Use kalman in Java

edit

if you are using opencv2.4, i got bad news: the kalmanfilter is not usable.

by default, the internal 'measurementMatrix' variable is set to all 0, your measurement input will be multiplied by that, - you see it's all dead from here on, so the prediction will be always 0, too.

if you are using opencv3.0, you can set it e.g. to Mat.eye(2,1,CvType.CV_32F);

double[][] samplesArr = {{0,0},{1,0},{1,1},{2,1},{2,2}};

Mat meas = new Mat(2,1,CvType.CV_32F);

KalmanFilter kf = new KalmanFilter(4, 2);

Mat mm = Mat.eye(2,4,CvType.CV_32F);
kf.set_measurementMatrix(mm);  // 3.0 only!
// you'll want to set other Mat's like errorCovPost and processNoiseCov, too,
// to change the 'adaption speed'

for(int i = 0; i < samplesArr.length; i++)
{
    meas.put(0, 0, samplesArr[i]);
    Mat corr = kf.correct(meas);
    Mat pre = kf.predict();
    System.out.println(pre.t().dump());
}

 [java] [0, 0, 0, 0]
 [java] [0.5, 0, 0, 0]
 [java] [0.80000001, 0.60000002, 0, 0]
 [java] [1.5384616, 0.84615386, 0, 0]
 [java] [1.8235294, 1.5588235, 0, 0]