DEVELOPMENT OF SEQUENTIAL DECISION ALGORITHMS FOR TARGET DETECTION IN SURVEILLANCE RADARS

Abstract

This paper presents the application of sequential analysis theory to ensure target detection quality characterized by the probability of correct detection and false alarm rate at surveillance radar stations. Based on the analysis and evaluation of common uncertainty factors and corresponding mitigation strategies, several sequential decision-making algorithm variants are proposed to suit practical deployment conditions. The study shows that, under uncertainty in the signal-to-noise ratio, detection quality can be maintained through various approaches. However, when real-time computational feasibility becomes a critical requirement, one of two strategies may be selected depending on the operational objective. If the goal is to achieve average detection performance with minimal observation time, the strategy using an averaged likelihood ratio over the full sequence of observation cycles is appropriate. In contrast, when robustness against worst-case scenarios is needed, the approach employing the likelihood ratio associated with the nearest neighboring hypothesis is preferred. Compared to other commonly used uncertainty-handling approaches, this method features a simpler structure, significantly lower computational load, and only a marginal increase in decision time. Both approaches are applicable to existing surveillance radars in Vietnam.