The goal is to implement a vision-based 3D pose estimator which estimates the position, orientation, linear and angular velocities of a Nano+ quadcopter. The drone is made to fly over a mat of April tags and the data relating to the image is provided.
The project initially involves implementation of projective transformation-based position and orientation estimation solved using singular value decomposition, based on the given image data where the method of Randon Sample Consensus(RANSAC) is implemented for the rejection of outliers while calculating velocities using optical flow
To execute this, clone this repo and open the directory within MATLAB.
- Run poseEstimation.m to obtain position and orientation of the MAV using a visual fiducial system
- Run OpticalFlow.m to obtain the velocities calculated using optical flow and further the outlier rejection