Adaptive normalization of active sonar data

Abstract

In active sonar detection the background interference is reduced through normalization prior to applying a detection threshold. From the two mainstream normalization algorithms, i.e. the sliding window and the split window, the latter places guard bands around the point of interest to avoid a spill of the target's energy into the background calculation. The size of the guard bands has a strong effect on normalization: if it is too narrow, the target's energy biases the background estimation; if it is too large, the background estimate at one location is based on background samples too far away from that point. During recent sea trials strong changes in system performance, attributed to inappropriate normalization, have been experienced. An approach to overcome this problem by means of an adaptive normalization window that takes into account the geometry of the scenario and the environmental conditions (time spreading) is presented.