Missile attitude in flight can be difficult to determine from ground-based images due to image resolution, lighting, object occlusion, and poor contrast. To address the need, UHV and UTARI propose a fusion based approach of 2 well proven methods to solve this problem. First, UHV Technologies will implement their existing innovative and state of the art machine learning methods for image segmentation of missiles during launch. These methods originated from an Advanced Research Project Association in the Department of Energy (ARPA-E) project to recycle metals, and was then improved by a United States Air Force (USAF) award to perform image segmentation based on UAV based aerial image data. Second, University of Texas at Arlington Research Institute (UTARI) will then take the segmented image data and then perform a feature-based pose estimation to determine the missile attitude using 3D alignment with a prior geometry database. The anticipated benefits include novel fusion-based approach deep learning algorithms to address this challenge, an image processing toolkit suitable for inclusion in current government owned analysis tools, and additionally commercial potential to address existing needs for low cost wide area ISR applications.
