1

我正在尝试使用 camshift 方法为 Android 实现对象跟踪应用程序。到目前为止，我成功地实现了 camshift 方法并在我的安卓平板电脑上运行它。不幸的是，它没有按预期工作......当我启动应用程序时，设备的摄像头被访问，并且将选择一张图像作为跟踪对象。使用该Imgproc.calcHist函数计算对象的直方图。

之后，每次相机给出一个帧时，Imgproc.calcBackProject都会在相机图像上使用并使用该Imgproc.calcBackProject函数计算对象的新位置。不幸的是，跟踪不是跟踪选定的对象，而是在屏幕的左上角漂移并停留在那里。以下是相关代码：

public Mat onCameraFrame(CvCameraViewFrame inputFrame) {

     //http://docs.opencv.org/3.1.0/db/df8/tutorial_py_meanshift.html#gsc.tab=0

     //take first frame of the video
    mCamera = inputFrame.rgba();

     if(objectTaken == false){
         // ich erstelle einen Beispielframe 
         trackingFrame = new Rect(mCamera.cols()/4,mCamera.rows()/4,600,500);

         System.out.println("SIZZEEE: "+trackingFrame.size().toString());
         System.out.println("Position: "+trackingFrame.x +" / "+ trackingFrame.y);

         object = mCamera.submat(trackingFrame.x, trackingFrame.x+trackingFrame.width, trackingFrame.y, trackingFrame.y+trackingFrame.height);
         hsvObject = new Mat();
        Imgproc.cvtColor(object, hsvObject, Imgproc.COLOR_BGR2HSV);


        lHSV = Arrays.asList(hsvObject);
        //MatOfInt mChannels = new MatOfInt(0, 1);
        mChannels = new MatOfInt(0);
        histogram = new Mat();
        histSize = new MatOfInt(180);
        ranges=new MatOfFloat(0,180);


        Imgproc.calcHist(lHSV, mChannels, new Mat(), histogram , histSize , ranges);
        //Core.normalize(histogram, histogram, 0, 255, Core.NORM_MINMAX, -1, new Mat());
        Core.normalize(histogram, histogram, 0, 255, Core.NORM_MINMAX);

        //return mCamera;
        objectTaken= true;

     }

     //set up the ROI for tracking
    // grab the ROI for the bounding box and convert it to the HSV color space
    //object = new Mat(mCamera,trackingFrame);

    backproj = new Mat();
    Imgproc.calcBackProject(lHSV, mChannels, histogram, backproj, ranges, 1);

    //box = Video.CamShift(backproj, trackingFrame, new TermCriteria(TermCriteria.MAX_ITER|TermCriteria.EPS, 50, 0.001));
    box = Video.CamShift(backproj, trackingFrame, new TermCriteria( TermCriteria.EPS | TermCriteria.COUNT, 10, 1 ));
    trackingFrame = box.boundingRect(); 
      pt1 =new Point(trackingFrame.x,trackingFrame.y);
      pt2 =new Point(trackingFrame.x+trackingFrame.width,trackingFrame.y+trackingFrame.height);
     RED = new Scalar(255,0,0);
    Imgproc.rectangle(mCamera, pt1, pt2, RED);
    return mCamera;
}

有人知道如何使跟踪工作吗？我看到看起来很有前途的凸轮轴视频，但我的应用程序远不如那些:(

4

0 回答 0