Adaptive target detection with multipath exploitation in heterogeneous environments

Adaptive detection of point like targets have been studied in detail during last decades. Most of these studies assume that the target steering vector is perfectly known. This assumption may be acceptable when the signal reflected from a target has only line of sight component. However, interactions between the signal backscattered from the target and the environment causes the signal to have multipath components. In this case, the conventional adaptive detectors which assume that the target steering vector is perfectly known may not be able to detect the target properly. In addition to the studies about target signal modelling, various studies have been carried out so far about modeling the noise signal. In this respect, homogeneous model is widely used in adaptive target detection algorithms where Gaussian distribution is used as the probability density function of noise. However, the theoretical studies and the measurements with high-resolution radars show that the probability density function of the noise deviates from the Gaussian distribution. In this thesis, adaptive detection of point-like targets is considered in the presence of multipath effect under the assumption of partially homogeneous and heterogeneous environments. Therefore, in the design of an adaptive detector to operate in partially homogeneous environments, an unknown scaling factor is used to represent the mismatch between the noise covariance matrices of test and training signals. In this scenario, it is assumed that each member of the secondary data set have the same spectral properties. On the other hand, in the design of an adaptive detector to operate in heterogeneous environments, the estimation of the covariance matrix in the cell under test carried out assuming that every individual member of the secondary data set have different noise scaling factors. Besides, the target echo is modelled as the superposition of direct and multipath components where multipath echoes are thought of as scattered signals from a glistening surface. Hence, the total multipath return is represented as a Gaussian distributed random vector. Target detection algorithm is constructed with GLRT method which is used for finding the maximum likelihood estimates of unknown parameters. The algorithm assumes that the total primary data covariance structure, in the target present case, resembles to the covariance matrix obtained from secondary data up to a degree (related to noise scaling factor and multipath contribution). The results highlight that the new detectors copes well with severe multipath conditions in the presence of non-homogeneous environments.