Nouvelles méthodes pour l’estimation des paramètres des distributions gaussiennes composées.

Abstract

RADAR (Radio Detection and Ranging) is a detection system that uses radio waves to determine the distance (ranging) angle or velocity of objects. Two tasks are very important in radar system; Parameters Estimation and Detection of objects. The main task in radar system is to detect an object in homogeneous and non-homogeneous environment. For that, this thesis proposes two contributions which are Parameters Estimation and Detection in Compound Gaussian clutter which accurately describes the behavior of the intensive sea echo ""Clutter"", in different s ituations. The first contribution proposed in this thesis is the estimation of the parameters of the Compound Inverse Gaussian (CIG) distribution which is described by the shape and scale parameters in absence of thermal noise and the results are compared with other estimation methods such as Method of Moments (MoM), Non-Integer Order Moments (NIOM), method of [zlog(z)] and the Maximum Likelihood Estimation (MLE). The proposed method consists of using an iterative procedure the MLE equations using two different methods are used; the MoM which is used as initial points and the MLE which is used to obtain the two equations of the shape and scale parameters for K iterations. After simulations, the results show that the IMLE has the best performance compared with MoM, NIOM, and [zlog(z)] and has similar estimation performance than the MLE method but requires lower computational time. The second contribution concerns the pplication of the Bayesian Approach to build a CFAR Detector in Pearson V distributed clutter in homogeneous and non-homogeneous environment. The obtained results are compared with Constant False Alarm Rate (CFAR) and its variants as Cell-Averaging CFAR (CA-CFAR), Greatest Ordered CFAR (GO-CFAR), Smallest Ordered CFAR (SO-CFAR) and Ordered Statistic CFAR (OS-CFAR). A new expression of the Probability of False Alarm (pfa) is derived which is different from that of the Neyman-Pearson a pproach. After that we proved that the Bayesian Approach yields a CFAR detector. Then, we use Monte- Carlo method to calculate pd of these detectors. The simulation results show that the Bayesian CFAR detector has the best performance compared with other detectors in different situations even in the presence of high level interferences.