Vision-Based Object Detection and Localization for Autonomous Airborne Payload Delivery

This paper follows the development of a vision–based object detection and localization system for implementation in an autonomous aircraft for payload delivery. Application of such a system could see use in the delivery of packages to offshore freighter ships and inaccessible inland areas. This system was developed in a modular fashion, such that it could be interchanged and adapted between various airframes. This system comprised of three core elements, namely autonomous flight control, vision–based object detection and localization and, payload release and delivery modelling. The final integrated system was tested and able to achieve fully autonomous flight, and simultaneously model a payload release trajectory from an altitude of 75 m to deliver the given payload with an average displacement of 1.8 m of the designated drop-zone. The drop-zone location was determined via the onboard vision system through the implementation of an object detection and localization algorithm.