Design of a constant false alarm rate (CFAR) detection scheme : using the sequential likelihood ratio test in a TBD approach

In this Master's Thesis the problem of variation in the probability of false alarm is treated, when testing the null hypothesis ``Target not present" against the alternative hypothesis ``Target present", using sequential likelihood ratio tests in noisy and clutter background environments. The hypothesis testing is performed with a Track Before Detect (TBD) approach, where the tracking is performed by a Sequential Markov Chain Monte Carlo (SMCMC) filter. Insurmountable problems with applying Wald's theory with filtered data are discussed. A new sequential detection scheme is designed that uses Neyman-Pearson theory in a sequential setting. Several models are proposed to control the probability of false alarm of this detection scheme by integrating these models in the threshold calculations. The models can be applied within a Gaussian noise environment. Furthermore, expressions are given to properly apply the models in a (sea) clutter. The results for applying the new sequential detection scheme, with integrated models, fulfill the desired Constant False Alarm Rate (CFAR) property in noise. Finally, a proposed implementation of the detection scheme in the SMCMC filter is given