New approaches to improved detection and tracking performance in radars

Abstract

The purpose of a Track While Scan (TWS) radar system is to detect, locate and track the various targets under its surveillance in the presence of interfering echoes. In the conventional approaches, the two basic functions in a TWS radar-detection and tracking-are viewed and designed independently. This thesis presents new approaches to improved detection and tracking performance in a TWS radar by integrating these two functions. First, the problem of tracking a single manoeuvring target in clutter is considered. The inadequacy of the existing framework in handling two problems-optimisation of decision threshold and tracking a manoeuvring target-is first brought out. A new framework is then presented to solve these problems. As a result, the tracking performance of weak targets is significantly improved and tracking a manoeuvring target in clutter becomes feasible. A novel and effective false track deletion criterion is also presented. This algorithm can be used to track multiple targets provided they are far apart. It is, however, not suitable for tracking multiple targets flying close to each other or crossing, and it is also not suitable for automatic track initiation. A new multi target tracking algorithm is then developed, possessing all the required features. Each of the existing algorithms caters only to a subset of the requirements. As a detection oriented algorithm, it shows how the a priori information provided by the track update filter can be used to improve detection characteristics. We then derive analytical formulas to indicate how the performance of a TWS radar system can be evaluated a priori. These equations indicate what choices are available for the designer to meet certain performance specifications relating to true track initiation, true track continuation, false track initiation and false track deletion. Although certain studies of this nature have been published, they deal with only a subset of desired characteristics and also ignore certain key aspects. Implementation aspects of this algorithm are then highlighted. A multi microprocessor architecture has been identified to make the algorithm viable for real time implementation. We then present a comprehensive solution to the combinatorial problems in multi target tracking. We demonstrate how, by using a search procedure different from gradient methods, one can locate N globally best hypotheses. The multi target tracking algorithm developed earlier has certain features that permit this solution. Finally, a modified version of an existing distribution free CFAR processor is presented to improve its performance in clutter edges and multi target situations.