Robotic Perception System for Casualty Pose Mapping

In response to an Army need for a human casualty perception system to be mounted on medical drones, Areté Associates proposes the development of AHEAD: the Areté Human Emergency Analysis and Determination system. The AHEAD system is a modular, JAUS-compliant unit consisting of a commercial, off-the-shelf (COTS) electro-optic (EO)-modality sensor, a spotlight for night and low-light operability, and a low-SWaP embedded processor. The core perception and machine vision algorithm is designed such that the three visual tasks—human identification, pose mapping, and surface volume mapping—complement each other, with each task increasing the accuracy of the subsequent one. The core element of Areté’s perception algorithm is the use of deep learning for human pose mapping. Areté plans to leverage DeepPose, a neural network designed for human pose mapping [4]. The pose estimated by the neural network will be used as a prior for the surface volume mapping using structure from motion, increasing the efficiency and accuracy of the solution.