OpenCV Android Poor results with camshift tracking

Question

I am trying to implement an object tracking application for Android using the camshift method. So far, i succeeded in implementing the camshift method and running it on my android tablet. Unfortunately, it does not work as intended... When I start up the application, the camera of the device is accessed and a piece of the image will be selected to be the tracking object. A histogram of the object is computed using the Imgproc.calcHist function.

Afterwards every time the camera gives a frame Imgproc.calcBackProject is used on the camera image und using the Imgproc.calcBackProject function the new location of the Object is computed. Unfortunately, instead of tracking the selected Object, the tracking drifts off in the upper left corner of the screen and stays there. Here is the releant code:

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;
}

Does anybody know how to make the tracking work? I saw camshaft videos that looked very promising, by my application is nowhere near as good as those :(