在Android上设置OpenCV的相机场景模式

Question

我正在寻找一种方法来为C++（Android NDK）中的网络相机设置场景模式值。

OpenCV在初始化时从网络摄像头获取值。 在下面的日志中，日志说Supported Scene Modes: auto,action,portrait,landscape,beach,fireworks,night,night-portrait,snow,sports,steadyphoto,sunset,theatre,barcode。

我想为相机设置其中一个值（如海滩）。

D/OpenCV::camera﹕ CvCapture_Android::CvCapture_Android(1) 
D/OpenCV::camera﹕ Library name: libopencv_java.so 
D/OpenCV::camera﹕ Library base address: 0x69f3b000 
D/OpenCV::camera﹕ Libraries folder found: /data/app-lib/gms.rewards.selfie-2/ 
D/OpenCV::camera﹕ CameraWrapperConnector::connectToLib: folderPath=/data/app-lib/gms.rewards.selfie-2/ 
E/OpenCV::camera﹕ ||libnative_camera_r4.1.1.so 
E/OpenCV::camera﹕ ||libnative_camera_r4.0.3.so 
E/OpenCV::camera﹕ ||libnative_camera_r2.3.3.so 
E/OpenCV::camera﹕ ||libnative_camera_r4.2.0.so 
E/OpenCV::camera﹕ ||libnative_camera_r2.2.0.so 
E/OpenCV::camera﹕ ||libnative_camera_r3.0.1.so 
E/OpenCV::camera﹕ ||libnative_camera_r4.3.0.so 
E/OpenCV::camera﹕ ||libnative_camera_r4.0.0.so 
E/OpenCV::camera﹕ ||libnative_camera_r4.4.0.so 
D/OpenCV::camera﹕ try to load library 'libnative_camera_r4.4.0.so' 
D/OpenCV::camera﹕ CameraWrapperConnector::connectToLib ERROR: cannot dlopen camera wrapper library /data/app-lib/gms.rewards.selfie-2/libnative_camera_r4.4.0.so, dlerror="Cannot load library: soinfo_relocate(linker.cpp:975): cannot locate symbol "_ZN7android17IConsumerListenerD2Ev" referenced by "libnative_camera_r4.4.0.so"..." 
D/OpenCV::camera﹕ try to load library 'libnative_camera_r4.3.0.so' 
D/OpenCV::camera﹕ CameraWrapperConnector::connectToLib ERROR: cannot dlopen camera wrapper library /data/app-lib/gms.rewards.selfie-2/libnative_camera_r4.3.0.so, dlerror="Cannot load library: soinfo_relocate(linker.cpp:975): cannot locate symbol "_ZN7android6Camera17setPreviewTextureERKNS_2spINS_22IGraphicBufferProducerEEE" referenced by "libnative_camera_r4.3.0.so"..." 
D/OpenCV::camera﹕ try to load library 'libnative_camera_r4.2.0.so' 
D/OpenCV::camera﹕ Loaded library '/data/app-lib/gms.rewards.selfie-2/libnative_camera_r4.2.0.so' 
D/OpenCV_NativeCamera﹕ CameraHandler::initCameraConnect(0x6a64eb6d, 1, 0x69359e18, 0x0) 
D/OpenCV_NativeCamera﹕ Connecting to CameraService v 2.3 
D/OpenCV_NativeCamera﹕ Instantiated new CameraHandler (0x6a64eb6d, 0x69359e18) 
I/OpenCV_NativeCamera﹕ initCameraConnect: [antibanding=auto;antibanding-values=off,auto,50hz,60hz;auto-exposure-lock=false;auto-exposure-lock-supported=true;auto-exposure-lock-values=false,true;auto-whitebalance-lock=false;auto-whitebalance-lock-supported=true;auto-whitebalance-lock-values=false,true;effect=none;effect-values=mono,negative,none,posterize,sepia,aqua,solarize,nv-vivid,nv-emboss;exposure-compensation-step=0.1;flash-mode=off;flash-mode-values=off,on,auto,torch,red-eye;focal-length=1.000;focus-areas=(0,0,0,0,0);focus-distances=0.95,1.9,Infinity;focus-mode=fixed;focus-mode-values=fixed;horizontal-view-angle=60.000;jpeg-quality=95;jpeg-thumbnail-height=240;jpeg-thumbnail-quality=90;jpeg-thumbnail-size-values=0x0,320x240,240x320;jpeg-thumbnail-width=320;max-exposure-compensation=20;max-num-detected-faces-hw=10;max-num-focus-areas=0;max-num-metering-areas=4;max-zoom=28;metering-areas=(0,0,0,0,0);min-exposure-compensation=-20;picture-format=jpeg;picture-format-values=jpeg,jfif,exif,yuv420p,yuv420sp;picture-size=1280x960;pic 
D/OpenCV_NativeCamera﹕ Supported Cameras: (null) 
D/OpenCV_NativeCamera﹕ Supported Picture Sizes: 320x240,480x480,640x368,640x480,800x600,960x720,1024x768,1280x720,1280x752,1280x960,1440x1080,1600x1200,1836x1080,1920x1080,2048x1152,2048x1360,2048x1536,2592x1456,2592x1520,2592x1920,2592x1944 
D/OpenCV_NativeCamera﹕ Supported Picture Formats: jpeg,jfif,exif,yuv420p,yuv420sp 
D/OpenCV_NativeCamera﹕ Supported Preview Sizes: 176x144,320x240,352x288,480x480,640x480,704x576,720x408,720x480,720x576,768x432,800x448,960x720,1280x720,1280x752,1280x960,1360x720,1440x1080,1920x1080,1920x1088,1920x1440 
D/OpenCV_NativeCamera﹕ Supported Preview Formats: yuv420p,yuv420sp 
D/OpenCV_NativeCamera﹕ Supported Preview Frame Rates: 5,8,10,15,20,24,25,30 
D/OpenCV_NativeCamera﹕ Supported Thumbnail Sizes: 0x0,320x240,240x320 
D/OpenCV_NativeCamera﹕ Supported Whitebalance Modes: auto,incandescent,fluorescent,warm-fluorescent,daylight,cloudy-daylight,shade,twilight 
D/OpenCV_NativeCamera﹕ Supported Effects: mono,negative,none,posterize,sepia,aqua,solarize,nv-vivid,nv-emboss 
D/OpenCV_NativeCamera﹕ Supported Scene Modes: auto,action,portrait,landscape,beach,fireworks,night,night-portrait,snow,sports,steadyphoto,sunset,theatre,barcode 
D/OpenCV_NativeCamera﹕ Supported Focus Modes: fixed 
D/OpenCV_NativeCamera﹕ Supported Antibanding Options: off,auto,50hz,60hz 
D/OpenCV_NativeCamera﹕ Supported Flash Modes: off,on,auto,torch,red-eye 
D/OpenCV_NativeCamera﹕ initCameraConnect: preview format is set to yuv420sp 
D/OpenCV_NativeCamera﹕ Starting preview 
D/OpenCV_NativeCamera﹕ Preview started successfully 

