Background noise distribution for noise object tracking in Track-Before-Detect systems using minimal chi-square statistic

Tracking of the noise signal in noise measurements needs special techniques. The difference between object and background noise is defined, using the noise distribution. The proposed technique is based on the model of the background noise. The window based approach is used for input signal preprocessing. The comparison of two distributions - empirical and observed is used. The global distribution is obtained using all measurements and the observed distribution is computed in the window area only. Minimal chi-square statistic is used as comparison criteria and results are processed by Spatio-Temporal Track-Before-Detect algorithm for tracking of the dynamic of the object and improved signal denoising. A few examples are shown for different objects that show possibilities of the proposed solution. Mean value suppression is possible using comparison of both distributions, what is important in application where the background estimation is not ideal.