Constructor Detail

• CVMKalmanFilter

  public CVMKalmanFilter(double[] X0,
                         double initStateCovariance,
                         double positionProcessStd,
                         double velocityProcessStd,
                         double positionMeasurementStd)

  Initialize a new Kalman filter with the specified initial state.

  Parameters:

  X0 - initial state estimate. Must a 6 elements double[] array with x0, y0, z0, vx0, vy0, vz0 with velocity in length/frame units.

  initStateCovariance - the initial state covariance. Give it a large value if you do not trust the initial state estimate (e.g. 100), a small value otherwise (e.g.1e-2).

  positionProcessStd - the std of the additive white gaussian noise affecting the position evolution. Large values means that the position undergoes heavy fluctuations.

  velocityProcessStd - the std of the additive white gaussian noise affecting the velocity evolution. Careful, we expect it to be in units of length/frame. Large values means that the velocity undergoes heavy fluctuations.

  positionMeasurementStd - the std of the additive white gaussian noise affecting the position measurement. Large values means that the positions measured are not accurate.

Method Detail

• predict

  public double[] predict()

  Runs the prediction step of the Kalman filter and returns the state predicted by the evolution process.

  Returns:

  a new double[] of 6 elements containing the predicted state: x, y, z, vx, vy, vz with velocity in length/frame units.

• update

  public void update(double[] Xm)

  Runs the update step of the Kalman filter based on the specified measurement.

  Parameters:

  Xm - the measured position, must be specified as a 3 elements double[]array, containing the measured x, y, z position. It can be null; the filter then assumes an occlusion occurred, and update its state based on solely the prediction step.

• getPositionError

  public double getPositionError()

  Return the root mean square error on position estimated through the state covariance matrix.

  Returns:

  the estimated error on position.

• getVelocityError

  public double getVelocityError()

  Return the root mean square error on velocity estimated through the state covariance matrix.

  Returns:

  the estimated error on velocity, in length/frame units.

• getNOcclusion

  public int getNOcclusion()

  Returns the number of occlusion events that occurred since the instantiation of this filter.

  Returns:

  the number of occlusions.