Passive Image Processing Algorithms for Automated Target Attitude Estimation

Weapons systems testing and analysis is crucial to engineering, understanding, qualifying, and deploying advanced weapons systems. With the boom in camera technology over the past decade image based methods to support weapons testing and analysis have become the state of practice for military test ranges. At the same time, there has been an explosion of research and development in computer vision algorithms that have demonstrated state-of-the-art performance on previously intractable problems. Despite this research, the weapons testing community has not significantly benefitted from these advances. Existing capabilities of rely largely on input from human analysts, pristine environmental collection conditions, and/or complex workflows for data collection and analysis. To provide a more automated, accurate and robust capability research and development in the area of automated computer vision algorithms for target attitude estimation are needed. To develop such a capability, Toyon and CU Boulder propose the research and development of a passive image processing algorithm for automated target attitude (PIP AATA) estimation that leverages both historically demonstrated statistical pose estimation in conjunction with state-of-the-art deep learning pose estimation techniques. The research will balance robust computer vision capabilities with statistical reliability to enable automated target position and attitude estimation with corresponding uncertainty estimates.