LINE OF SIGHT ANALYSIS USING A FEEDFORWARD NEURAL NETWORK AND ONE-METER RESOLUTION DIGITAL ELEVATION MODEL (DEM) MAP DATA

Abstract

The importance of maximizing one's Line of Sight (LOS) while minimizing enemy LOS is of critical importance in war. LOS between an observer and a target exists if a straight-line vector between the observer and target is not intersected by terrain. Many sensors and kinetic or non-kinetic weapons and enablers require intervisibility between the shooter and target for employment. A means to analyze a terrain map and determine one's LOS would aid route planning onboard aircraft to minimize exposure to ground-based sensors. Furthermore, most LOS programs are computationally expensive to run at scale, making any such analysis on board small aircraft generally unavailable to analyze a large terrain set or to analyze many LOS vectors between formations of sensors/shooters and targets. An LOS machine-learning estimate may solve this problem by reducing computational time, allowing a large number of LOS calculations to be performed with relatively small computation resources found on a laptop. Rapid and computationally efficient LOS calculations would aid warfighters in either maximizing their LOS (such as for an anti-aircraft missile placement) or minimizing their LOS (such as for a vulnerable helicopter needing to hide from potential enemies). The goal of this work is to determine whether such a machine-learning model can reduce the computation time for a large set of LOS calculations as compared to traditional LOS calculation methods with minimal loss in accuracy.