture();
texture_surfacetextture1.setDefaultBufferSize(1920, 1080);
textureview1_surface = new Surface(texture_surfacetextture1);
//创建imagereader
imagereader1 = ImageReader.newInstance(1920, 1080, ImageFormat.JPEG, 2);
imagereader1_surface = imagereader1.getSurface();
imagereader1.setOnImageAvailableListener(imagereader_available1, null);
Log.i(TAG, "opencamera1 start!");
open_camera("0");
Log.i(TAG, "opencamera1 finish!");
}
};
二、打开相机
记得在这里做好区分就行，还有获取camera权限；
private final void open_camera(String cameraId) {
try {
if (ActivityCompat.checkSelfPermission(this, Manifest.permission.CAMERA) != PackageManager.PERMISSION_GRANTED) {
requestPermissions(new String[]{Manifest.permission.CAMERA}, 100);
Log.i(TAG,"[open_camera] 正在申请权限");
return;
} else {
Log.i(TAG,"[open_camera] 已经获取权限");
}
CameraId = cameraId;
Log.i(TAG,"opencamera id is :"+cameraId);
if (cameraId == "0") {
mcameramanager.openCamera(cameraId, mStateCallback, mBackgroundforcamera0Handler);
} else {
mcameramanager.openCamera(cameraId, mStateCallback2, mBackgroundforcamera1Handler);
}
} catch(CameraAccessException e) {
e.printStackTrace();
}
};
三、创建session
从上面的openCamera()函数可以看出，针对不同的cameraId，创建不同的回调，或许这就是傻瓜式开发的我吧，莫学。createCameraPreviewSession就是创建session函数，我们进入其中。当然再来句废话，mStateCallback2和mStateCallback是一样的；
private final CameraDevice.StateCallback mStateCallback2 = new CameraDevice.StateCallback() {
@Override
public void onOpened(@NonNull CameraDevice camera) {
mCameraDevice1 = camera;
Log.i(TAG,"[open_camera:1] get CameraDevice! cameraId:"+mCameraDevice1);
createCameraPreviewSession("1");
}
};
这里面的核心思想就是针对不同cameraId，用不同的cameradevice和各种surface，创建session，直接过；然后这里你又会看到傻瓜式的msessionstatecallback1和msessionstatecallback0，哈哈；
private void createCameraPreviewSession( String cameraId) {
Surface previewsurface = null;
Surface imagesurface = null;
if (cameraId == "0") {
try {
Log.i(TAG,"[------createCameraPreviewSession-------] choose surface:"+previewsurface.toString()
+" imagesurface:"+imagesurface.toString()+" device:"+McameraDevice.toString());
mCameraDevice.createCaptureSession(Arrays.asList(textureview1_surface,imagereader1_surface),msessionstatecallback0,mBackgroundforcamera0Handler);
} catch (CameraAccessException e) {
e.printStackTrace();
}
} else {
try {
Log.i(TAG,"[+++++++createCameraPreviewSession++++++++] choose surface:"+textureview2_surface.toString()
+" imagesurface:"+imagereader2_surface.toString()+" device:"+mCameraDevice1.toString());
mCameraDevice1.createCaptureSession(Arrays.asList(textureview2_surface,imagereader2_surface),msessionstatecallback1,mBackgroundforcamera1Handler);
} catch (CameraAccessException e) {
e.printStackTrace();
}
};
四、创建预览Request
记得添加的surface和创建的session中的surface对应就行，一般操作；至于预览回调，无所谓，不用写，整个null；
CameraCaptureSession.StateCallback msessionstatecallback0 = new CameraCaptureSession.StateCallback() {
@Override
public void onConfigured(@NonNull CameraCaptureSession session) {
try {
Log.i(TAG,"[createCameraPreviewSession] create session successful for camera: 0");
CaptureRequest.Builder PreviewRequest_builder = mCameraDevice.createCaptureRequest(CameraDevice.TEMPLATE_PREVIEW);
PreviewRequest_builder.addTarget(textureview1_surface);
//创建request并开始预览
CaptureRequest request = PreviewRequest_builder.build();
Log.i(TAG,"[camerasession] begin preview!---"+session);
session.setRepeatingRequest(request, mPreviewCallback, mBackgroundforcamera0Handler);
mcamerasession0 = session;
Log.i(TAG,"[camerasession] send preview request!");
} catch (CameraAccessException e) {
e.printStackTrace();
}
}
}；
五、创建CaptureRequest
给button创建click事件，然后点击拍照，注意的是要将拍照的回调数据添加到imagereader中，不赘述，最后在imagereader回调中将我们早期创建imageview利用上，要把回调的数据压缩成bitmap并设置到imageview中，就像下面。
private void take_camera1() throws CameraAccessException {
Log.i(TAG,"开始camera0拍照");
CaptureRequest.Builder picture_request_builder = mCameraDevice.createCaptureRequest(CameraDevice.TEMPLATE_STILL_CAPTURE);
picture_request_builder.addTarget(imagereader1_surface);
picture_request_builder.set(CaptureRequest.CONTROL_MODE, CaptureRequest.CONTROL_MODE_AUTO);
CaptureRequest picture_reuqest = picture_request_builder.build();
//session
mcamerasession0.capture(picture_reuqest, mPreviewCallback,mBackgroundforcamera0Handler);
Log.i(TAG,"camera0 完成拍照，开始图片回调");
}
private final ImageReader.OnImageAvailableListener imagereader_available1
= new ImageReader.OnImageAvailableListener() {
@Override
public void onImageAvailable(ImageReader reader) {
//TODO zhb
Log.i(TAG,"camera 0 回调已经获取到imagereader 图片");
Image image = reader.acquireLatestImage();
ByteBuffer buffer = image.getPlanes()[0].getBuffer();
int number_buffer = buffer.remaining();
byte[] bytes = new byte[number_buffer];
buffer.get(bytes);
image.close();
Bitmap bitmap = BitmapFactory.decodeByteArray(bytes, 0, number_buffer);
Log.i(TAG,"camera 0 完成bitmap生成");
image_view1.setImageBitmap(bitmap);
}
};
总结
其实demo就是在实践验证同时打开两个相机的功能，以上代码不可能ctrl+c和ctrl+v就能够跑起来，你们按照自己工程，double打开一个相机的流程就行，当然前提是机器支持这样的功能。最后附上一张图验证结果。
（谢谢大家，欢迎指正）