Online De-Noising of Radar Data using Multi Resolution Analysis

Abstract

Target Tracking is an active research area, which encompasses various applications in Defence as well as Commercial applications. For estimating state vectors of tracked objects, Kalman filtering techniques are widely used, and the performance of Kalman filter depends on priory assumptions like state transition models and measurement uncertainties. In practical real time applications, all these priory assumptions are not available always and existing models are not suitable for target dynamics, which have an impact on the tracking quality, and some times filter, may diverge also. Recently Wavelet based multi resolution analysis has become a powerful tool, for image compression and de-noising applications and does not require explicit priory knowledge like Kalman filter for noise suppression. However, It is found that during real time de-noising, wavelet analysis exhibits poor performance due to certain artifacts. In order to improve the performance, a method is proposed and implemented that utilizes variable moving window and symmetric extension techniques.