Abstract

Civilian and military aircraft face significant threats from passive missiles, such as short-range or within-visualrange air-to-air missiles (SRAAMs or WVRAAMs) and manportable air defense systems (MANPADS), which do not emit radio frequencies (RF) and thus evade detection by an aircraft's Radar Warning Receiver (RWR). In this paper, we present a deep learning-based passive missile detection algorithm utilizing simulated Solar Blind Ultraviolet (SBUV) signatures of missiles, which offer inherent advantages over infrared (IR) signatures. The proposed algorithm employs the YOLOv5 (you only look once) framework of Convolutional Neural Networks (CNNs) and moving object tracking to detect and classify the simulated passive missile UV signatures from a sequence of images. Data synthesis techniques, using 3D missile and aircraft combat scenario simulations in the SBUV spectrum, are applied to overcome the challenge of limited training data. Our findings demonstrate that the proposed algorithm effectively detects the simulated passive missile UV signatures, de-cluttering them from the background and classifying detected missiles as threatening or approaching threats. Consequently, this deep learning-based approach provides a promising approach for improving the detection and direction assessment of passive missile threats, ultimately enhancing aircraft safety and security.