Selfie Drone V 2.0

Abstract

Computer vision/Image processing has become a vast field. It has enabled us to automate many everyday tasks, no matter how mundane or extraordinary. This project aims to use image processing to localize the UAV using pose estimation and capture a user's picture. The UAV will rise a few feet before the user. An image will be selected from an existing pose model and the UAV will move autonomously and capture a selfie depending on it. There will be involved some pose estimation algorithms that will create a relation between drone’s space coordinates and the different link length ratios of the subject. Mediapipe pose will be used for pose estimation, and ROS will be used for the programming. After the processing, the drone will place itself into a position that best matches the template selected by the user. The project will be supported by a pixhawk2 flight controller.