Autonomous Target Tracking Of A Quadrotor UAV Using Monocular Visual-Inertial Odometry

Unmanned Aerial Vehicle (UAV) has been finding its ways into different applications. Hence, recent years witness extensive research towards achieving higher autonomy in UAV. Computer Vision (CV) algorithms replace Global Navigation Satellite System (GNSS), which is not reliable when the weather is bad, inside buildings or at secluded areas in performing real-time pose estimation. Thecontroller later uses the pose to navigate the UAV. This project presents a simulation of UAV, in MATLAB & SIMULINK, capable of autonomously detecting and tracking a designed visual marker. Referring to and improving the state-of-the-art CV algorithms, there is a newly formulated approach to detect the designed visual marker. The combination of data from the monocular camera with that from Inertial Measurement Unit (IMU) and sonar sensor enables the pose estimation of the UAV relative to the designed visual marker. A Proportional-Integral-Derivative (PID) controller later uses the pose of the UAV to navigate itself to be always following the target of interest