Object Recognition to Refine Drone Positioning

摘要

This article is about using object recognition to more accurately position drones. A description is given of using an autopilot with a Kalman filter to reduce the error in positioning the drone. The use of a quaternion for positioning a drone is described. The implementation of the marker recognition module based on the calculation of moments using the OpenCV library and the Python programming language is described, which allows to find the offset relative to the marker, as well as allowing to calculate the positioning error over time. This allows you to synchronize the data received from the camera and autopilot to reduce the error in positioning the drone.