When a radar is used for tracking, targets can be unresolved if they are close to each other or if a target is close to the sea surface it can be unresolved from its reflection. When targets are unresolved, conventional monopulse processing extracts a single “merged” measurement depending on the targets' relative position and their relative signal strength with possibly very large direction of arrival (DOA) errors. Such targets can be resolved only by using a detection algorithm that can differentiate between single and merged measurements and then employing a special signal processing suitable for the merged measurement case.; If two targets are present in the same resolution cell, conventional radar signal processing algorithms can extract only a single set of merged measurements, i.e., one elevation and one bearing angle (even though there are two targets). A new signal processing approach using the maximum likelihood (ML) criterion is presented. The ML extractor is developed for the Swerling I model as well as Swerling III model of radar cross section (RCS) fluctuation of the unresolved targets. For the first type of targets an analytic expression for the estimated angular position of the targets as well as the Cramer Rao lower bound (CRLB) on estimation error standard deviation are obtained. Detection of the presence of unresolved measurements is very important as it tells the tracking algorithm whether to use the special signal processing or not. In this research, a generalized likelihood ratio test (GLRT) detector is developed for this purpose. The angle estimation and detection algorithms are compared with an existing ‘superresolution’ algorithm.; Extraction of the elevation angle of targets in the presence of sea-surface induced multipath is a difficult problem due to the presence of diffuse and specular reflection components in the radar returns. These returns follow from the direct as well as three different reflected paths. The state of the art monopulse ratio based angle estimation produces a negative bias (sometimes very high, depending on the scenario). A modified ML elevation angle extractor is developed for the sea-surface induced multipath. This angle extractor reduces the root mean square error (RMSE) of estimated angle by a factor of 2–3.; In the case of two unresolved targets, the effect of using the measurements for the interfering (secondary) target is studied. For targets in the presence of sea-surface induced multipath a hybrid ML technique is developed that switches the signal processor to one of the two ML elevation angle extractors depending on the ratio between the observed signal to noise ratio (SNR) to the tracker predicted SNR. This results in a further 20% reduction of RMSE. Finally, the signal processing algorithms are applied to the most realistic tracking and radar scheduling benchmark problem to date. The tracking performances are evaluated and compared with the tracker using the state of the art signal processing techniques.
展开▼
