Artificial intelligence for helicopter safety: Head-pose estimation in the cockpit

This paper discusses a novel method for improving rotorcraft safety. A new algorithm is proposed for estimating the head position of helicopter pilots and copilots using onboard cockpit videos. Cockpit videos offer the ability for crash analysts or incident investigation to understand not only the aircraft state but also the pilot and copilot's actions in a potentially unsafe situation. In addition, head pose information can also be used to improve the overall scanning techniques used by pilots or to research which technologies can assist the pilots in focusing more attention out of cockpit, rather than down at the instrument panel. Two algorithms were created to provide possible solutions to the problem of head pose estimation: a hybrid computer vision algorithm that utilizes deep learning based detectors, and a purely deep learning algorithm. The purely deep learning algorithm was able to correctly classify 91.49% of copilot head positions in a real-world flight video.