Electro-optical (EO) systems with digital image processing and computer-aided detection are increasingly coming into use for maritime surveillance, reconnaissance and search and rescue. EO systems have the potential to improve the consistency of detection, reduce operator workload and fatigue, and improve search efficiency. However, quantifying their performance versus more traditional approaches is problematic, because of the differences in how performance is specified for traditional systems versus modern computer-aided designs. In maritime search applications, system performance is commonly specified in terms of the lateral range curve (LRC). The LRC is a plot of the probability of detection versus horizontal range from the search platform. This metric has a long history, rooted in visual searches by trained human observers. However, it is specified without reference to any false-alarm rate or probability of false alarm. Computer-aided EO performance, on the other hand, is usually specified in terms of Signal-to-Noise Ratio (SNR), Receiver Operating Characteristic (ROC) curve, or some equivalent metric. In this paper, we demonstrate a methodology for estimating LRCs from SNRs or ROC curves. This methodology provides a consistent, quantifiable means for comparing the performance of new and legacy systems.
This investigation centered on the most challenging search and rescue requirements: finding small targets in high seas. Our course of action was to investigate the capabilities of known hyperspectral and LWIR sensors in realistic conditions of target and environment to drive the design of a sensor system capable of substantially improving search efficiency and efficacy for these conditions. The relevant results from this study include demonstration of significant power in clutter rejection (e.g., whitewater and wave complexity) by the LWIR sensor. In addition, several factors combine to indicate that a modest implementation of HSI and IR sensors would provide significant improvement in search efficiency and efficacy for small targets in high seas. These factors include high PD, low PFA, and the untiring nature of the sensors when combined with the potential of real-time automatic target/background discrimination. This modest implementation would translate directly into faster, more complete coverage, at lower overall costs to the USCG, and a more likely probability of a successful search mission.