Perception Testing for Maritime Surveillance

This paper describes a perception study designed to measure the capability of human observers to perform uncued detection of a ship in simulated video imagery. We generated a set of video clips depicting a surface ship in calm water at various ranges, levels of visual contrast, and camera panning rates (which are related to the time the targets remained on screen). Using custom designed test software, we presented these videos to a group of active duty military service members and also to a demographic proxy group, and asked the test subjects to indicate for each clip if it did or did not contain a single target of unknown range and contrast. A fraction of the clips contained no target at all, so that the subjects' false detection rates could be estimated and individually compensated. The test subjects' true positive fractions were compensated for false alarms, averaged into cohorts of similar levels of experience interpreting surveillance video, and correlated against contact size, contrast, and time on screen. The results indicate the relative importance of the parameters varied: target size in pixels, contrast of the target with the background, time on screen, and operator experience with surveillance video. Based on these results, we derived a set of logistic curves, and used them to determine V-50 and E values suitable for use with the U.S. Army Night Vision Electronic Systems Directorate NV-IPM package. We also developed a parametric model for the effect of limited view time. Through use of NV-IPM, the results of this study can enable more accurate predictions of human detection capability against maritime targets in more general situations.