Target Tracking with Glint Noise

Christer Johansson

Master Thesis, Report UPTEC F00064, August 2000.

Abstract:

Target tracking represents estimation of position, velocity and acceleration of a target. The estimation must handle different pertubations. One of the pertubations is glint noise. Glint noise occurs when a radar is used in target tracking because of interference between the refelected radar waves.

Glint noise is non-Gaussian which make the estimation more difficult. One of the most common target tracking methods today is the Kalman filter. This method assumes the pertubations to be gaussian but this will not be true for glint noise.

This report studies three methods, the Kalman filter, the Kalman smoother and the use of Layered Interacting Multiple Models (LIMM), for target tracking with glint noise. The glint noise is modelled in two different ways. One model is a statistical mixture between Gaussian noise and Laplacian noise. The second model assumes the target to be well described by six point reflectors, whereafter the pertubations are calculated. The behaviour of the methods is compared by using two different target tracks. The relative order of performance between the methods is independent of target track and the method that has the best performance is LIMM.

Company:

SAAB (Bofors Missiles)

Thesis Advisor:

Mikael Sternad