SEARCH

SEARCH BY CITATION

Keywords:

  • sidescan sonars;
  • detection probabilities;
  • binning;
  • logistic regression;
  • cubic splines

Abstract

Sidescan sonars are used in detecting hostile targets on the seabed. An accurate representation of the sensor performance is required to carefully plan and execute searches. We studied several confidence interval methods for assessing the accuracy of detection performance as a function of range from hostile targets. Using simulation, we determined that the best approach for modeling sensor data was piecewise logistic regression with cubic B-splines at carefully selected knots. This method was demonstrated on real data collected from the CITADEL sea trial. © 2012 Wiley Periodicals, Inc. Naval Research Logistics, 2012