Performance of Fusion Algorithm for Active Sonar Target Detection in Underwater Acoustic Reverberation Environment

Abstract

Classically, automatic detection of targets in active sonar system is addressed by a matched filter processing followed by a constant False Alarm Rate (CFAR) thresholding method. Even though, various CFAR techniques viz. CA CFAR, GO CFAR and SO CFAR etc. are available in literature, none of them alone is sufficient to eliminate the false echoes. In certain applications, such as active sonar where the probability of false alarm pfa requirements are very stringent, the performance of CFAR alone cannot be used as detection criteria. Further, the choice of a particular CFAR algorithm is also a complex task, as the non-homogenous nature of the acoustic medium is difficult to predict. In this paper, a fusion algorithm is proposed for active sonar application where, in addition to CFAR technique, a support vector machines (SVM) based classification algorithm is also used to eliminate the false echoes. The performance of the algorithm is verified using practically measured data.