Stochastic Radiative Transfer Model for Contaminated Rough Surfaces: a Framework for Detection System Design

We developed a framework to evaluate the performance of a detection system for contaminated surfaces. We employ the radiative transfer model for contaminated surfaces (Ben-David and Davidson, ECBC-TR-1084, 2013) and transform the physical model into a stochastic probability model with which detection probability and false alarms can be estimated for scenarios of interest. Our algorithm employs a data fusion approach known as a distributed binary integration system (also known as a double-threshold detector, or m-out-of-n detector) in order to combine the individual detection results from multiple scans over several potentially contaminated areas. With our probability model we can explore the parameter space (e.g., number of measurements, time to detect, area to monitor, sparsity of the contamination, field of view, etc.) and study the tradeoffs between parameters that affect the overall system detection performance. We can also use the stochastic model to set sensor requirements for a contamination scenario. We presented plots that demonstrate the interaction between parameters and an example for the detection of a potassium chlorate contaminated “car” with a CO2 tunable laser system. Subject Terms Radiative transfer, contaminated surfaces, detection, rough surface, BRDF, reflectance, fill factor, distributed binary integration, CFAR, data fusion, distributed sensor system, double-threshold detector, m-out-of-n detector, potassium chlorate, probability theory, system performance, probability of detection and false alarm.