Status on Video Data Reduction and Air Delivery Payload Pose Estimation

Abstract

The paper discloses a current status of the development and evaluation of an autonomous payload tracking capability for determining time, state and attitude information (TSPI) for all types of airdrop loads. This automated capability of accurately acquiring TSPI data is supposed to radically reduce the labor time and eliminate man-in-the-loop errors. The paper starts with the problem formulation and then reviews the developed software. The algorithms rely on commercial off-the-shelf feature-based video-data processing software adopted for obtaining TSPI. Having the frame coordinates of the centroid of a tracking item available from no less than three ground stationary surveyed cameras at each instant during a descent together with these cameras azimuth and elevation information allows solving the position estimation problem. If more known-geometry features of the airdrop load can reliably be extracted, the pose (position and attitude) estimation would also be possible. The paper primarily addresses the status of the payloadﾒs position estimation portion providing examples of processing video data from up to six cameras. Yet, it also discusses the applicability of more recent computer-vision algorithm, based on establishing and tracking multiple scale-invariant keypoints. The paper ends with conclusions and suggestions for the further development.