实施例的代码（不完全是，我使用相同的码）：

cv::Mat inframe = cv::Mat(); 
    cv::VideoCapture mCamera; 
    mCamera.open(1); 
    mCamera.set(CV_CAP_PROP_FRAME_WIDTH, 400); 
    mCamera.set(CV_CAP_PROP_FRAME_HEIGHT, 300); 

// mCamera.set("scene mode", "beach"); // <-- looking for a way 

    while (mCamera.isOpened()) { 
    bool grab = mCamera.grab(); 
    if (grab) { 
     mCamera.retrieve(inframe, CV_CAP_ANDROID_COLOR_FRAME_RGB); 

     // To do something for the iframe 

    } else { 
     LOGE("--->Cannot grab a frame from camera."); 
    } 
    } 
    mCamera.release(); 

Answer 1

唯一的问题是，不存在用于由Android NDK处理相机任何标准化API。 有必要通过Java，因为NDK访问将特定于特定的硬件。

虽然有VideoCapture类OpenCV中的C++接口，你可以使用它，如果您实现本地活动的做法，但由于NDK不从本地代码访问摄像机提供的API，所以VideoCapture使用无证接口，这使得VideoCapture不通用。它不适用于某些旧设备或新设备。只要看看OpenCV本机相机here上的问题。

我建议在管理端使用setSceneMode(String value)，您还可以发送从机端的请求管理，从而设定场景模式使得在管理方面的SCENE_MODE_#之间的转换，并在本机端对应的枚举（你应该实现自己）。

通过这里的方式是为Android相机的特性（VideoCapture的set()法）：videoio.hpp #L263 - #L325

---

更新＃1：

你可以尽量延长的功能OpenCV的camera_wrapper通过CameraParameters设置场景模式（如焦点或白平衡），但可能有一些Android版本之间的差异。

还要注意，您必须将androidcamera模块编译到所有Android版本才能使用它们。这就是为什么您在文件夹OPENCV_ANDROID_PATH\sdk\native\libs\armeabi-v7a下看到很多libnative_camera_r#.#.#.so。

Answer 2

经过多次测试，我发现答案之一就是从OpenCV退出。 这意味着直接控制V4L2。

https://github.com/openxc/android-webcam

http://www.linuxtv.org/downloads/legacy/video4linux/API/V4L2_API/spec-single/v4l2.html